The Intricate Beauty of Pecking Orders

Everything you think about pecking orders in animals is probably wrong

Most people view pecking orders in animals as rigid hierarchical arrangements based upon differences in individual ability. The superior individuals – the strongest, most aggressive, and genetically gifted – are at the top, and those less able – the weak, the timid, and genetically inadequate – are at the bottom. Common sense tells us that this must be true. But my research of nearly fifty years suggests this view is a distortion of reality. Instead, my research proposes that pecking orders emerge from a continuous intricate and beautiful dance of interactions among the individuals in groups. Rank orders arise from the dynamics of behavior, not the qualities of individuals. If I am correct, the individual-based view inhibits progress in the study of pecking orders. But it goes beyond that. Many people use the same view to explain their own or other individuals’ ranking in society – that it is down to their intelligence, grit, confidence, work ethic, social connections, or more perniciously to their ethnicity, gender, or genetic background. Out of all the different kinds of social organization in animals, pecking orders seem to have the best chance of being explained by individual differences. But if the individual-based view is incorrect for pecking orders, I suggest that both as scientists and lay people, we need to change our views of hierarchy, status, and rank in human societies.

In this essay in the form of a dialogue I will do three things. First, I will talk about why the depiction of pecking orders in contemporary research is distorted and the experimental and theoretical evidence against that view. Second, I will present a more accurate view of pecking orders based upon the latest experimental evidence. This view, as noted above, is that the dynamics of aggressive behavior in groups, instead of differences among individuals, produce the rank structure of pecking orders. Third, I will sketch out some ideas about how these behavioral dynamics lead to the formation and maintenance of pecking orders.

I will explain my assertions through a talk I had with my friend Peter. This allows a less technical, and I hope more enjoyable discussion, than a formal academic essay. The informality of this talk also lets me make a few digressions, for example, about how contemporary social and behavioral science supports and rewards research that presents an inaccurate picture of pecking orders and other kinds of social organization. Peter is a lawyer who has partially retired from a successful career in New York. One of his specialties is intellectual property law, and he has worked with many well-known artists and musicians to protect unauthorized use of their music and art. He likes creative people and wants to know how their minds work. Peter and I get together from time to time, and unlike most non-scientists I know, Peter usually asks me a few thoughtful questions about my research. He now spends most of his time in the Hamptons on Long Island in New York where he supplies much of the food he and his wife need by cultivating a large garden and fishing several days a week. He recently decided to get a few chickens for eggs and entertainment. The antics of the chickens captivated Peter, and he began making notes on their behavior. Like a good lawyer, he tried to figure out what his “clients” were doing, and he developed a theory. One day last summer, he called me to talk about what he had come up with.

Peter: I think I just observed one of those pecking orders that you study. When I watched my chickens for a while, I saw that one of them, Clucky, pecked all the others and no one pecked her in return. Another chicken, Veronica, pecked everyone but Clucky. Annabella, my third chicken, only pecked Gordo, the fourth chicken, and Gordo didn’t peck anyone. I also noticed that a couple of times after Clucky had pecked Gordo, Veronica would come over and peck Gordo too, like Veronica was using Clucky’s attack to take advantage of Gordo. And once, after Veronica pecked and chased Gordo and Annabella, she – Veronica – went over and attempted to peck Clucky. She wasn’t successful, but it looked like she gained some, I don’t know, maybe confidence, after she was aggressive toward Annabella and Gordo. I can hardly believe I’m talking about what might be going on in the heads of chickens.

Linear Pecking Orders

Ivan: You have indeed observed a pecking order, or to be a bit more formal, a dominance hierarchy, in your chickens. In particular, you have observed a hierarchy with the classical “linear” structure. Pecking orders like yours are called linear because the individuals can be ranked in a straight line from top to bottom by the number of other individuals that they dominate. In your chickens, as I’m sure you realize, the ranking starting from the top was: Clucky, Veronica, Annabella, and Gordo. Animal behaviorists have reported linear pecking orders in many, many species – some insects, reptiles, and crustaceans, and many fish, birds, and mammals, including human children and adolescents. You may not be surprised that you observed a linear pecking order, but, in a way, finding a linear arrangement is remarkable. If the relationships between your chickens had formed by chance – if we flipped a coin to see who dominated whom – the odds of getting a linear arrangement would have been only moderate. There must have been some forces at work to make your pecking order linear rather than non-linear.

Peter: You’re right. When you said I had a linear pecking order, I didn’t think it was anything special. I just assumed that pecking orders were always linear – even though I didn’t know the term. Maybe this has been figured out a long time ago, but what are the forces, as you call them, that make pecking orders linear? Could they be caused by differences among the chickens? Like differences in aggressiveness? Or genetics? Or maybe even in their past experiences? 

The Prior Attributes Hypothesis

Ivan: What you are proposing is known as the “prior attributes” hypothesis. The idea, as you suggested, is that differences in dominance ability among the chickens determine their ranks in the dominance hierarchy. The chickens could differ in genetic constitution, physiology, personality, or even in ways that we aren’t aware of yet. The prior attributes hypothesis proposes that you observed a linear pecking order because Clucky was the highest on dominance ability before you assembled your flock and then she gained the top rank in the hierarchy, Veronica was the second highest on dominance ability and she took the second-to-the-top rank, and so on.

            Animal behaviorists have carried out research using this hypothesis for decades. They still use it today, and the idea behind the hypothesis has made its way into our thoughts about success in sports, business, and life more generally.

But even though the ranks of individuals on certain prior attributes are correlated with their ranks in pecking orders, this hypothesis is not an accurate explanation of why we have linear arrangements. Let me give you an analogy. You’re a sports fan, and you know it’s almost impossible to precisely predict the rankings of professional baseball and basketball teams in the final playoffs, given what we knew about them at the beginning of the season. Predicting a final outcome in sports from earlier information is a version of the prior attributes hypothesis. We can’t know how the rankings will shape up.

Besides being incorrect, this hypothesis prohibits us from seeing how intricate and beautiful the process of pecking order formation is. Sure, animals grubbing out their places in the hierarchy by pecking, hitting, threatening, and biting may not be beautiful per se. But here’s the beauty of it: I have become convinced that pecking orders emerge spontaneously and unexpectedly from the ever-changing behavioral dynamics in groups and that these dynamics are not simple expressions of differences in the abilities of individuals. Does that make sense?

Peter: I hear what you’re saying, but I find it hard to believe. Can you elaborate? You just said that the ranks of individuals on prior attributes are correlated with their positions in hierarchies – isn’t that enough to mean that the prior attribute hypothesis is correct?

Ivan: In the prior attributes hypothesis, the individual scoring highest on attributes must become the alpha in the pecking order, the one scoring second highest must become the beta, and so on. This result is logically true no matter what kinds of attributes or groups of attributes are involved; they can be physical, genetic, cognitive, physiological, or even personality factors. If the correlation between rank on attributes and rank in the pecking order is not perfect, then, the prior attributes formulation predicts that many non-linear pecking orders would result – something that we think does not often happen in small groups of animals. Researchers have measured the correlations between many kinds of attributes in animals and their subsequent ranks in hierarchies. These correlations are, as far as I am aware, always in the more moderate range and do not approach the level of total accuracy required by the prior attributes hypothesis.

If the prior attributes hypothesis is not correct, then not only are dominance researchers looking for something they can never find, but the way some popular commenters explain success in many areas such as in attracting mates, gaining top jobs in business, and becoming popular in groups is based upon a scientific delusion. You probably haven’t heard of him, Peter, but Jordan Peterson, a well-known pop psychologist, is one of those commenters. He says human society is one giant “pecking order.” He declares that the dominance hierarchy in life is both permanent and real. Those who have the individual qualities that place them in the top ranks will experience lives of success and those that do not will have abysmal lives. Sadly, many people use such misconceptions in evaluating their own success or lack of it in their lives.

Peter: Yes, you’re right. I haven’t heard of Jordan Peterson. But I do feel that the dominance hierarchy of my chickens seems permanent and real. It’s not that I don’t believe what you just said about perfect correlations being necessary for the prior attributes hypothesis to be correct, but I’m not very mathematical, and so arguments like that don’t feel intuitively persuasive to me. So tell me, is there some other evidence that suggests the prior attributes hypothesis is not correct?

Testing the Prior Attributes Hypothesis

Ivan: How about this: Let’s say we allow a group of animals to establish a pecking order, then we wipe out their memories of whom they dominated in the hierarchy, but don’t change them in any other way, and finally let them form a second pecking order. In this experiment, we would be ‘‘rewinding the tape’’ of the individuals to the extent possible and letting them form a second pecking order as if they were starting from scratch. If rank on prior attributes determined rank in the first pecking order, then rank on the attributes should also do so in the second one. In other words, we would expect identical rankings for individuals in both hierarchies in a large percentage of groups. If, however, rank on the prior attributes did not determine pecking order rank, the first and second pecking orders should not be identical, and a considerable proportion of individuals should change ranks between them. Do you follow me, so far?

Peter: Yep, so far.

Ivan: But how large should the proportion of groups with identical hierarchies be to support the prior attributes hypothesis? It would be too much to expect that 100% of the groups had identical first and second pecking orders, but how about a more moderate number like 75%?

My colleagues and I actually carried out this experiment with groups of cichlid fish. Like chickens, these fish readily establish pecking orders in the lab. We observed groups forming their first pecking orders, and then we separated the individuals in the groups for two weeks. Two weeks of separation are sufficient for fish of the species we used to forget their previous dominance relationships. When we reassembled the groups to form second hierarchies, we found that only about 27% of the groups had identical first and second pecking orders – very different from the 75% we thought would indicate strong support for the prior attributes hypothesis. And in one dramatic instance, the bottom-ranked fish in the first pecking order in one group rose to the top rank in the second pecking order, and the alpha fish in the first pecking order fell to the bottom rank in the second pecking order.

Now, let’s consider people, not fish. It is notoriously difficult in general to predict how well individuals will do in a new situation given information about their performance in an earlier setting. Think about the difficulty of making good hiring decisions for top positions in corporations and recruiting people for sports teams. For example, the Harvard Business Review reports that large companies fail to choose managerial candidates with the right mix of talents for their positions over 80% of the time. And researchers find only a weak correlation between the evaluation of college football quarterbacks when drafted and their subsequent performance on National Football League teams. (Footnotes to this research)

So, that’s the mathematical and experimental evidence for why the prior attributes hypothesis fails in accounting for linear pecking orders. But let’s go a little deeper and consider the basic assumption upon which the prior attributes hypothesis rests – it’s that people and animals are like physical objects, that they aren’t affected by changing situations, their recent experiences, or the composition of groups. Both common sense and abundant scientific evidence argue against this assumption. While it’s easy to understand why early research on pecking orders used the prior attributes hypothesis, I think it’s a failure of scientific imagination and nerve to continue to do so. At this point, reliance on the prior attributes hypothesis has become a lazy habit that researchers need to leave behind.

The results of our experiment forming and reforming hierarchies indicated that although there was some influence of a fish’s rank in the first pecking order on its position in the second, that influence was relatively small. But here’s the interesting thing. Even though the contribution of differences in prior abilities seemed weak, over 90% of the first and second hierarchies were still linear. That was something we didn’t expect, but there it was in our findings. Clearly, something else was working to create linear structures besides simple rankings on attributes.  

Peter: That’s really interesting, Ivan. I thought the reason why linear hierarchies were so common would indeed be simple. Is this lack of a simple explanation what makes you say that hierarchy formation is intricate and beautiful? Can you tell me more about that?

Ivan: Well, it might be too early for me to talk about why I see pecking orders as intricate and beautiful. I don’t want to get too ahead of myself.  But I will say that their predominantly linear quality is part of my fascination; pecking order formation is mysterious. For the time being, let me say that I think the secret to understanding where linearity comes from lies in observing how individuals interact when they establish groups. Somehow, linearity emerges out of the dynamics of interaction as individuals form groups.

Peter: That’s not something I would have thought of. What makes you think the linearity comes from the ways individuals interact as they set up pecking orders?

Social Dynamics and the Formation of Pecking Orders

Ivan: In the study in which we formed and then reformed pecking orders, we also did a second experiment with cichlids that suggests an answer to your question. In this experiment, we formed hierarchies in two ways: round robin competition and the usual way by group assembly. In round robin competition, the four fish in a group met each other only as separate pairs – first fish A and B, then C and D, etc. In group assembly, all fish in a group met together at the same time. We did this experiment to contrast the importance of individual differences versus social dynamics in establishing pecking orders. In round robin competition, the individual differences between fish were free to influence the outcomes of dominance encounters. If A ranked higher on dominance ability than B, A could dominate B when they met as a pair. But social dynamics could not play a part in how well A and B did in the hierarchy. That is, C and D could not observe the contest between A and B and then act based upon what they had seen. For example, C and D could not attack B just after she had received several nips by A, and A couldn’t take advantage of her “momentum” by attacking C or D after she had chased B several times in a row. But in the group assembly method of forming hierarchies, whatever social dynamics were important to pecking order formation were free to influence the establishment of the hierarchy. Of course, in group assembly, in addition to social dynamics, individual differences could also play a part in the formation of dominance relationships.

We reasoned that if individual differences in dominance ability were central to the formation of linear arrangements, then all or nearly all the pecking orders established through round robin competition should be linear. But if not, then relatively few would be linear. On the other hand, if social dynamics were crucial to linear pecking order formation, all or nearly all the pecking orders formed through group assembly would have linear structures.

We found that round-robin competition delivered moderate proportions of linear structures (about 56% in groups of four and 50% in groups of five fish), but group assembly produced nearly all linear pecking orders (92% in groups of four and 91% in groups of five). As in the previous experiment, individual differences in dominance ability could sometimes account for linear structures, but social dynamics virtually guaranteed them.

Peter:   Okay, I still think that differences in dominance ability must be important, but I’m beginning to get why you think that interaction in groups – what you call social dynamics – is crucial. What do you know about these social dynamics? Do you think they could have anything to do with the behaviors I observed in my chickens that I thought might be chain reactions – the ones when Veronica pecked Gordo just after she – Gordo – had been attacked by Clucky and when Veronica attempted to peck Clucky following her – Veronica’s – attacks on Annabella and Gordo?

Ivan: I think you’re onto something.  An experiment I did with chickens early in my career suggested something about the social dynamics that might be involved in creating linear pecking orders. I assembled groups of four chickens, and a research assistant and I recorded the identity of the initiator and receiver of every aggressive behavior in each group for 12 hours over two days. I found that some sequences of behavior – social dynamics – were common and that others that at least were theoretically possible, were not common. For example, a common one was for a chicken to attack one group member and follow that quickly by attacking another. A second example was for one chicken to peck a group member and for the one pecked to be attacked by yet another group member. Like Veronica pecking Gordo just after Clucky had pecked her. The interesting thing was that the common sequences promoted linear pecking orders, but the uncommon ones could have led to non-linear pecking orders.

Winner, Loser, and Bystander Effects

I don’t want to get too technical in describing how some of the sequences promote linear arrangements and some don’t. Let me just say that I discovered a way I could explain the presence of the common sequences and the absence of the uncommon ones more simply using what I called “winner,” “loser”, and “bystander” effects. A winner effect means that an individual winning one dominance encounter increases its probability of winning against a second opponent. When there is a loser effect, an individual that loses one encounter elevates its probability of losing against a subsequent opponent. In a bystander effect, an individual observing a contest improves its probability of dominating the loser and decreases its probability of dominating the winner in later contests.

Some people think that winner and loser effects – that success begets more success and loss promotes further loss – explain outcomes in human life. Let me return to sports as an example of this. Commenters talk about a player being on a roll or having a hot hand – for example, by scoring a string of baskets or having hits on successive at-bats in baseball. Or the converse – being in a slump or losing the winning touch. The belief behind the so-called “Matthew Effect” is similar – the rich get richer, and the poor get poorer. This effect is also called cumulative advantage, and social scientists use it to explain wide-ranging forms of social inequality – from the distribution of wealth in societies to the number of publications by scientists.   

After I published the article suggesting that winner, loser, and bystander effects might account for the formation of linear pecking orders, work by several researchers showed that combinations of these effects could indeed generate linear structures. I was elated. It was wonderful to see other researchers extend these ideas and to demonstrate, at least in principle, that these effects could do what I had proposed they could.

Two’s Company, Three’s a Crowd

But as time went by, I began to question the mathematical models and computer simulations behind these studies. One problem was whether winner and loser effects really occurred in groups situations – when three or more animals were together. Many researchers had observed a loser effect in isolated pairs of animals – when a loser met just one other individual – but they had not investigated it in groups. The winner effect was more elusive in experiments. Researchers had tested the effect in pairs of animals, and sometimes they found it, and sometimes they didn’t. But like the loser effect, they had not looked for it in groups. The models and simulations simply assumed that both effects did occur in group settings. I expected that the effects did, and my colleagues and I launched some experiments to make sure that the models and simulations were on firm ground.

In our experiments, we tested for winner and loser effects in isolated pairs and in group situations using cichlid fish. To verify the effects in pairs, we let two individuals establish a dominance relationship. We considered the dominant fish as the winner and the subordinate as the loser. Then we separated the winner and the loser and paired them with new individuals. In this experiment we found a strong loser effect – a fish with a losing experience usually lost again when it met a new individual. But we did not find a winner effect in the species we used. As I just mentioned, researchers do not find a winner effect as consistently in their experiments as they do a loser effect.

In the group situations experiment, we started with the same procedure as in the pair experiments: we let two individuals set up a dominance relationship. Then we added either one or two additional fish to the original pair to compose groups of either three or four individuals in total.  In these experiments, we found that there was no winner effect: the initial winner had no advantage against a third or fourth fish. This was not surprising since we did not find a winner effect in the pairs. But, more surprising, we found that in group settings the loser effect disappeared. Losers from the initial pairs did not lose at a level more than chance to the individuals added to make a group. Our experimental work showed no evidence for the basic assumptions upon which the models and simulations were based. As far as I am aware, there has been no further study of winner and loser effects in group settings, but research assuming that these effects do occur in groups continues, nevertheless.

I was initially surprised by our findings of group situations interfering with the loser effect, as well as with several additional effects that we tested for. But in retrospect, I think that I should not have been. We’re all familiar with cases in our own lives when adding a third or fourth person to a pair changes everything. Think of this situation: You’re young, having your first date with someone you’re really attracted to, when one of your parents or a friend joins the two of you and inserts themselves into the conversation. It’s that kind of change in the dynamics of the situation that we discovered in our experiments.

Our experiments were with fish, but for me, they raise a question about labeling other people as “losers”:  assuming just because they have done poorly in one situation, they will do poorly in all future situations. That’s been popular in politics lately, especially with one of our former presidents. At least in our experiments, a fish that was a “loser” in one setting was often a “winner” in its next situation.

Simulated Dynamics versus Actual Ones

A second problem with the mathematical models and computer simulations was that they had predicted the behavioral dynamics that animals used to form linear hierarchies, but they had not compared their predictions to what happened in real life. So, my colleague Brent Lindquist and I decided to make the comparison. We juxtaposed the records of aggressive behavior that the models and simulations said animals should use in establishing linear hierarchies with records of behavior that real chickens actually did use in creating hierarchies. When we did this, we found that the interaction records from the models and simulations looked nothing like those in the real chickens. The models and simulations had produced the right “answer” – linear pecking orders – but they had done it in a way unrelated to how it happened in real life. Unfortunately, this is not a rare example. It’s all too common in the social and behavioral sciences for researchers to claim that they have discovered the dynamics that generate a social pattern without comparing the dynamics predicted by their simulations and models with those used in real groups of humans or animals establishing those social patterns. I know how this might sound, Peter. But, sadly, it is far easier and more academically rewarding to publish papers describing models and simulations purporting to produce social organization than it is to carefully collect data on the behavioral dynamics occurring in real groups and to explain how those dynamics actually generate the social organization.

Early in my career, a senior professor, who was a friend and supporter, gave me a warning along these lines. He said that just because I had shown that the prior attributes hypothesis was not correct, I wasn’t under any obligation to try to discover how animals did form linear orders. He worried that I might get bogged down in the attempt to do so and that it would be bad for my academic advancement. Fifty years on, I think he might have had a point.

Even though I suggested in my earlier research that winner, loser, and bystander effects could account for linear pecking orders, I now think using them is a lot like employing the prior attributes approach. In both cases, researchers are proposing that linear structures are the result of individuals being ranked from top to bottom on their capabilities to perform aggressive acts against other individuals. Under the prior attributes hypothesis, individuals have these differences in capabilities before they join a group, and with the winner, loser, bystander framework, the individuals develop these differences in capabilities during interaction early in group formation. Both explanations agree that differences in the abilities of individuals generate linear orders, they just differ on when individuals come to possess those differences.

Peter: Good for you, Ivan. But I must tell you I really was rooting for the winner, loser, bystander explanation. Still, I don’t know about the beautiful part yet, but it sounds as if the true explanation might be more intricate than I’ve been imagining.

I’m assuming from the experiments that you told me about earlier that you still think that social dynamics produce linear pecking orders. Do you know what those social dynamics are?

The Myth of Stable and Enduring Pecking Orders

Ivan: Before I try to answer your question, I have to tell you something that might be upsetting. So far you and I have been talking about efforts to explain how pecking orders come to have linear structures. And although we haven’t said it, we have been assuming that these structures are stable – once the individuals form a particular linear arrangement, it stays in place for a relatively long period of time. I now must tell you that static and enduring pecking orders do not exist. We have been talking about explanations – prior attributes and winner, loser, and bystander effects – for something that doesn’t really occur.

Peter: I have to say that I am troubled and confused to hear this. It contradicts what I’ve always assumed about pecking orders, and it denies what I saw with my own eyes when I observed Clucky and my other chickens. Why didn’t you tell me this to begin with? It would have saved a lot of time.

Ivan: Maybe I was wrong, but I believed it would not have worked if I had told you from the beginning that your assumption that pecking orders were stable and long-lasting was incorrect. We took a winding road, maybe, but I wanted to showcase the problems with the prior attributes hypothesis and winner, loser, and bystander effects before I introduced the idea of social dynamics and a completely different way of thinking about pecking orders.

Recently, my colleagues and I published a study that shows that dominance hierarchies are not as static and long-lasting as most people assume. In this study, we observed groups of four individuals from three species – chickens, cichlid fish, and mice – forming hierarchies. These species are from three very different biological classifications (birds, fish, and mammals) and thus represent a broad swath of the animal kingdom. We observed these groups from the introduction of members for 12 hours in the chickens, 18 hours in the fish, and 72 hours in the mice. Over these periods of time, we recorded every aggressive interaction and the time at which it occurred among group members using categories of behavior most relevant to dominance in each species. As far as I am aware, these are the longest, continuous records of aggressive interactions ever recorded for any species establishing pecking orders.

When we examined the interaction records, we discovered that pecking order formation in all three species was an exceptionally dynamic process. During the observation periods, the individuals in the chicken and mice groups interacted hundreds of times, and in the fish groups, they interacted thousands of times. (The median number of interactions in the chicken groups was about 460, in the mice groups about 365, and in the fish groups about 6,000.) All this interaction created continuing changes in the pecking order structures of the groups.

Let me show you a figure displaying an excerpt of the aggressive interaction in a group of chickens. It uses a technique I call “music notation” to indicate when one chicken pecked another. Each arrow indicates a peck, and it goes from the line representing the aggressor to the line representing the receiver. The diagrams over the music notation “score” show the different pecking order structures that the group developed over time. They diagrams are placed over the specific peck that changed the pecking order from one structure to another. You can see that the interactions produced several different linear pecking order structures and one non-linear one. The ellipses marks between arrows indicate that several arrows of the same direction were left out to condense the graph. This group made more changes in its pecking order structures, but there isn’t enough space to show you the whole record.

You can see two things in this figure: First, there was not a stable and enduring pecking order structure. The group had several different pecking order structures, some lasting for shorter and some for longer periods of time. This pattern was repeated in virtually all the groups we observed. Initial linear structures were replaced by different linear structures or by non-linear ones. But even though specific pecking order structures were not stable and enduring, at any given point, the groups usually had one or another linear configuration. Specific linear orderings had limited lifetimes, but the class of linear structures was robust. A new linear arrangement usually arose phoenix-like after a previous linear or non-linear configuration was destroyed.

For nearly fifty years I had been doing research on pecking orders, and I had always assumed that a group quickly developed a linear arrangement and kept that arrangement for a long time – certainly as long as any experiment that I had carried out. The observation of changing, mostly linear, hierarchies was always there to be made, but other researchers and I had just assumed what pecking orders were like and had never bothered to look more carefully. After we completed our study, I still didn’t understand how pecking orders were established, but I knew that a very different explanation would be needed than those that other researchers and I had been pursuing. 

Second, the aggressive interactions among the individuals directly created the stream of changing pecking order structures. As the directions of interactions among the group members changed, so did the pecking orders. More generally, events – interactions – that occurred in time and space created the changing pecking order structures that occurred in the groups. Recognizing that interactions generated the ongoing flow of hierarchy transformations changes the conceptual framework needed to account for the organization of dominance behavior. Rather than trying to explain how individual differences or over-simplified social dynamics like winner, loser, and bystander effects produce stable, long-lasting linear structures, we must now discover how interactions are organized to produce changing, but mostly linear orders.

Peter: I’m amazed by these findings. It never occurred to me that pecking order structures could change so frequently. I just assumed that you got a pecking order and that was it. But I wonder whether you would have seen such dynamic changes in pecking order structures if you had used different species in your study. Maybe you just got lucky in the species you chose.

Ivan: That of course is possible, but it doesn’t seem likely to me. As I mentioned earlier, we used animals from three different biological classifications – birds, fish, and mammals – representing a broad range of species. So, I’m betting that not forming stable, long-lasting linear orders is common across a great many species.

Peter: Do you think that your groups would have settled down to a single, long-lasting linear structure if you had observed them long enough?

Ivan: That’s a good question, and I don’t know the answer. But let me point out, that if someone has the assumption that there really is one, true linear pecking order based upon differences in ability among the members of a group, this is of course a reasonable expectation. The idea would be that after many interactions, the “true” nature of the individuals’ attributes would make themselves manifest. Over time “quality will out”.

My best guess is that while pecking order structures might become more stable over time, there still would be episodic challenges, reversals in relationships, and re-reversals in relationships. Even if a group’s pecking order structure did become stable, we would still need to understand the dynamics of interaction that led to the stable arrangement and those that managed to maintain the stable order when challenges arose.

Peter: This is the most puzzling thing about your findings: How could I and so many dominance researchers observe what we thought were stable linear hierarchies if they were in fact not always stable and linear?

Ivan: I should point out that until recently, I had also assumed that there was a single, long-lasting linear ordering in a group of animals – as long as the group was small in number. It wasn’t until my colleagues and I began to analyze the data from our study that I became conscious of how often pecking orders changed their configurations. I simply never thought to question the assumption of stable and enduring linear arrangements. Like most other researchers, I had unintentionally used certain methodological techniques that made it almost impossible to discover the extent of hierarchy changes over time. For example, I recorded pecking order structures during periods when I thought that the “dominance relationships” among group members were unambiguous and stable. That is, I waited until one individual in each pair making up a group directed aggressive behaviors at the other individual in the pair several times in a row (usually six or eight times) without being counterattacked. Then I declared the aggressor dominant to the receiver. If there was any back-and-forth aggressive behavior in a pair, I assumed that their dominance relationship had not yet stabilized, and I waited until the aggressive behavior in the pair met my requirements. After I had delayed for a while – sometimes a few minutes or hours or until the next day – I would find a period in which, by my criteria, all the relationships in the pairs of a group were “stable”.  These stable relationships would usually compose a linear ordering.

But I didn’t realize two things: First, during those periods when I thought that the dominance relationships were not stable, the pecking order structure was changing and some of the structures were probably non-linear. Second, if I had continued to observe the aggressive interactions after I had declared that a linear structure was in place, I almost certainly would have seen new pecking order arrangements emerge – most linear but some non-linear. If the findings of changing but mostly linear hierarchies from our study of chickens, fish, and mice generalizes across species, then researchers who observe hierarchy structures at one point in time across many different species would have a high probability of seeing linear arrangements. It’s a short step from there to assume that the linear ordering they observed would persist over a long period. I don’t want to get into the weeds too deeply, but there are other techniques that researchers use that inadvertently reinforce the idea of stable and long-lasting linear hierarchies.

Observing a Clip versus the Whole Movie

In other words, Peter, most researchers in this field are, in effect, observing a short clip from an action-packed movie. They derive findings from what they glimpse in the clip and never see other parts of the movie. The cartoon, below, illustrates these two ways of observing aggressive interactions in a group of chickens to determine hierarchy structure. The first one views interactions just long enough to determine a structure and the second one views the interactions over an extended period.

Of course, these two different approaches are not just limited to the study of pecking orders. Rather than patiently observing how a social process unfolds and evolves over time, many social and behavioral scientists, in their eagerness for answers (and publications!), rush in, quickly gather some convenient data, and construct a superficial, and often conventional, account of the process. Many people like these accounts, but they do us a disservice by clouding our understanding of the social world that we live in. I think we should be guided by Yogi Berra’s wise advice for scientists: “You can observe a lot by just watching.” We need to take the time that’s required to gather the dynamic data needed to understand social processes and to develop explanations that may be at odds with more traditional thinking but are closer to how the processes actually work.

Peter: After hearing all this, I’m a little disheartened. It’s been a winding road, all right. And I’m not so sure that you were right in taking me down it. Aren’t we just reduced to explanations that don’t work for a kind of pecking order structure that doesn’t exist, a new and very different picture of the organization of dominance behavior in groups, and a claim that this organization arises from the social dynamics of aggressive interactions among the individuals in a group? Have I got that right?

Ivan: First, Peter, I appreciate that you find our talk disheartening. It would have been much better if I could have told you about a new and much better way to explain how pecking orders form and develop over time. I know I would be happier if I could do that. It’s something that I’ve been on the trail of for many years. But let me point out that getting rid of ideas that are incorrect but intuitively appealing and that influence popular thought and discussion, including the way people think about themselves and others, is itself an important step. I know that these ideas will linger on in popular discussions and in the work of dominance researchers, but at least for those who want it, there is a path forward in dominance research.

What’s Next?

Now the central question about dominance hierarchies becomes: What are the organizing principles of aggressive behavior in groups that produce changing but often linear structures of hierarchy order? How do these principles ensure the dynamic stability of linear forms? I have looked at chickens and fish interacting in real time and at video records of their interaction for thousands of hours, and I have some ideas about the organization of those interactions. But I have far from a complete picture. It may be my own limitations, but I believe this is a difficult conceptual problem. Perhaps some of the people best able to make further progress will have backgrounds in highly quantitative fields such as applied mathematics, physics, or computer science. But quantitative knowledge will not by itself be enough. To work effectively on this problem, researchers will also have to immerse themselves in the raw interaction data and to develop an intuitive feeling for the rules of aggressive interaction.

As I said when we began our conversation, I believe that the prior attributes idea (and the winner, loser, and bystander model as well) impoverishes our scientific imagination and our understanding of what an intricate and beautiful process pecking order formation really is. As I see it, groups are pulsing with complex streams of interaction, and somehow, out of these actions and reactions, pecking order structures emerge and vanish to be replaced by other pecking order structures that emerge and vanish in their turn. Instead of “simple” creatures using “simple” means to form a “simple” type of social organization, the reality is a group of individuals interacting with each other in complex ways that nonetheless generate a series of usually well-ordered configurations. This is the heart of it, isn’t it? That the mystery and, yes, beauty of the interaction process is how individuals in groups across very different species, each pursuing their own agendas, following some rules, but probably with many variations, can still produce these pecking order structures.

And given this view of the organization of dominance behavior, I find myself now wondering about social organization more generally. How many other kinds of social organization in humans and animals have what we think are stable and enduring forms created by individual differences or simple dynamics, but are continually being produced by ongoing events – processes in time and space? I have to say that the idea of social organizations that I can observe and might even be part of, are teeming with actions and events just to stay in place, really intrigues me. Even more, I believe this view opens new problem areas for scientists, a new appreciation of social organization by non-scientists, and a new awareness of what it means to be a participant in the social world.

Peter: Thanks, Ivan, I do sense your excitement in this. And I get the mysterious part. We don’t yet know how pecking orders emerge from the aggressive interactions in groups. I also get the intricate part. You observe large numbers of interactions, they apparently have complicated and convoluted patterns, nevertheless, the interactions produce changing but mostly linear pecking orders. But I’m still having trouble with what you call the “beautiful” part.

Ivan: I think you know that when I say “beautiful”, I don’t mean regular visual or aesthetic beauty. I mean a kind of scientific or intellectual beauty. At one level, the interactions producing pecking orders appear so protean as to lack any long-term organization. They seem only rooted in aggression, competition, and individual self-interest. Yet, when you analyze the interactions in the right way, these apparently protean interactions produce ever-changing but almost always linear forms. This is what I mean by “teeming with actions and events just to stay in place.”

 I marvel that relatively simple creatures pursuing their own interests, and with no plan (I’m assuming) for coordination at the group level, can, through their collective actions, generate hierarchy structures. The beauty for me lies in how the consistency of pecking order structures in the groups arise from what appears to be fluid, wavering, and unpredictable behavior on the part of the individuals involved. I believe that a process that can compile the behavioral events of individuals that seem so jumbled into a flow of pecking orders that are in fact supremely ordered must be most beautiful.  

Peter: I think I’m beginning to understand what you’re saying. How about this? Let’s say that I took a trip and observed small dance troupes all over the world: Scotsmen wearing kilts on oil drilling platforms in the North Sea, Sri Lankan teenagers in saris on sunny beaches under palm trees, Siberian reindeer herders dressed in traditional malitsas on the frozen tundra, modern dancers on the Lower East Side on Manhattan covered in metallic paint and feathers, and on and on. All the performances build up to a part filled with dazzling and complicated movements by the dancers. In each dance, this part continues for the rest of the performance, changing over time, but almost always keeping a pattern similar to the one it began with. All the dances went on for a very long time. None of the groups had a choreographer, and none knew any of the other groups. But the remarkable thing was that although none of the dances were exactly the same, the dazzling and complicated movements were astonishingly similar across all the troupes. Is this at least an approximate analogy to what you’re saying about the intricacy and beauty of pecking orders?  

Ivan: I think it is, Peter. And now that you’ve come to that, I believe you owe Clucky and her friends an apology.