the identity of human social groups

16
THE IDENTITY OF HUMAN SOCIAL GROUPS by Mike Robinson Social Synthesis Unit, 61, Kings Road, Kingston, Surrey, United Kingdom KEY The existence of human social groups is taken as problematic. What is it that constitutes a group “identity”? What is it that stays the same despite changes in membership? The everyday notion that a group is the same if it stays about the same size and continues to do the same sorts of things is taken seriously. “Size” and “activity” are key variables that must be kept within limits if the group is to maintain its identity. This, in turn, implies a stabilizing process amenable to cybernetic analysis. From observation and experimentation with natural and laboratory groups, a model of this process is developed. It incorporates mechanisms for preserving values of key variables, and enables us to identify sources of instability. The assumptions of the model are tested by computer simulation, the GROUP-1 program, and the results are found to compare well with historical data from natural groups. Processes of special interest are a triple size-regulation mechanism, and the appear- ance of the “early leaver phenomenon” in informal groups. “his means that new members have the highest probability of leaving, and this is well known in industrial situations. It is found to be an essential correlate of stability. WORDS: human social system, stability,normative structure,preference set, computer simulation. r+z UMAN SOCIAL groups that exist for long H periods of time develop recognizable identities. We refer to them by the same name. The membership may change, but the group does not, or so we assert in our ordinary language usage. Our everyday no- tions tell us that if a group stays about the same size and continues to do the same sorts of things, then, despite considerable change of membership, it is the same group. This description is not too far from a rudi- mentary cybernetic analysis: The identity of groups consists in the maintenance of (at least) two variables, size and activity, within limits. If this is the case, then we would also expect to be able to identify regulatory systems that operate on these variables to keep them within the limits. It turns out that this can be done, and the basic regulatory systems of human social groups are described in this paper. Historically, the research was in the tra- dition of Homans’ (1950, 1961) attempt to elucidate a set of general social organiza- tional principles, and was much influenced to Roy Rappaport’s (1968) investigative methodology. The community studies of Roger Barker (1968) and his associates in developing concepts for studying the envi- ronment of human behavior played a major role in shaping the form of the research. 114 Behavioral Science. Volume 26. 1981 THE MODEL A schematic representation of the way in which groups regulate the major variables of size and activity is given in Fig. 1. The model was derived from extensive research involving both natural and laboratory groups, and was tested by computer simu- lations, the results of which were mapped onto the histories of the natural groups. The labels in the diagram, and other rele- vant terms, are defined below. Human social group In order to concentrate on size and activ- ity as related to the internal functioning of groups, constraints were introduced on the types of group to be studied. The groups were voluntary and informal. Members were not paid, nor did they employ paid functionaries, nor did they have legal obli- gations to each other (as in a family group). Although the findings have implicationsfor work groups and organizations with formal structures, it was felt that there were too many powerful extraneous factors at work to take these as the subject matter. The groups were also permeable: Members could join or leave at will. This is an obvious requirement for the study of size change.

Upload: mike-robinson

Post on 06-Jun-2016

212 views

Category:

Documents


0 download

TRANSCRIPT

THE IDENTITY OF HUMAN SOCIAL GROUPS

by Mike Robinson Social Synthesis Unit, 61, Kings Road, Kingston, Surrey, United Kingdom

KEY

The existence of human social groups is taken as problematic. What is it that constitutes a group “identity”? What is it that stays the same despite changes in membership? The everyday notion that a group is the same if it stays about the same size and continues to do the same sorts of things is taken seriously. “Size” and “activity” are key variables that must be kept within limits if the group is to maintain its identity. This, in turn, implies a stabilizing process amenable to cybernetic analysis.

From observation and experimentation with natural and laboratory groups, a model of this process is developed. It incorporates mechanisms for preserving values of key variables, and enables us to identify sources of instability. The assumptions of the model are tested by computer simulation, the GROUP-1 program, and the results are found to compare well with historical data from natural groups.

Processes of special interest are a triple size-regulation mechanism, and the appear- ance of the “early leaver phenomenon” in informal groups. “his means that new members have the highest probability of leaving, and this is well known in industrial situations. It is found to be an essential correlate of stability.

WORDS: human social system, stability, normative structure, preference set, computer simulation.

r+z

UMAN SOCIAL groups that exist for long H periods of time develop recognizable identities. We refer to them by the same name. The membership may change, but the group does not, or so we assert in our ordinary language usage. Our everyday no- tions tell us that if a group stays about the same size and continues to do the same sorts of things, then, despite considerable change of membership, it is the same group. This description is not too far from a rudi- mentary cybernetic analysis: The identity of groups consists in the maintenance of (at least) two variables, size and activity, within limits. If this is the case, then we would also expect to be able to identify regulatory systems that operate on these variables to keep them within the limits. It turns out that this can be done, and the basic regulatory systems of human social groups are described in this paper.

Historically, the research was in the tra- dition of Homans’ (1950, 1961) attempt to elucidate a set of general social organiza- tional principles, and was much influenced to Roy Rappaport’s (1968) investigative methodology. The community studies of Roger Barker (1968) and his associates in developing concepts for studying the envi- ronment of human behavior played a major role in shaping the form of the research.

114

Behavioral Science. Volume 26. 1981

THE MODEL

A schematic representation of the way in which groups regulate the major variables of size and activity is given in Fig. 1. The model was derived from extensive research involving both natural and laboratory groups, and was tested by computer simu- lations, the results of which were mapped onto the histories of the natural groups. The labels in the diagram, and other rele- vant terms, are defined below.

Human social group

In order to concentrate on size and activ- ity as related to the internal functioning of groups, constraints were introduced on the types of group to be studied. The groups were voluntary and informal. Members were not paid, nor did they employ paid functionaries, nor did they have legal obli- gations to each other (as in a family group). Although the findings have implications for work groups and organizations with formal structures, it was felt that there were too many powerful extraneous factors at work to take these as the subject matter. The groups were also permeable: Members could join or leave at will. This is an obvious requirement for the study of size change.

THE IDENTITY OF HUMAN SOCIAL GROUPS

TECHNIQUE

NORMATIVE

2 2 I-

a lu

115

Size Size was defined as the number of mem-

bers in the group. Since the notion of mem- bership is not always well defined (holding a card, paying a subscription), membership was further defined as “being recognized as a member by other members of the group.” Group size, at any given point in time, was thus the sum of members who were present and members who were absent. Although this accords with common sense, it should be noted that group size cannot be estab- lished by a simple head count of people at meetings.

Absent Absent is simply the number of group

members who are absent from any meeting or activity.

Inactive Inactive is the term used for the number

of group members who are present at any

given meeting or activity, but who do not have a role that contributes to the activity.

Leaveljoin Leave/join is the ratio of the number of

people who leave the group (cease to be members) to the number of people who join the group (become members) in a given time period. The most natural and conven- ient time period to use was the week for the groups studied. Thus (t:10(1/2)) and (t: 11(0/0)) would mean 1 person left and 2 joined in week 10, and no one joined or left in week 11. Obviously size in week t+l is determined by size and join/leave in week t.

Activity As with leave/join, activity levels are

quantized into weeks. Activity is defined as the total number of person-hours spent per week on group-defined activities. In some cases it was not practicable to measure this

Behavioral Science, Volume 26. 1981

116 MIKE ROBINSON

directly, and it will be stated whenever alternative indices are used.

Normative structure Normative structure, for any group, is its

entire set of beliefs-moral, social, behav- ioral; prescriptive and descriptive. Follow- ing Bates and Cloyd (1956), a belief was assigned to the normative structure if it was believed by the majority of group members to be held by the majority of group mem- bers. It should be noted that this definition allows for false norms: The group may be- lieve a majority holds a particular view when it does not.

Normative strength Normative strength is the ratio of those

actually holding a set of beliefs to those believed to hold them. Normative strength is represented in the model as a threshold, since its effect is all or nothing. At any point in time an external observer may establish gradual changes in normative strength by interviews. To the group itself, the over- throw of a false norm is instantaneous. At one moment the group believes the major- ity to hold a given view; suddenly it realizes it does not; a new norm is “revealed.”

Technique Technique is a prescriptive subset of the

normative structure. I t is defined as a defi- nite, repeatable, and integrated set of pro- cedures that are to be carried out in suc- cessfully attaining specific goals. This con- cept is similar to Barker’s (1968) account of a behavior setting program. He says, “The complete program of a setting is usually stored within the inhabitants of penetration zone 5/6; parts of it are stored within the inhabitants of more peripheral zones. The program is sometimes written out, as in the lesson schedule of a teacher or in the agenda and operational guides of a business meeting” (p. 168). He also comments, “A central problem of behavior setting opera- tion is to get the proper program stored within performance zone inhabitants.. . . This is accomplished by formal training and/or experience in the setting, and it requires time” (p. 80).

Environment This is the human environment of the

group. It is the pool from which new mem- bers may join and into which members who leave may disappear.

Size: Step increase/step decrease These labels indicate that size has in-

creased or decreased beyond the limits at which the group can maintain its stability and identity. These limits are determined by a relation between technique and activ- ity (described later), and have to be empir- ically determined for any given group.

Collapse The group ceases to exist.

External intervention The intervention of persons or forces ex-

ternal to the group that prevents activities being carried out or successfully completed.

Factionalization: Restructuring/ destructuring

Factionalization is the decomposition of a single group into two or more groups. This event is characterized by the appear- ance of two or more distinct normative structures where there was only one before. Unlike false norms, the event of factionali- zation is obvious to the group itself, and it characterizes itself as two or more groups. Factionalization is an unstable state, and leads to restructuring or destructuring. In either event, the group changes its identity. Restructuring means that stable relation- ships are established between the newly emergent groups. It marks the transition from informal to formal organization. De- structuring means that the group activity patterns are disrupted to the point at which the group collapses.

THE RELATIONS IN THE MODEL

For explanatory purposes, it is conven- ient to split the model into three parts: the normative system and the size regulating subsystem (which together comprise the stable states of a group), and the unstable states (the area in which change occurs).

Behavioral Science, Volume 26, 1981

THE IDENTITY OF HUMAN

TECHNIQUE

t NORMATIVE STRUCTURE

f

SOCIAL GROUPS 117

FIG. 2. The normative system of a group.

The normative system The first point to note here is that the

way in which the normative structure is maintained has much in common with the logic of the self-fulfilling prophesy. A set of techniques defining activity are evolved from the common beliefs of the group. The resulting activity maintains the normative strength-aggregate agreement with the normative structure-and thus the struc- ture itself. The group sets out to “confmn,” not “falsify,” its beliefs. A direct conse- quence of this, to be illustrated later, is that the impetus for change in any stable group must come from outside that group.

Thus we may say that techniques deter- mine activities (and patterns of activities), and the success of the activities reinforces the techniques. Stable groups are charac- terized by the ability to produce regular patterns of activity. It is mainly by this that they are recognized. When patterns of ac- tivity are represented by weekly quantities of activity (person-hours) the range of ac- tivity for the group has been determined. The numerical values of this range will differ for different groups-but activity levels that fall outside this range mean that the group is becoming unstable. Its tech- niques are threatened.

The role specifications inherent in the set of techniques serve as a template that de- termines the limits of group size. Function- ally, these may be defined in the following way: lower size limit: the minimum number

of members capable of filling the maximum number of roles specified by any technique; upper size limit: the maximum number of members capable of being absorbed by the maximum number of roles specified by any technique.

From this it follows that any level of activity may be achieved from any level of size. The independence of size and activity (within their normal ranges) is a character- istic of stable groups. On the other hand it is assumed that group activities are socially visible; hence the connection between ac- tivity and environment. The connection be- tween environment and size means that some people in the environment are at- tracted by the group activities and join. Since the assumption is made (and will be empirically justified) that joining rate is constant, this does not affect the indepen- dence of size and activity. We now have to explain how the leaving rate is (approxi- mately) matched to the joining rate.

The size regulating subsystem We have said, to be justified later, that

the joining rate is constant. We now have to explain how the leaving rate is approxi- mately matched to the joining rate so that size remains within limits.

Factors affecting leaving rates have been plumbed by social and industrial psycholo- gists in some depth. Absenteeism, “span of control,” group size, sociometric structure, frustration and work disruption, participa-

Behavioral SEience, Volume 26, 1981

118 MIKE ROBINSON

INACTIVE < ++ SIZE

D ENVIRONMENT

FIG. 3. The size regulating subsystem of a group.

tion and responsibility, size of the standard deviation from the production mean, “in- tention to quit,” and various attitude sets have all been related to leaving behavior. When terminological differences and inap- propriate factors had been screened out, size, absenteeism, and inactivity were found to be most deeply connected with leaving behavior. In stable groups, size was a good predictor of joining and leaving-in other words, of size in the following week. This is, of course, a characteristic, not a cause of stability.

In informal, voluntary groups a dual mechanism of size regulation is found. Size increase means there is less activity avail- able per member. Since the groups are vol- untary, it can be assumed that participation in activity is rewarding in itself. In this sense activity is a resource. As groups ap- proach the upper limits of their normal size range, the groups become less rewarding to their members, and two things happen: Ab- senteeism increases. Members are, so to speak, put in reserve against a time when the group may be “short” of members. In- activity increases. Members thus deprived of activity tend to leave. There is an inverse relationship between these two processes: Increased absenteeism means less inactiv- ity and less leaving; increased inactivity means more leaving and less absenteeism. The simulation brought to light a further

sophistication of these mechanisms, the “early leaver phenomenon”, and this will be discussed later.

Unstable states Unstable states follow from a block on

group activity, or from an increase or de- crease in membership that takes the group outside its normal size range. Both cause a crisis of technique, a step change in the normative strength, and the disruption of the normative structure.

The intervention of external forces that prevent the realization of technique as ac- tivity may result in inactivity or hyperac- tivity. If the latter is not successful in re- storing the status quo, then inactivity fol- lows. This results in doubt about the nor- mative structure in which the techniques are embedded, and factionalization or a step decrease in size may follow. Faction- alization, the appearance of two or more normative structures where there was only one before, may lead to the restructuring of the group. The relations between the emer- gent groups may become normalized. If this does not occur, further inactivity precipi- tates leaving (people look for more satisfy- ing company) and the group collapses.

A sudden large increase in group size results in inactivity, factionalization, and its consequences. A sudden large decrease in group size again means that the resources

Behavioral Science, Volume 26, 1981

THE IDENTITY OF HUMAN SOCIAL GROUPS 119

are lacking to realize the techniques; inac- tivity precipitates leaving, and the group collapses.

THE EMPIRICAL BASIS OF THE MODEL

The model was developed from three ma- jor (participant observation) studies of nat- ural groups, and aspects of the processes were tested on various laboratory and ex- perimental groups. Since these studies were carried out to develop the model, they do not count as “tests” of it. Nevertheless, they provide important illustrative and support- ive data. The assumptions of the model, in rigorous but restricted form, were tested by their ability to reproduce (in simulation forms) the historical behavior of these groups.The three natural groups were:

(i.) KDIL-a subdelinquent community in South London of about 60 members di- vided into subgroups. Although this study was important in the development of the model, it was undertaken before the tech- niques of variable measurement were stan- dardized, and will not be cited here.

(ii.) BRP-A housing community action group, “the Brighton Rents Project,” of about 60 members, divided into subgroups, and located in the seaside town of Brighton in the South of England. Although this group was selected to examine process of stability, the major interest here will be in the way stable processes broke down or were broken down, and the way this related to the effects produced by simulation runs. A brief description and history of the group is in order before we consider the quanti- tative data.

The BRP was formed in May, 1969, and effectively collapsed in December of the same year, although meetings finally ceased in March 1970. The BRP was formed as an umbrella organization (covering the politi- cal spectrum from liberals through social- ists to anarchists) to agitate for better local housing conditions. Its range of activities was reflected in its subgroup structure. The Rents Registration Group gave legal advice and support to poor tenants in bad condi- tions. The Investigation Group sought to. demonstrate the relationship between bad housing and local vested interests. The Squatting Group was set up with the simple

aim of placing homeless families in munic- ipal property that was standing empty. The Pamphlet Group was organized to docu- ment and publish the findings of the Rents Registration, Investigation, and Squatting Groups. The Propaganda Group was to pro- vide backup services, in the form of leaflets and posters, for the other groups. Finally the Coordinating Committee was set up, as its name suggests, to link the activities of all the other groups. In theory, all groups followed a policy mandate laid down by majority vote at public meetings. The for- mal organization of the BRP is illustrated in Fig. 4.

The informal, actual organization of the BRFJ was very different. Most of the sub- groups formed were unable to create effec- tive techniques, failed to regulate size and activity, and collapsed. Only two subgroups achieved continuing indentity and stability. These were the Rents Registration Group and the Squatting Group. Both operated in a semiautonomous way. The actual orga- nization of the BRP is illustrated in Fig. 5.

The Squatting Group stabilized inter- nally, but the success of its techniques for placing homeless families in empty prop- erty provoked (as one could expect) a strong reaction from the municipal author- ities. Finally, court action blocked these activities. The normative crisis that fol- lowed also drew in the Rents Registration Group, and ultimately caused the collapse of the BRP as a whole. The details of this will be considered in conjunction with un- stable simulation runs. (i.) IREv-the autonomous local branch

of a national left-wing group. Its size fluc- tuated between 5 and 10 members over a three-year observation period. The group exhibited all the characteristics of stability described in the model. This was somewhat paradoxical, since its explicit and stated aim was to proselytize and grow. The de- tailed functioning of the group will be re- viewed with the results of simulation runs.

GROUP-1: A SIMULATION OF GENERAL GROUP PROCESSES

In order to test the general model and its consequences, the simulation program GROUP-1 was developed. It was written in

Behavioral Science. Volume 26. 1981

120 MIKE ROBINSON

FORTRAN IV and run on the ICL 1900 series computer at Brunel University. The pro- gram and full specifications can be found in Robinson (1977).

The elements of the simulation were a set of “individuals” and a set of “activities.” “Individuals” were assigned to the “group” or the “environment,” and could move be- tween the two. Each “individual” was char- acterized by three parameters: (i) a specific, but variable, “level of commitment,’’ rep- resenting the number of hours per week he/ she can devote to group activity; (ii) an ordered set of preferences on the available activity; (iii) (for “individuals” in the “ g r ~ ~ p ” ) the name of the activity in which he/she is currently involved. Each activity was characterized by two parameters: (i) The number of “individuals” and (ii) the number of “person-hours” needed for its successful completion.

The basic rules of the simulation were as follows:

(i.) One or more subset(s) of “individ- uals” in the “group” “decide” to carry out one or more activities.

(ii.) The “level of commitment” of those “individuals” who cannot participate in

their most preferred activity decreases. There is a provision for increase or decrease in “level of commitment” in the course of activity, depending on the number of “in- dividuals” and the person-hour require- ments of the activity.

(iii.) An activity is deemed to be success- fully completed if the aggregate “level of commitment” of those “individuals” who participate matches the person-hours needed to complete it.

(iv.) Those “individuals” who partici- pate in successful activities increase their preference for that activity, while those who participate in unsuccessful activities decrease their preference for that activity.

(v.) “Individuals” with a very low “level of commitment” leave the group.

(vi.) “Individuals” join the group accord- ing to a random schedule that can be mod- ified by the experimenter, or linked to the activity level of the group.

(vii.) “Normative structure” is repre- sented in the simulation as an n-person payoff matrix, in which the outcome to each player is determined by the choice of all players. The matrix itself is the set of pref- erence orderings of the “individuals” in the

Behaviural Science, Volume 26, 1981

THE IDENTITY OF HUMAN SOCIAL GROUPS

9

RENTS REGISTRATION GROUP

- RENT REGl!XRAlION - - INVESTIGATION - - PAMPHLETS -

FIG. 5. Informal organization

group. Table 1 shows a typical preference matrix. The numbers in the matrix repre- sent the preference level of each “individ- ual” for each activity. Thus “individual” (IND) 1 prefers activity (ACT) 6 to activity 2, activity 2 to activity 7, activity 7 to activity 3, and so on.

(viii.) “Decisions” on which activities to carry out are made in the following way. The first preference of each “individual” is assumed to be hisher choice of activity. If an activity receives a sufficient number of first choices, it is carried out. If no “deci- sion” is made on first choices, then second choices are included, and so on. To illus- trate this in the Table 1 matrix, we can simplify, and assume that each activity needs at least three participants in order to be attempted. We see immediately that ac- tivity 5 wil l be chosen as it receives 3 first choices. Activity 2 will also be chosen since it receives 2 first and 1 second choice. (“In- dividual” 1 must be contentwith his second choice of activity, since his first choice of activity 6 is not shared at the first or second level of preference. “Individuals” 7 and 8, whose second choice it was, have already opted for activity 5.) The remaining two ‘‘individuals” (2 and 4) who have not achieved their first or second choices must

n SQUATTING GROUP

- ANARC Y - SQUAl

hy :HISTS -

’TIN6 -

\ of the Brighton 1

- 3ents Project.

121

now join in activity 2 or 5 or do nothing. (ix.) “Individuals” who participate in ac-

tivities that are successfully completed in- crease their preference for that activity. Alternatively, if the activity is not success- fully completed, the “individuals” who par- ticipate in it decrease their preference for it.

It should be noted that the psychological assumptions in the simulation are minimal. People put less effort into things they did not want to do in the first place; they prefer things that work to things that do not.

SIMULATION RESULTS

The normative system As in the model, runs that characterized

“stable groups” also showed size and activ- ity to be independent. In no case was the 2 probability of H, less than .2. By com- parison, the size/activity correlation of the most stable natural group, IREV, gave a x2 probability of H, of .9, and the data is reproduced in Table 2.

A condition of the independence of size and activity is the determination of size range by technique. To examine this rela- tion we introduce the notion of role. In natural groups this was defined as a subtask

Behavioral Science, Volume 26.1981

122 MIKE ROBINSON TABLE 1

GROUP 1 PREFERENCE SET. IND. ACT” 1 ACT2 A C T 3 ACT.? ACT5 ACT6 ACT7 ACT8

I 6 2 4 8 5 1 3 7 2 6 4 8 5 3 7 2 1 3 5 2 3 8 1 7 4 6 4 8 4 2 1 6 3 5 7 5 4 1 6 7 3 5 8 2 6 5 1 6 7 4 8 3 2 7 2 7 6 3 1 2 8 5 8 7 5 3 8 1 2 4 6

‘IND = individual: **ACT = activity.

(specified by a technique) carried out by an individual. Since the number of subtasks involved in carrying out an activity corre- lated highly with the number of person- hours needed to carry it out, subtasks, or roles, provided an alternative measure of activity. Using this relation in the simula- tion, roles were simply defined as person- hours (“level of commitment”) divided by a constant (usually, for technical reasons, 5) . The size range and distribution for IREV is given in Table 3.

Since there was no correlation between number of roles (an alternative measure of activity) and size, we wish to establish the maximum number of roles that can be filled by the minimum group size. The mean number of roles was 6.4, with a standard deviation of 3.6. Thus the maximum likely number of roles was (approximately) 13, and the minimum size 5, an average of 2.5 roles per person. In fact, active members filled on average 2 roles, so one can see that the minimum size is capable of filling the maximum likely number of roles. Similarly, it is clear that the upper size limit (10) is near to the maximum number of people that could be absorbed by the maximum likely number of roles. Various simulation runs of “stable groups’’ generated similar relations. Thus we may say that although size and activity are independent, technique serves as a “template” to determine the range of both.

TABLE 2 IREV: SIZE AND ACTIVITY LEVELS.

Level of Activity

0-9 10-19 20+

5-6 7 8 2 7-8 m 29 13 !#-I0 15 22 11

Slze

TABLE 3 SIZE RANGE AND DISTRIBUTION OF IREV.

Sue Frequency fin weeks)

6 11 34 29 45

5

Next we have to justify the assumption, stated earlier, that joining rate in stable groups is constant. Joining behavior of in- dividuals is affected by the social visibility and the attractiveness (Newcombe, 1961) of the group to be joined. Both of these factors must be underpinned by the type and level of activity. In the case of IREV, regression showed that the best predictor of joining behavior was the maximum activ- ity level of the previous month. (The cor- relation coefficient was .32 with a f signif- icance level in excess of .001. The constant was 6.6 and the increment .06 of the maxi- mum activity level, giving a predicted in- crease of 2 members soon after the maxi- mum activity level (36) was reached.)

Joining behavior is associated with high levels of activity, but the effect is randomly time lagged. Since high levels of activity were regular (there was a pattern of activ- ity) and other unknown factors affected joining, joining rate is best represented as a constant.

The constant joining rate assumption was confi ied by the simulation, where it proved impossible to generate “stable groups” when joining was tied to activity levels (with or without random time lags). A joining rate based on activity levels de- stabilized the simulated groups in several interesting and realistic ways. These wil l be considered later under the heading “Unsta- ble States.”

Lastly, we have to consider how the no- tions of “normative structure” and “nor- mative strength” are represented in the simulation model. The process of activity selection by consideration of the first to nth preferences of “individuals” in the simula- tion replicates, in a simplified way, the pro- cess of discussion in a natural group. Al- though no allowance or weighting is made for the status of “individuals,” this is effec- tively recreated in the course of the simu-

Behavioral Science, Volume 26, 1981

THE IDENTITY OF HUMAN SOCIAL GROUPS 123

lation. The preferences of “individuals” who have been in the group longest come to show a similar profile in the deep struc- ture of preference orderings. This may be convergent, as in the case of a homogeneous group, or it may be divergent, as in the case of role differentiation in connected activi- ties. The similarity of preference orderings frequently allowed older “individuals” to dominate the decision process. Although the preference set in the simulation only mirrors that part of the “real” normative structure we have termed techniques, it is reasonable to assume that the function of general norms is to provide a conversational context for decision making. This is re- placed by operations within the choice set. “Normative strength” is represented in the simulation as the mean level of preference for a chosen activity. As in natural groups, successful completion of activities increases normative strength, while failure decreases it. Finally, disruption of preference order- ings that have been established over time in the simulation has similar effects to the disruption of normative structures in natu- ral groups. Both effects will be considered under “Unstable States.”

The sue regulating subsystem In stable groups, size fluctuates within

limits, and the leaving and joining rates are approximately matched. Figs. 6 and 7 show

i 15

the size and numbers of people who joined and left for IREV and a simulated stable group. On the gross phenomenon the match between natural and simulated groups is close. In terms of detail, we have to account for the absenteeism and inactivity/leaving regulators.

Operation of the absenteeism sue regu- lator was observed in the Squatting and Rents Registration Groups of the BRP and in IREV. In general the number of members absent rose faster than the size of the group, However, this effect could only be shown statistically for IREV, where it was signifi- cant at the .01 level by 2. In terms of proportions, absenteeism rose from 10% at the lowest size level to 25% as the upper size levels were reached. There was also a statistically significant (.001) inverse rela- tion between absenteeism and leaving. At high size levels, although leaving was the major regulator, its effects were damped by the effects of absenteeism.

In the simulation, only the major regu- lator of leaving behavior was used; hence the slightly higher join/leave rate in Fig. 7. In IREV, the correlation between inactivity and leaving was significant at the .001 level (2). In the figures, 18 of those who left had a below average level of involvement in terms of person-hours, while only 2 had an above average level of involvement. This was written in as a simulation rule, and

/-.L--- _- ..... /- ....... /. .........

~. ~ .. , ... *,< .... <... , <..’

I I I I I I i 20 40 60 80 100 120 140

TIME (IN W E E K S ) - FIG. 6. Group IREV: Joining and leaving rates; resultant sizes.

Behavioral Science. Volume 26, 1981

124 MIKE ROBINSON

........... , ............... I .--

0 ’ I I i 20 40 60 80 100 120 140 16a

T IME SIMULATED W E E K S )

FIG. 7. GROUP-I simulation: Joining and leaving rates; resultant sizes.

operated as a satisfactory regulator in sta- ble runs. More interestingly, the simula- tions also drew attention to (by reproduc- ing) a facet of the original group history that had not, at first, been considered im- portant, the “early leaver phenomenon.”

The “early leaver phenomenon” is well known in industrial situations, and was originally discovered by Rice, Hill, and Triste in 1950. It now seems that this pro- cess also characterizes social groups: The last to join are the most likely to leave, and, conversely, the longer members stay in the group, the less likely it is that they will leave. We have already noted that size in- crease means there is less available activity per member. In practice, the activity avail- able to less active members is reduced more than that available to more active members (see Fig. 8 for the activity distribution in IREV). Less active members tend to be newer members, and thus newer members have a greater propensity to leave. This effect can be clearly seen in Fig. 9. The advantages of this arrangement are that, in Roger Barker’s terms, the storage of the proper program, which requires time and experience to establish, is not disrupted. In our terms, it means that those members with operational knowledge of techniques are the least likely to leave.

In the GROUP-1 simulations, stable con- ditions were most frequently achieved when new members entered at a randomly

disturbed constant rate, and had randomly distributed levels of commitment-very much the sort of situation that groups en- counter in real life. Four typical cases of new entry can be described:

If the new member has mismatched pref- erences and a low level of commitment, he is very likely to leave.

If the new member has mismatched pref- erences and a high level of commitment, he is likely to be assimilated after changing his preferences and lowering his level of com- mitment.

If the new member has matched prefer- ences and a low level of commitment, he may either be assimilated or leave (after a

7 M E A N NUMBER OF INDIVIDUALS

’1 \+ \

‘* 0 I I

2 4 6 8 lo+

ACTIVITY LEVEL (HOURS RR WEEK)

FIG. 8. Group IREV: Distribution of activity among group members.

Behavioral Science. Volume 26. 1981

THE IDENTITY OF HUMAN SOCIAL GROUPS 125

LEFT ’I] 5 NUMBER

TIME IN GROUP ( W E E K S ) FIG. 9. Group IREV: Distribution of time-in-group before leaving.

period of inactivity), depending on whether the group is near its lower or upper size limit.

If a new member has matched prefer- ences and a high level of commitment, he is almost certain to be assimilated, but may displace older members. Provided such oc- currences are relatively infrequent, prefer- ence set and activity are not disrupted.

In stable conditions, these processes re- sult in the appearance of the characteristic “early leaver” pattern, an example of which is shown in Fig. 10. The processes underly- ing other types of leaving pattern force out (or prevent the establishment of) older members, and cause variability in the pref- erence set and instability in the group.

Unstable states

The GROUP-1 simulation was especially useful in identifying unstable states, since these are unpredictable and difficult to ob- serve in natural groups. Several “danger points” were identified.

(i.) Formation. When a group first forms, in the simulation as in real life, there is a good chance that preference profiles

‘ O l

wil l be radically dissimilar, even where the first choices are congruent. Unless the av- erage level of commitment is very high, any upset or failure leads to the termination of activity, the dissolution of any emergent norms, and the disintegration of the group. This outcome was a common simulation result (preference orderings for “individ- uals” were randomly assigned), and was also observed in three of the seven sub- groups of the BRP. The Pamphlet, Propa- ganda, and Coordinating groups never agreed on aims, techniques, or activities (sometimes, when there was agreement, the activities were “usurped” by other groups), and disintegrated after very few meetings. Mapped onto the model, this simply means that the core of stability, the normative structure, failed to gell.

(5.) Influx of new members. This source of disruption had several variants, all of which had a high probability of wrecking the group.

(a.) If the entry of new members was tied to the activity level of the group (even with a random time lag), the group almost al- ways became unstable. This could take two forms. In the first, the group crystallized its

TIME IN GROUP (SIMULATED WEEKS)

FIG. 10. GROUP-I simulation: Distribution of time-in-group before leaving.

Behavioral Science, Volume 26.1981

126 MIKE ROBINSON

activities below the level of social visibility. No new members were attracted. This meant that, although the group might con- tinue for some time, any member loss would result in serious activity disruption, loss of commitment, and disintegration. In the sec- ond case, high activity levels resulted in rapid growth to well above “normal” size bounds. High activity (and hence further growth) would be maintained with a very low average level of commitment. Inactivity would grow rapidly. Whole subgroups would suddenly drop out. Having been in the group for some time, these subgroups would have influenced the preference struc- ture (and activity pattern) in their own direction. Their disappearance disrupted activity. The low level of commitment of the remaining members was usually insuf- ficient to continue the activity, or to toler- ate the process of normative restructuring. More inactivity was followed by unreplaced member loss, and usually by disintegration. This process underlies the simulated group imaged in Fig. 11, and was very similar to the history of the BRP Squatting Group (although the latter had the additional haz- ard of an externally imposed block on activ- ity). The instability effects of tying joining rates to activity levels point to the need for a third type of size regulation mechanism, If “entry requirements” are imposed, tightened at high size levels, and relaxed at low size levels, this form of instability is much harder to generate. If social visibility is assumed to be a factor affecting joining

rate, then this third mechanism reduces the fluctuations to a (randomly disturbed) con- stant rate that can be regulated by the other two mechanisms of absenteeism and leaving. In terms of the simulation, the “entry requirements” were represented by “degree of match’ between preference pro- files of would-be and existing members. The “natural” counterpart of this will be dis- cussed in the concluding sections.

(b.) A sudden influx of new members, and/or (c.) a consistent influx qf “individ- uals” with a high “level of commitment.” Both of these events had the effect of changing the established preference struc- ture towards the preferences of new mem- bers. Usually this resulted in a new pattern of activity-and older members left after being deprived of their “favorite” activity. The formation danger point is repeated. It is almost certain that the deeper preference of old and new members will be dissimilar. Compromise activities are initiated that are unsatisfying to both. Inactivity and disin- tegration follows. Sometimes the simulated groups recovered from an influx of new members after preference and activity re- structuring, but a consistent influx of highly committed individuals would cause succes- sive crises terminated only by the collapse of the group.

(iii.) Activity blocks originating outside the group. Both the natural and simulated groups were resilient in the face of occa- sional activity disruption. Neither could tol- erate prolonged blocks on their major activ-

11, .... ..*.... ,

.... v)

SIZE

0 5 10 15 20 25 30

TIME (SIMULATED WEEKS) -> FIG. 11. GROUP-I simulation: size and activity against time in an unstable group.

Behavioral Science, Volume 26. 1981

THE IDENTITY OF HUMAN SOCIAL GROUPS 127

ities. In natural groups such blocks arise from many causes: lack of finance, competition for resources, deliberate inter- vention by an external agency, and so on. In the simulated groups, blocks were cre- ated simply by raking the person-hour (“level of commitment”) requirements of major activities to impossible levels. In all cases the re-orientation of the normative structure required to establish a new activ- ity pattern was so drastic that members became inactive and left long before it could be achieved.

In the simulated groups, where the nor- mative structure was the preference struc- ture, and variables could be easily manip- ulated, crises were easy to achieve. In nat- ural groups, even with longitudinal studies, crises were rather rare. None were observed in IREV; only one (albeit terminal) in the BRP; and an interesting “near-crisis” was observed in one of the short-term experi- mental groups. Since disruption of the nor- mative structure is a central feature of in- stability, it is worth recounting these two events to illustrate the process.

In the BRP, a change in one central norm was quickly followed by the collapse of the group as a whole. The BRP functioned as a set of cooperating subgroups. Decisions on which techniques to actualize, and how fre- quently, were the prerogative of each sub- group. The subgroup that specialized in squatting, the extra-legal placement of ten- ants in empty properties, developed refined techniques in this area. After a period of time, the Municipal Authority took action that blocked these techniques. This pro- voked a crisis in the subgroup-members were deprived of activity. The BRP as a whole attempted to reorientate itself by centralizing decision making in an executive committee. The norm of subgroup auton- omy was annulled. Many members, espe- cially those in the Squatting Group, refused to accept the change. Faction fighting and argument disrupted the activities of the BRP as a whole, and especially those of the Rents Registration Group, the most viable subgroup. Inactivity resulted in members leaving. The group diminished in size to the point where former techniques could not be actualized. The group collapsed.

The second instance is taken from an experimental group set up to examine ef- fects of membership change on normative set. Group members were placed in a situ- ation in which they could win money, or lose chances to win money. Rewards, pen- alties, and the “elimination” (removal from the group) of members were randomized. The randomness of the situation was not disclosed to group members. They were left to “discover” the rules governing the distri- bution of rewards and penalties, and to devise their own norms. A stable norm set developed that was rational in terms of the underlying structure of the situation. Mem- bers attempted, cooperatively, to maximize their rewards by minimizing group size, and by distributing activity (and hence rewards) evenly among all members. This pattern almost suffered a catastrophic (in the sense of Zeeman, 1976) disruption when a sub- group formed that decided it was “OK to double-cross newcomers.” (As the group, not the experimenter, explained the “des ,” this was easily possible.)

From a strategy of equal reward in a cooperative small group, members teetered on the brink of changing to a strategy of exploitation in a large group. At one point (on the normative strength measure) the exploitative faction were in the majority. In terms of perceived majority belief (norma- tive structure), they did not realize this. With further turnover of members, the co- operative norm recovered. It was, however, clear that a change in this central norm would have changed all norm, and resulted in an entirely different mode of functioning.

In general, we can say that the normative structure and techniques of any group are maintained by the success of the activities that they prescribe. Conversely, two types of event threaten the normative structure. Size levels above “normal” limits entail an inactive subgroup that is not satisfied by normal activity. The ensuing division of opinion may become public (normative strength effects normative structure), and significant changes in functioning (usually collapse) occur. Similar effects occur if ac- tivity is blocked by an external agency. Thus we see that when group identity is maintained, normal size, normal activity,

Behavioral Science, Volume 26, 1961

128 MIKE ROBINSON

and normative structure are inextricably linked.

Summarizing, we can say that if a group survives its formative period, it is likely to become stable. Change, in the form of growth, retructuring, or collapse, tends to follow from external factors: member influx or activity blocks.

COMMENTS

The dual system of size regulation, absen- teeism and inactivity/leaving, is of partic- ular interest as it adumbrates two distinct growth strategies. Organizations concerned with the development of an activist base commonly impose a schismatic require- ment on their branches. Units of a certain size must be split into two new branches. In this way overall organizational growth can be maintained while the stability re- quirements of each subgroup or branch con- tinue to be met. In the absence of, and sometimes despite an overall organization, schism in the traditional sense (irreconcil- able factionalization) is a common conse- quence of growth beyond “normal” limits. Schism, intentional or not, corresponds with our model. The notion of steady linear growth (held by many group organizers, including those in IREV) does not corre- spond to the model, and we know of no case where it has worked in practice. Growth in activist organizations is a discontinuous and a risky matter.

A second and more common growth phe- nomenon is the development of the absen- teeism size regulator. The simulation showed this to be less powerful and reliable than the inactivity/leaving regulator, but this axiomatic simulation relation may not hold in all cases. The absenteeism regulator undoubtedly operated in two natural groups, IREV and the Rents Registration Group of the BRP, but in simulated condi- tions it tended to cause groups to exceed their upper size bounds and become unsta- ble. This could be taken as an indication thzt, in developmental terms, emphasis on the absenteeism regulator would lead to large organizations with a small active nu- cleus. Although, to date, only informal observations have been made, the indica- tions are that this developmental variant is

common in the branches of established po- litical parties and trades unions. Broadly speaking, large increases in membership can be tolerated if they do not disturb the effective size of the active group. Beyond a certain level, absenteeism rises exponen- tially with size. Although the exact nature of the exponential function has not yet been determined, the following formula has been found to work well in predicting attendance at meetings for unitary groups in the 60- 800 size range:

where A is attendance A = R + M1’2,

R is the number of

M is the total mem- formal roles

bership. A third size regulating mechanism is that

of “entry requirements.” Here a level of agreement with group norms and values is required before individuals can be admit- ted. These requirements become stricter as size increases. This mechanism was not tested for in the studies reported here, but has been amply supported by the work of Roger Barker (1968) and others. In our model it would augment the size regulators in the event that the constant joining rate assumption were violated. It is mentioned here as it forms the basis of a third “higher” group strategy for size and activity main- tenance, best exemplified by the selection committee procedures of various exclusive clubs and societies. Under this strategy, member activity is maximized and leaving minimized. The application rate is itself a function of the group’s success and social visibility. This third regulatory mechanism is consistent with our contant joining rate assumption. It can be regarded as a special and limiting case of early leaving.

Lastly, it is of interest that the “early leaver phenomenon” was found both in the informal, natural groups, and in their sim- ulated counterparts. We have explained earlier how this forms a coherent part of the system of group stabilizing mechanisms, essentially by safeguarding the expertise of longstanding members. Here one wonders if the anxieties of industrial managers about early leavers are ill-founded. It may well be

Behavioral Science, Volume 26, 1981

THE IDENTITY OF HUMAN SOCIAL GROUPS 129

that a “cure” for the problem would be a disaster for the organization.

In conclusion, we may say that the st8- bilizing systems in informal, human groups seem both elegant and eminently suscepti- ble to cybernetic analysis, which in turn may be able to provide a descriptive and diagnostic basis for a more general under- standing of group processes. Herbert Simon (1969) once reflected that animals, viewed as behaving systems, were quite simple. The apparent complexity of their behavior over time was largely a function of the complexity of the environment in which they found themselves. It seems that this observation is also true of human social groups. Their apparent diversity and com- plexity has more to do with their diverse intentions and environments than with the fundamental mechanisms that enable them to persist through time-to maintain their identities.

REFERENCES Barker, R. G. Ecological psychology. Stanford, Cali-

fornia: Stanford University Press, 1968. Bates, A. P., & Cloyd, J. S. Towards the development

of operations for defining group norms and member roles. Sockmetry, 1956,19,26-39.

Homans, G. C. The human group. New York Har- court, Brace, and World, 1950.

Homans, G. C. Social behavior: Its elementary forms. New York Harcourt, Brace, and World, 1961.

Newcomb, T. The acquaintance process. New York Holt, Reinhart, & Winston, 1961.

Rappaport, R. A. Pigs for ancestors. New Haven, Connecticut: Yale University Press, 1968.

Robinson, M. J. Human social groups. Doctoral dis- sertation, Brunel University, United Kingdom, 1977.

R i c e , A . K . , W , J . M . M . , & T & t e , E . L . T h e representation of labor turnover as a social process. Human Relatwm, 1950,3.

Simon, H. A. The sciences of the artificial. Cambridge, Massachusetts: MIT Press, 1969.

Zeeman, E. C. Catastrophe theory. Scientific Ameri- can, 1976,23465-83.

(Manuscript received April 24, 1978; revised Novem- ber 1, 1979)

Behavioral Science, Volume 26.1981