Case of Conflict

Social Process and Hierarchy Formation in Small Groups: A Comparative Perspective Author(s): Ivan D. Chase Source: American Sociological Review, Vol. 45, No. 6 (Dec., 1980), pp. 905-924 Published by: American Sociological Association Stable URL: http://www.jstor.org/stable/2094909 Accessed: 28-08-2017 20:21 UTC

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide

range of content in a trusted digital archive. We use information technology and tools to increase productivity and

facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at

http://about.jstor.org/terms

American Sociological Association is collaborating with JSTOR to digitize, preserve and extend access to American Sociological Review

This content downloaded from 146.186.111.108 on Mon, 28 Aug 2017 20:21:51 UTC All use subject to http://about.jstor.org/terms

SOCIAL PROCESS AND HIERARCHY FORMATION IN SMALL

GROUPS: A COMPARATIVE PERSPECTIVE*

IVAN D. CHASE

State University of New York at Stony Brook

American Sociological Review 1980, Vol. 45 (December):905-924

The author investigates the theory, methods, and findings of animal behaviorists and social

scientists studying dominance hierarchies in small groups. In both disciplines: (1) the literature argues that the explanations of hierarchy structure are based upon differences in individual characteristics among group members; (2) although critical examination reveals that these explanations require stringent conditions to account for commonly occurring kinds of hierarchies in humans and animals, the available data indicate such conditions are not met; and (3) the hierarchy-formation process has not been adequately studied. In an attempt to alleviate the current problems in hierarchy research, I present the results of a particular animal study and develop a general explanation as to how hierarchy structures arise. This approach applies to both humans and animals and serves as a model of how the cumulative patterns of interactions among individuals produce group social structures.

Despite great differences in social be- havior in many human and animal groups, their dominance hierarchies are re- markably similar (see Brown, 1975; and Wilson, 1975, for general background on dominance hierarchies). Why should spe- cies so diverse as animals are from one another, and as all the animals are from

* Direct all correspondence to: Ivan D. Chase; Department of Sociology; State University of New York (SUNY); Stony Brook, New York 11794.

An earlier version of this paper was presented at the 1978 Annual Meeting of the American Sociologi- cal Association in San Francisco. Many colleagues have commented on and contributed to this paper, and I am grateful for their support. In particular, I would like to thank Jack Hailman for his incisive discussion and general encouragement. Jeffrey Baylis, Ann Becker, Shelly Cohen, Ken Feldman, Mark Granovetter, Gene Weinstein, Bill Wimsatt, and one anonymous reviewer provided many helpful criticisms of earlier versions of the paper, and I am most appreciative of their efforts. I would like to thank Gordon Stephenson for help with the SSR System, Sara Edwards and Nancy Raffetto for as- sistance with data collection, and Erica Jen for com- puter wizardry and assistance with techniques of analysis. Veronica Abjornson, Carole Roland, and Rosemarie Sciales speedily turned rough draft into final copy, and Mary Ann Huntington drew the fig- ures. Part of the research reported here was con- ducted under a grant to the author from the Harry Frank Guggenheim Foundation, and the observa- tional data were gathered in the Department of Zool- ogy, University of Wisconsin-Madison. The support of both the Foundation and the Department are gratefully acknowledged.

humans, form dominance hierarchies so similar in structure? And, even more ba- sically, how are we to study and explain the formation of dominance structures in small groups of men and animals? The goal, here, is to make a start toward an- swering these questions by developing a new approach to the study of hierarchy structures. This approach explains how hierarchies emerge from the interaction among group members rather than being generated by differences among those in- dividuals. It treats hierarchy formation as a developmental process in which the out- comes of previous interactions influence the course of successive ones, and it indi- cates how patterns of interaction fit to- gether to form the kinds of dominance hierarchies commonly observed.

This paper is in five parts: (1) a sum- mary of the data collected on dominance hierarchies and the methods used to gather those data in both the animal be- havior and the human social sciences lit- erature; (2) a critical evaluation of the theories that have been proposed to ex- plain hierarchy structures in both disci- plines; (3) a presentation of the results of a study of hierarchy formation in an animal group; (4) a new approach to the theory of hierarchy formation; and (5) an evaluation of the new approach as well as a discus- sion of the implications of the results for continuing research.

905

This content downloaded from 146.186.111.108 on Mon, 28 Aug 2017 20:21:51 UTC All use subject to http://about.jstor.org/terms

906 AMERICAN SOCIOLOGICAL REVIEW

HIERARCHY STRUCTURES IN ANIMAL AND

HUMAN GROUPS

The Animal Behavior Approach

One of the distinguishing features of the animal behavior or ethological tradition is the emphasis on behavioral observation and the use of ethograms (or catalogs of behavior) developed for each kind of ani- mal or human group studied. In a typical ethological study, a group is observed for a period of time sufficiently long to dis- cover the direction of dominance relation- ships among all possible pairs in the group. The direction of a dominance re- lationship in a pair is typically determined by asymmetries in the agonistic (a cate- gory including both aggressive and sub- missive behaviors) behavior of the indi- viduals toward one another. For example, the individual who delivers all or the majority of aggressive acts in a pair is considered dominant, and the individual who receives the acts or initiates most of the submissive gestures is considered subordinate. The pattern of the relation- ships in a group is frequently represented in a "dominance matrix." A dominance matrix has either a "0" or a "1" in all off-diagonal cells: a "1" in the i, j cell indicates that individual i dominates indi- vidual j and a "0" indicates that j domi- nates i.

All the dominance relationships in a group, taken together, form the hierarchy structure. Theoretically, hierarchy structures can vary from linear, at one extreme, to the other extreme in which each individual in the group dominates an equal number of other individuals. In a linear hierarchy there is an individual A who dominates all others, an individual B who dominates all but A, and so forth, down to the last individual who dominates no one. An example of a dominance ma- trix for a linear hierarchy with five indi- viduals is given in Table 1, and a matrix for a hierarchy as far away from linearity as possible-in which each animal domi- nates an equal number of animals-is given in Table 2.

A surprisingly uniform finding by ethologists is that hierarchies in small groups (less than about ten members) are frequently and perhaps predominantly

Table 1. Dominance Matrix Showing a Linear Hierarchy

Dominant Number Animal Dominated Animal Dominated

A B C D E

A - 1 1 1 1 4 B 0- 1 1 1 3

C 0 0- 1 1 2 D 0 0 0 - 1 1

E 0 0 0 0- 0

linear or near-linear. This finding is stable across many individual researchers' work and across an extremely broad range of animal species from insects to primates. For example, linear and near-linear hierarchies have been found among cer- tain kinds of wasps and bumble bees (Wil- son, 1971); various birds-such as chick- ens (Guhl, 1975), chaffinches (Marler, 1955), and red crossbills (Tordoff, 1954); domestic mammals-such as cows (Schein and Fohrman, 1955) and ponies (Tyler, 1972); wild mammals-such as coyotes (Bekoff, 1976) and buffaloes (McHugh, 1975); and primates in the wild and in captivity-including rhesus mon- keys (Sade, 1967), baboons (Hausfater, 1975), and vervets (Struhsaker, 1967). In the primates, hierarchies can become more complex through both the formation of coalitions and what is known as depen- dent rank (Jolly, 1972). In dependent rank, a particular individual (animal) is able to achieve dominance over another animal in the presence of a third animal (frequently its mother or consort) but not able to do so if the third animal is not present. How- ever, even in those groups where coa- litions and dependent rank are found, there is often a linear or near-linear

Table 2. Dominance Matrix Showing a Hierarchy in Which Each Animal Dominates an Equal Number of Other Animals

Dominant Number Animal Dominated Animal Dominated

A B C D E

A - 1 1 0 0 2 B 0- 1 1 0 2 C 0 0- 1 1 2 D 1 0 0- 1 2 E 1 1 0 0- 2

This content downloaded from 146.186.111.108 on Mon, 28 Aug 2017 20:21:51 UTC All use subject to http://about.jstor.org/terms

SOCIAL PROCESS AND HIERARCHY FORMATION 907

hierarchy if relationships are examined between pairs when a coalition partner or third party conferring dependent rank is not present.

Researchers in the ethological tradition who have studied human groups have fre- quently focused on hierarchical relation- ships; many of these studies have been on preschool children (McGrew, 1972; Missakian, 1976; Strayer and Strayer, 1976), but adolescent groups have also been examined (Savin-Williams, 1977; 1979; 1980). In addition to using nonver- bal, behavioral measures of dominance, as in the animal studies, some studies have also employed verbal indices of dominance-for example, if A gives an order to B and B obeys, then A is rated as dominant to B (Savin-Williams, 1977; 1979; 1980). As in the animal studies, findings indicate that human hierarchies are frequently linear or near-linear (Missakian, 1976; Savin-Williams, 1977; 1979; 1980). In those studies where data have been collected for some but not all possible pairwise interactions, the data indicate that only a few interactions de- viate from those expected in a linear hierarchy (McGrew, 1972; Strayer and Strayer, 1976).

Another uniform trend is that most studies by ethologists have been static. That is, in the great majority, studies have described a given hierarchy structure at some particular time rather than either explaining how that hierarchy structure was established or examining the proc- esses by which members of the group changed rank over time.

The Human Social Science Approach

The research of sociologists and social psychologists, like that of ethologists, has provided, for the most part, static de- scriptions of social structure in small groups. There are a few notable ex- ceptions (such as Newcomb's [1961] work) but these tend to utilize self- reported preference (friendship) data rather than observational data on asym- metric relationships. These static de- scriptions indicate a strong hierarchical component in the social structure of a

broad spectrum of human small groups. Three areas of research are of relevance here: the study of human small groups in laboratory settings, the study of these groups in "natural" situations, and the analysis of sociometric data.

When groups of unacquainted individu- als are assembled in laboratory settings, a differentiation of members quickly emerges along such dimensions as fre- quency of originating and receiving vari- ous kinds of behavioral acts and ratings of leadership and likeability by fellow group members (see Collins and Raven, 1969; Gibb, 1969; and Hare, 1976, for com- prehensive reviews). Some members show behavioral profiles that give them relatively high control over the actions of their fellows and over group activities in general, while other members exert rela- tively little influence on either group ac- tivities or their colleagues (e.g., see Bales and Slater, 1955, and the reviews cited above). Researchers using the status ex- pectation paradigm have shown that ex- ternal status characteristics-such as so- cial class, race, and gender-influence the distribution of participation and prestige in these groups (see Berger et al., 1972, for a general statement of the paradigm and Fennell et al., 1978, for a provocative dis- cussion of differences in male and female groups).

Although ethologists and laboratory small-groups researchers frequently col- lect the same kind of data-individual A directs an act of type X to individual B-their analyses of the data are different. As mentioned above, ethologists look at the hierarchy structures produced by all possible pairwise relationships among group members. Laboratory small-groups researchers, however, tend to concentrate on rank measures-the comparison of in- dividuals in terms of total interaction rates or group preference scores. Bales et al. (1951) is a partial exception in that they give pairwise interaction frequencies for 18 aggregated sessions of six-man groups.

Studies in natural settings-such as those by Blau (1955) of a bureaucratic agency, Hanfmann (1935) of kindergarten children, Homans (1950) of an industrial work group, and Whyte (1955) of gangs-provide further confirmation of

This content downloaded from 146.186.111.108 on Mon, 28 Aug 2017 20:21:51 UTC All use subject to http://about.jstor.org/terms

908 AMERICAN SOCIOLOGICAL REVIEW

the hierarchical structure of human small groups. These studies all find clearly dif- ferentiated status rank systems based upon both the behavioral actions and sen- timents of group members. Although most of these studies, like the laboratory small-groups studies, do not provide data on all possible pairwise relationships in their groups, the work of Hanfmann (1935) is an exception. Her data indicate the presence of a near-linear hierarchy in kin- dergarten children, comparable to those reported by ethologists for other human and animal groups. (See Mazur, 1973, for a pioneering comparison of hierarchies in humans and animals; his approach and conclusions, however, differ from those presented here.)

The studies of sociometric data by Davis (1970), Hallinan (1974), and Holland and Leinhardt (1970; 1971; 1972) provide some of the most sophisticated de- scriptions available of social structure in small human groups. Although one might expect preference choices to yield structures considerably different from those produced by dominance relation- ships, there are in fact considerable similarities. These researchers have dis- covered (with high consistency across studies and samples) that there is a strong trend toward transitivity in the preference relations of group members (for example, if A chooses B as a friend and B chooses C, than A will choose C also). As will be explained in greater detail, there is a similar strong trend toward transitivity in the dominance relationships of human and animal groups. In a linear hierarchy all possible triads have transitive dominance relationships; and, in a near-linear hierar- chy almost all the possible triads have transitive relationships. Davis (1970:850) suggests that, with more refined data where the stronger preference relation for every pair in a group could be determined, "sociometric data would tend to fit the model of a transitive tournament." A transitive tournament is another term for a linear hierarchy and, if Davis's suggestion is correct, then social structures produced properly refined preference relationships, and dominance relationships would be identical.

THEORIES PROPOSED FOR THE

EXPLANATION OF HIERARCHY

STRUCTURES

I will now examine theories pro- posed-explicitly and implicitly-in both the animal behavior and human social sci- ences literature, to explain the formation of hierarchy structures. This examination is based upon previous research by the author (Chase, 1974). The procedure used in that study was this: various theories of hierarchy structure were expressed in their basic (core) form, and that core form was converted into a mathematical model, a simple mathematical expression of what the core form implied. Then the stringency of the conditions needed by each mathematical model to predict the empirically common linear and near-linear hierarchies was examined and the avail- able data inspected, to determine if the conditions were met.

The two mathematical models that proved to encompass most of the expla- nations of hierarchies were a correlational model and a pairwise interaction model. For example: explanations that are bas- ically correlational models are those that indicate that individuals' positions in a hierarchy are determined by their physical attributes, their genetic endowment, their hormonal state, their past social perform- ance, their personality traits, the social labels that they have been given, or any composite of these factors. In order to account for linear and near-linear hierar- chies, correlations between ranks in a hierarchy and their scores on any of the above-mentioned variables or composites would have to be .9 or greater (Chase, 1974). This theoretical result holds for linear and near-linear hierarchies no mat- ter what species of animal, including hu- mans, are involved and no matter what variables or composite of variables are used to predict rank in a hierarchy. This is a high correlation coefficient, indeed, and one which indicates that any correlational explanation must fulfill a stringent math- ematical condition.

An examination of the literature indi- cates that correlations of .9 or higher are not usually found between a factor pre-

This content downloaded from 146.186.111.108 on Mon, 28 Aug 2017 20:21:51 UTC All use subject to http://about.jstor.org/terms

SOCIAL PROCESS AND HIERARCHY FORMATION 909

dieting dominance and rank in a hierar- chy. Some of the best correlational data available are for animals and, specifically, for the premier animal of dominance studies-the chicken. Correlations vary in size for different investigations and, in studies that examined the association between an individual's aggressive- ness-as indicated by its success in fights with other animals, and its place in a hierarchy-correlations ranged from about .4 to .8 (Chase, 1974). These corre- lations, then, were not as high as the target value of .9, although some were near.

Explanations of hierarchies that are pairwise interaction models in their ca- nonical form are, for example, differences in fighting ability for animals, theories dealing with the interaction of personality types for humans, and exchange theories for humans. In this model it is assumed that each member of a group has a pair- wise contest with each other member, that the winner of a contest dominates the loser in the group hierarchy, and that an individual has a particular probability of success in each contest. This model re- quires that there must be one individual with a .95 probability of dominating each other individual, a second individual with a .95 probability of beating every one but the first individual, and so forth down to the last individual who has only a .05 probability of beating any other individual (Chase, 1974). These results hold, re- gardless of the species of animal and re- gardless of what factors are assumed to explain success in pairwise interactions. As in the correlational model, these re- sults indicate that stringent mathematical conditions must be fulfilled with the pair- wise interaction model. Again, it has been found that data which best determined if these conditions were met were from ani- mals, specifically chickens, and that the required conditions were not fulfilled (Chase, 1974).

It is not that correlational and pairwise encounter theories and measurements give us no useful information about the success of individuals in dominance in- teractions or the places of individuals in hierarchies; they do, indeed; but the kind

of information they give us is not suffi- cient to account for a group level social structure, the overall hierarchy found in a group. Although individuals in a group may differ, it appears that the differences are not strong enough to explain empiri- cally common dominance structures (see Hallinan, 1974, for a similar argument concerning individual differences and sociometric structures).

TRIADS AND THE FORMATION OF

HIERARCHIES

If many of the currently accepted theories of hierarchies and dominance re- lations are not adequate, then what might explain the presence of linear and near- linear hierarchies so frequently found in human and animal groups? As discussed above, theories that attempt to explain hierarchy structure by correlations (or the characteristics of single individuals) and theories that use pairwise interactions (or the characteristics of pairs of individuals) appear to be inadequate. As Simmel realized long ago, three is a very impor- tant number in sociology, and I shall show that the structure of interaction in triads is very important for explaining the forma- tion of linear and near-linear hierarchies. In order to understand the importance of triadic interaction, one must be aware of a simple mathematical fact: in a linear hierarchy all possible triads have transi- tive dominance relationships, and if a hierarchy is not linear it contains at least one triad with intransitive dominance relationships-and the fewer the intransi- tive triads, the closer the hierarchy to linearity. In a triad with transitive domi- nance relationships, if individual A domi- nates B and B dominates C, then A also dominates C. So, A dominates the other two members, B dominates C, and C dominates no one. Figure 1A shows a triad with transitive dominance relation- ships. In a triad with intransitive domi- nance relationships, the three individuals cannot be arranged in rank order because each individual dominates one other indi- vidual. That is, A dominates, say, B; B dominates C; and C dominates, rather than is dominated by, A. Figure 1B shows

This content downloaded from 146.186.111.108 on Mon, 28 Aug 2017 20:21:51 UTC All use subject to http://about.jstor.org/terms

910 AMERICAN SOCIOLOGICAL REVIEW

A A

B _C B _C

A. Transitive Triad B. Intransitive Triad Figure 1. The Configuration of Dominance Relationships in a Transitive and an Intransitive Triad

A A

B C B C

A. B. Figure 2. (A.) Final Configuration of Dominance Relationships in All Triads Having Two Extant Dominance

Relationships at the End of the Observation Period. (B.) Initial Configuration of Dominance Relationships in Triad with One Reversal

a triad with intransitive relationships. In a linear hierarchy all individuals can be ranked uniquely from top to bottom by the number of group members they dominate, and, therefore, any three can be ranked with respect to one another and form a transitive triad (this can be seen, for example, by taking all possible subgroups of three in Table 1). In a hierarchy which is not linear, not all individuals are uniquely ranked by the number of group members they dominate-there is at least one tie and, therefore, at least one in- transitive triad. Table 3 shows the domi- nance matrix for a group in which there is a tie in rank between A, B, and C and one intransitive triad involving A, B, and C.

Table 3. Dominance Matrix Showing a Hierarchy with One Intransitive Trait Involving A, B, and C

Dominant Number Animal Dominated Animal Dominated

A B C D E

A - 1 0 1 1 3 B 0- 1 1 1 3 C 1 0- 1 1 3 D 0 0 0 - 1 1 E 0 0 0 0- 0

Experimental Design and Data Collection

Could linear and near-linear hierarchies be so common because animals and hu- mans use behavioral processes which

This content downloaded from 146.186.111.108 on Mon, 28 Aug 2017 20:21:51 UTC All use subject to http://about.jstor.org/terms

SOCIAL PROCESS AND HIERARCHY FORMATION 911

favor the formation of transitive triads and discourage the formation of intransitive triads? (Cf. Davis, 1970; Feld, 1980; Halli- nan, 1974; Holland and Leinhardt, 1970; 1971; 1972 for discussions of forces pro- moting transitivity in preference relation- ships.)

In order to answer this question, I de- signed an experiment to determine the actual extent of transitive dominance re- lationships in groups of three individuals and, if transitivity were common, to dis- cover the behavioral processes by which transitive relationships were generated. The species of choice for this experiment was the chicken; hierarchies in chickens are often linear or near-linear as are hierarchies in more socially complex spe- cies such as primates and humans, but at the same time the behavior of chickens is simple enough so that it is more tractable to analysis. In addition, the most exten- sive dominance research on any species is on the chicken.

In this experiment, groups of three pre- viously unacquainted chickens were placed in a neutral cage, and all aggressive actions among the individuals were re- corded. Twenty-four triads were ob- served; each chicken took part in three triads, and each chicken in a group had been in the same number of triads pre- viously. The data recorded consisted of the identity of the chicken initiating an aggressive action (the attacker), the iden- tity of the chicken attacked (the receiver), the kind of aggressive action (peck, feather-pull, claw (scratch), or jump-on), and the real (clock) time -at which each at- tack occurred. Each triad was observed for four hours to give a grand total of 96 hours of observation time. A combined total of 2,801 aggressive acts was re- corded, and the resulting average rate of aggression was 29.2 acts per hour or 116.7 acts per triad.

An SSR Keyboard was used to record the data. The data output from the Keyboard was converted to an electronic signal recorded on audio tape with a stan- dard tape recorder, and the resulting data tapes were transcribed by computer to produce hard copies of the data record and files for later analysis by computer (see Stephenson et al., 1975, and Stephen-

son, 1979, for a description of the SSR Keyboard and its capabilities). An excerpt from a data record is presented in Table 4.

The Extent of Transitivity in Triads

When groups of previously unac- quainted chickens are assembled, there are usually interchanges of aggressive ac- tions between individuals although, in some pairs, one animal immediately sub- mits to the other animal without ever fighting back. After either a short or more protracted series of interchanges, the asymmetric pattern of aggressive acts seen in established hierarchies develops: the interchanges stop and all or almost all of the aggressive actions are performed by one member of a pair against the other. At this point, one can think of a "decision" having been reached: a dominance re- lationship has been formed with one ani- mal being dominant and the other subor- dinate.

In order to determine when a domi- nance relationship had been formed and, thus, to discover the extent of transitivity in the experimental triads, two criteria were used. One animal was considered to be dominant over another if: (1) it deliv- ered any three aggressive actions in a row (any combination of pecks, feather- pulls, jump-ons, or claws) against the other animal and (2) there was a 30-minute period, following the third aggressive act, during which the receiver of the aggres- sion did not attack the initiator. The in- volvement of either member of the pair

Table 4. Data Excerpt

Act Time Aggressive Number Real Time Interval Acts

0 09:30:25.1 Session Begins 1 09:31:01.8 36.7 sec. 3PIa 2 09:31:03.5 1.7 3P1 3 09:31:31.8 28.3 1P3 4 09:32:06.8 35.0 1P2 5 09:32:08.6 1.8 1PU2b 6 09:32:10.1 1.5 1C2C 7 09:32:14.9 4.8 3P1 8 09:32:18.6 3.7 1P2 9 09:32:21.2 2.6 3J2d

a "pl, is the code for a peck. b "PU" is the code for a feather-pull. b "C" is the code for a claw (scratch). d ,J is the code for a jump-on.

This content downloaded from 146.186.111.108 on Mon, 28 Aug 2017 20:21:51 UTC All use subject to http://about.jstor.org/terms

912 AMERICAN SOCIOLOGICAL REVIEW

with the third member of a triad was not considered in these criteria and neither was the length of time necessary for the completion of the three acts. Once formed, a dominance relationship could be reversed if the subordinate animal ful- filled the criteria. While somewhat arbi- trary, these criteria have support as a valid index of stable dominance relationships-of the 54 dominance re- lationships meeting the criteria in this study, only two were later reversed within the four-hour observation period (after- ward, each chicken was returned to its home cage). The average time required to fulfill the criteria in the 52 initially formed relationships was 82.6 minutes, and only two relationships were established within less than 30 minutes before the end of an observation period. There was, then, a sufficient time period after the formation of most relationships for reversals to occur.

An examination of the configuration of relationships at the end of each observa- tion period indicated that 1 triad had no dominance relationships, 17 had two, and 6 had all three possible relationships. The 17 triads with two dominance relation- ships all had the same configuration-one animal dominated each of the other two. If the two dominance relationships were determined randomly, this configuration would have a probability of .25 and would be expected to occur, by chance, about four times in 17. This configuration is shown in Figure 2A. It is important to note that this configuration guarantees a triad with transitive dominance relationships regardless of which subordinate later comes to dominate the other (preliminary analysis of data from several groups of four chickens each, observed for 12 hours, indicates that subordinate animals tend to form dominance relationships among themselves, in time). If B later dominates C, there is a transitive triad with A on top, B in the middle, and C on bottom. If C later dominates B, there is a transitive triad with A on top, C in the middle, and B on the bottom.

Similarly, the 6 triads with three domi- nance relationships all had the same con- figuration, a triad with transitive relation- ships as shown in Figure 1A. A transitive

triad has a probability of .75 of occurring if relationships are determined randomly- and would be expected to occur between four and five times out of six. Taken to- gether, these findings demonstrate a marked tendency toward transitivity in the experimental triads, and this finding corroborates those of the studies of human and animal groups reviewed ear- lier.

An examination of the two groups with reversals in previously formed dominance relationships gives additional support for the tendency toward transitivity. In one of these triads, animal A initially dominated B but was herself dominated by C (see Figure 2B). With this configuration an in- transitive triad would have developed had B come to dominate C or a transitive triad if C had come to dominate B. In actual fact, A and C reversed their relationship, and this produced the configuration shown in Figure 2A which guaranteed a transitive triad. In the other triad, the ini- tial configuration of relationships was in- transitive, as shown in Figure 1B. By the end of the observation period, A and B reversed their dominance, with the triad having a transitive relationship shown in Figure 1A resulting.

Behavioral Processes Leading to Transitivity

I now present a data analysis that shows how the interaction process in triads helps to explain the tendency toward transitive dominance relationships. The first part of the analysis treats the long-term structure of the interaction process and the second part, the short-term structure.

1. Long-term Processes. This analysis of long-term processes includes the 17 triads with two relationships and the 6 with three relationships and excludes the 1 triad with no relationships. The analysis examines the effect of "winning" and "losing''-that is, the impact of either gaining dominance or becoming subordi- nate upon the further aggressive behavior of a chicken.

The analysis begins with the 16 triads in which only two dominance relationships were formed (no reversals) and 1 triad with three relationships in which one ani-

This content downloaded from 146.186.111.108 on Mon, 28 Aug 2017 20:21:51 UTC All use subject to http://about.jstor.org/terms

• Contents
• image 1
• image 2
• image 3
• image 4
• image 5
• image 6
• image 7
• image 8
• image 9
• image 10
• image 11
• image 12
• image 13
• image 14
• image 15
• image 16
• image 17
• image 18
• image 19
• image 20
• Issue Table of Contents
• American Sociological Review, Vol. 45, No. 6, Dec., 1980
• Volume Information [pp.1039-1045]
• Front Matter
• The Presidential Address: The Challenge and Opportunities of Applied Social Research [pp.889-904]
• Social Process and Hierarchy Formation in Small Groups: A Comparative Perspective [pp.905-924]
• Weber's Last Theory of Capitalism: A Systematization [pp.925-942]
• Immigration and the Expansion of Schooling in the United States, 1890-1970 [pp.943-954]
• The Meek Shall Not Inherit the Earth: Self-Evaluation and the Legitimacy of Stratification [pp.955-971]
• Research Notes
• Suburban Status Instability [pp.972-983]
• Generating Newsworthiness: The Interpretive Construction of Public Events [pp.984-994]
• Comments
• The Relay Assembly Test Room: An Alternative Statistical Interpretation [pp.995-1005]
• Worker Productivity at Hawthorne [pp.1006-1027]
• Welsh Political Mobilization: A Comment and a Note [pp.1028-1031]
• Aggregation Gain and Loss in Electoral Research [pp.1031-1036]
• Back Matter [pp.1037-XVI]