SOCIAL PROBLEMS, Vol. 8 ( 2 ) , 102-111, 1960.

Initial Interaction of Newcomers in Alcoholics Anonymous: A Field Experiment in Class Symbols and Socialization.

John F. Lofland University of California, Berkeley Robert A. Lejeune Columbia University

This paper reports an exploratory field experiment on one aspect of one of the alcoholic therapy organizations, Alcoholics Anonymous. A.A. enjoys the reputation of being one of the more effective organizations in helping alcoholics arrest their sickness. Attention has naturally turned to the problem of why a larger number of alcoholics do not affiliate with A.A. (5,8). We are concerned with the sociological features of this problem and specifically with what features of the social structures of A.A. groups may facilitate or deter affiliation. Field observation in about half of the approximately seventy A.A. groups in the Manhattan Borough of New York City revealed that A.A. groups are quite heterogeneous in their social class composition across groups but that within groups they are relatively homogeneous. This we took to be an important structural fact. It seems clear that members, to some extent, self-select themselves to groups of approximately their own social class level. However, beyond selfselection, we wondered on the one hand, if A.A. groups themselves differed by social class on their reception of new persons to A.A. and on the other hand if newcomers might not be received differently depending on their own social class (6,pp.39-40; 7,p.llS). – – An~alcoholicwho is not in an institution may attend his first A.A. meeting in various ways. One of the more frequent ways is that he decides on his own, or through the advice of a friend, to try A.A. He enters his first meeting alone and without knowing anyone in the group or in A.A. We are concerned with this class of first contacts with A.A. and specifically with some of the properties of this phenomenon in large urban centres. In an urban setting the objective social class ranks of actors have low visibility. In lieu of viewing the actual ranks which determine an actor’s social class, symbolic means of communicating social class occupancy develop. These social class symbols select for a given actor the social class that is to be imputed to another actor.* Alcoholics alone at their first A.A. meeting, where they * This formulation is taken from Goffman (4). Form and Stone (3) have conducted a study indicating that symbols are used to infer social class and discuss which symbols are used to infer which classes.

know no one and before they have spoken to anyone, are in a situation where the “others” present must assign social class to them on the basis of a limited number of social class symbols, mainly those presented by the actor’s clothes, postural behavior, grooming and ethnicity. Those social class symbols presented by one anonymous actor to another before they have spoken we shall call an actor’s presentation. A.A. norms call on members to seek out those at meetings whom they perceive as newcomers. The decision by a member to approach a newcomer may well be, to a large extent, a function of the positive or negative evaluation by the members of the presentation of the newcomer. The latter’s presentation may thus determine which newcomers become socialized into A.A. and which do not. Newcomers, of course, have the option of initiating interaction with members, but most are probably not disposed to do so due to their unfamiliarity with A.A. and – if member accounts of their first meeting are at all general – because they come to their first meeting with feelings of fear and anxiety. We, therefore, assume that newcomers, though they may desire interaction with group members, are not likely to be the initiators of this interaction. On the basis of our field experience in A.A. we feel safe in saying that if there is to be interaction between members and newcomers at these first meetings, then in most cases the members must take the initiative. The experiment reported below deals with three variables. Social class presentation of the group and social class presentation of the newcomer are the two independent variables and initial socialization of newcomers is the dependent variable. The experiment consisted in sending six male agents in different social class presentations to A.A. open meetings where they posed as alcoholic newcomers. We term the agent and group class level as “high” and “low” but by this we do not mean “high class” or “low class”; but rather, that relative to one another the levels are high or low. Since the two independent variables are taken at two levels only, the experiment has four conditions (or treatments): 1) agents in low social class presentation attending low social class A.A. groups, 2) agents in low social class presentation attending high social class A.A. groups, 3) agents in high social class presentation attending low social class A.A. groups and 4) agents in high social class presentation attending high social class groups. The next section reports the methods of determining high and low A.A. groups, and is followed by a description of the content of the agents’ treatments of the groups. Following these are the means of measuring initial socialization, the design employed to reduce extraneous variation, the results and a discussion of the results. On the basis of the notion ubiquitous in sociology that actors of like rank interact more than actors of unlike ranks, we, before undertaking the research, stated the general hypothesis that if the newcomer and the A.A. group display similar social class presentations, then the initial socialization will be higher, and conversely, if the newcomer and the A.A. group display different social class presentation, then the initial socialization will be lower.

RANKING THE GROUPS BY SOCIAL CLASS Although we had reason to believe that A.A. groups in Manhattan were dispersed on social class composition, we wanted to be more precise and state more exactly the nature of these differences. During the field work we developed a series of direct observation indicators that appeared to specify, at least in part, the notion of social class as it is presented to the observer who has no more than simple observational knowledge. From these we built a weighted index, which is similar in its logical index to Chapin’s Living Room Scale (2). It should be made clear that the ranking of the groups was by the social class symbols they displayed. We assume that the symbols currently associated with specific social classes are not subject to such anomalous display (l,pp.158-163), that any group’s index score would have no correlation with the average objective class ranks of the people composing the group. Our indicators fell into three main groups: 1) individual properties of individual A.A. group members (e.g. style of dress and grooming), 2) properties of the group’s immediate action (e.g., amount of money in the collection plate and quantity and quality of food served after the meeting) and 3) properties of the meeting place (e.g., condition of floors and walls). To help keep the sample homogeneous on variables other than social class, we decided to rank only those meetings that 1) started at 8:3OP.M., 2) were open meetings, 3) not in institutions, 4) were not discussion groups, 5) were ninety per cent or more “white” members. (there is racial self-segregation in A.A. ), and 6) were attended by both sexes. In other words, we limited ourselves to what is known in A.A. as “typical A.A. meetings.” At the time of the ranking, twenty-one groups met our specifications. We attended a meeting of each of the groups and independently scored them on the index. Apart from this, before calculating the score, we gave the group an independent intuitive relative rank. No attempt was made in the intuitive ranking (or by the index) to assign the groups to social class categories only to determine which groups stood higher or lower than others. As a further check, we asked an informant member of a Manhattan A.A. group to rank the groups in our sample by their social class. He was familiar with, and was able to rank, fourteen of the twenty-one groups. We found it interesting that the informant had never attended over half of the fourteen groups that he could confidently rank. We took this as an indication that specific groups have informal “statuses” in A.A. (reputations) with which he had become familiar as a result of being an A.A. member. He reported that some of these groups were referred to in A.A. as “snob groups,” where piped music, showmanship and furs were the norm at the meetings, whereas other groups, often referred to as “real A.A.”, were held in modest surroundings and were composed mostly of working and lower class members. Five rankings of the meetings resulted. Table one presents the intercorrelations among the rankings.

TABLE 1 INTERCORRELATION AMONG THE FIVE SOCIAL CLASS RANKINGS OF A.A. GROUP Observer 1 Observer 2 Index Intuitive Index Intuitive Informant .65* .67* .71* .79* Observer 1 Index -90 -86 .90 Intuitive Observer 2 Index -89

All correlations computed by Kendall’s Tau. * Based on correlation of 14 groups; all others based on 21 groups. The intercorrelations among the observers are all fairly high. The correlations with the informant are lower. Examination of the various rank orders reveals that the correlations are reduced by rather small rank shifts of the groups between the observers and the informants’ ranks. The shifts are smallest at the extremes and larger for the ranks of the groups in the middle. Because of this, we took as our sample the six highest and the six lowest groups. There was no disagreement in delineating the extremes; the difference between the six highest and six lowest was quite striking. However, since we could not reliably or confidently order all of the groups within these extremes, they are treated as being simply “high” or wlow.n THE MANIPULATED TREATMENTS: PRESENTATION AND ACTION The manipulation we imposed upon the high and low groups consists in preparing the agents in certain presentations and having them perform in certain ways while they were in the meetings. Presentation. It appears to us that their are four basic styles of male dress in contemporary urban society (1960), that these styles are, of course, seen in A.A., and that they are associated with certain broad social classes in U.S. society. To describe these styles exactly is a difficult and lengthy effort, but somehow most of us are able to discern them. The presentations that actors make are composed of facts of so great a number and variety that delineation in terms of a few attributes is quite difficult. We believe that each of the four types briefly described below are associated not only objectively but also perceptually in most people’s inferences to social class from observing anonymous actors. Evidence of the perceptual association of types of actors presentation is outlined in Form and Stone (3).

Salient descriptive attributes of male presentation types: 1. Upper-middle and Upper Class: well groomed, clean, latest style clothes (esp. suits and ties), subdued coloured clothes, which are neatly pressed, of good quality, not worn and are of matching colours (using as criteria the present man’s fashion advertisements). 11. Lower-middle Class: less well groomed, clean, wearing suits and ties but of out-dated style, colours not so subdued, clothes showing some wear, fair to poor quality, less well pressed, articles of “clashing” colours. 111. Working Class: not so well groomed, clean, wash trousers, no suit coat (a waist jacket, usually) no ties, “clashing” colours. IV. Derelict: poorly groomed, dirty, dirty wash pants, suit coat (usually out of style, worn, unpressed, dirty). Since the Type IV presentation is confounded by the fact that occupants are usually older than any of our agents were, we choose to use the Type I and Type I11 presentations as the values on the variable of agent presentation. Type I is called the “high presentation” and Type I11 we call the “low presentation.” The agents themselves were all white male graduate students in sociology, primarily of upper-middle class origin and ranging in age from 24 to 34. Action. Instructions for agent action at the meetings were of two kinds: what they were to do physically and what they were to say when interacting with a member. agents arrived alone at a meeting at 8:20. In the first week of a group’s treatment they sat mid-way in the room in the centre of a row of chairs on the right side. In the second week, they sat on the left side. They were carefully instructed not to initiate interaction with anyone and to sit through the meeting looking tense and uncomfortable. If they were contacted at anytime they were serious and sincere, polite but not gracious. Following the end of the speaking, they went over to the literature table and browsed through the literature for five minutes. If no contact was made they went over to the main concentration of members where coffee was being served and stood around in the concentration for fifteen minutes. They did not take coffee or smoke during that time. If, at the end of this period, there was no contact, they left the meeting. When contact was made, the agent was instructed to let his movement be directed by the member. In connection with other work, we had field notes available on what went on in first contacts at A.A. meetings. From this we isolated the probes that the agents could expect when they were contacted. Our problem was, interestingly enough, not in passing the agents as alcoholic newcomers – this was more often than not assumed in interaction – but to give them standard responses that would not get them too involved with the members, but at the same time would not put off the members. The responses to probes on alcoholism, A.A. familiarity, where they lived, their marital status and their phone numbers were the same for all conditions. We

made them as vague but as plausible as possible. The only item that changed was the occupation they gave if they were asked, which varied appropriately by high and low presentation. All agents used assumed “American” names. We role-played with the agents before their meetings to get them used to their status and familiarize them with their responses. In interaction, the burden of carrying the conversation forward was laid completely on the member. The agents were purposely bland, as one might expect in an alcoholic who is new to A.A.

Three kinds of measurements were made on each visit to a group: 1) time spent in interaction with members, 2) numbers of persons in contact, and 3) interaction content. Each agent was equipped with a cumulative pocket stop watch. He kept this in his pocket at the meeting and switched it on while someone was talking to him. The resulting figure was the group’s total time in interaction with the agent in that trial (visit). The agent was instructed to consider himself in interaction when an individual was attending to him whether listening or talking or if he was part of a larger system (a dyad or larger in which actors addressed comments to him and/or to the group). If the conversation lulled but had the direct potential of resuming, he was to consider himself in interaction. From the field notes of some twenty earlier observers in A.A., we developed a list of interaction content that appeared to be particularly relevant to the socialization of the newcomer. They were set up in dichotomous form so that the agent upon leaving the meeting could score whether each member said the item or not. A separate form was filled out for each member who interacted with the agent. In addition, other characteristics such as age, sex and presentation type of interacting members were noted. The agents filled out the forms as soon as they left the meetings and arrived home. The next morning the forms and the qualitative characteristics were reviewed with one of us. Ideally, each agent would have carried a hidden recorder from which the content could have been scored more accurately and reviewed in detail. However, we found that this was not feasible. THE DESIGN

To partial out the structural effects of group social class composition and agent presentation type, it was necessary to design the agents’ visits to groups such that 1) meeting-specific variations (size, idiosyncratic members, etc.) apart from any effects of their social class composition, would not obscure the experimental manipulation and 2) agent-specific variations (age, physical features, personality, etc.) would not obscure the fact of their differentpresentations. In addition to these, the design had to be such that the possibility of the agent being seen in different presentations by the same members was minimized. To reduce the meeting specific variation, each of the twelve

meetings was treated (visited) with a high and a low agent. To reduce agent-specific factors each agent administered the four treatment conditions. On substantive grounds we decided that each agent should attend a meeting two weeks in a row, since he might not be recognized as a newcomer in the first visit or if he was, we could see how the members followed-up contacts in the second visit. Time, economy and the possibility of the agents being seen by members in a different presentation led us to assign each of the six agents to four groups under each of the four conditions and to have him visit these four groups during two consecutive meetings, for a total of eight visits per agent. Thus, each of the twelve groups was visited four times, twice ( on consecutive weeks) by one low agent and twice by one high agent. The experiment was executed in two two-week phases. In the first two weeks three of the agents administered high treatments and three low treatments. In the second two weeks they switched presentations. Also, in the first two weeks three of the low and three of the high groups received low treatments and the other three of the high and low groups received high treatments. In the second two weeks each group received the opposite treatment. The agents were rotated in this manner to help control for possible effects on agents due to the fact that they administered the high or low treatment first. Likewise, groups were allocated so that one half were first treated with high and one half with low to help control for any effect due to the order in which they were treated. One visit by one agent in one presentation to one meeting is considered a trial; therefore, there are twelve replications (N=48). The rotational patterns outlined above were set up formally; groups and agents were randomly assigned to them. RESULTS Before presenting the results we must specify more clearly our concept of socialization. If socialization is viewed in terms of the dimensions determining its effectiveness, and if we only consider the dimensions generic to A.A. groups, we may, for these purposes, specify one broad dimension determining effective socialization. Effective socialization into A.A. is defined as the newcomer maintaining relative sobriety and becoming active in some way in A.A. The dimension is a given A.A. group’s degreeofactivitydirected towurd linking the newcomer into the socialsystem.Amount and kind of activity are considered here simply as degree. At any point in time the dimension has a value. The unit used in speaking of the value of the dimension is one attendance by a newcomer. This value, which results from an attendance, we shall call an activity outcome. We are concerned with initidactivity outcomes. A reformulation of our hypothesis in these terms reads: If the newcomer and the A.A. group display similar (different) social class presentations, then higher (lower) initial activity outcomes will occur. Our data are accurate on three indicators of the value of initial activity outcomes: 1) number of members interacting with the agent., 2) amount of time they spend in interaction and 3) amount of commitment they make to help the agent. The data trend on

each of the three indicators, contrary to our expectations, does not support the hypothesis. As shown in Table 2, the agents.. interacted with 118 members* over the forty-eight trials,** of which sixty four percent were in the “incongruent status” conditions. The low agents in the low groups interacted with 24

TABLE 2

Relative Group Rank Low High ($1 (%I Agent Low 24 32 56

Presentation High 32 12 44

per cent of the total members contacted and the high agents in the high groups had 12 per cent of the total. This relationship is significant by the xZ test ( x a =9.88, .01>P>.001). This, of course, could have been due to longer interactions in the congruent status conditions which would have precluded a large number of interactions. Table 3, which shows the total interaction times for

* Member interactions lasting less than one minute in which the verbal content was no more than just asking directions, an “excuse me” in a crowd, asking the time of day, etc., were considered irrelevant to activity outcomes and not counted. ** In the analysis, the 48 visits are considered independent trials. We had hoped to study member follow up in second visits, but each second visit turned’out to be a first visit in terms of contacts. The agents usually in the second week did not see those present who contacted then in the first week. Those few that they did see did not seem to notice the agent.

each treatment condition, indicates that the relationship is the same as in Table 2, the number of members interacting. Table 3 shows the actual amount of time in interaction for all visits in each condition relative to possible amount of time in interaction.

TABLE 3 TOTALTIME A.A. MEMBERSINTERACTED WITH AGENTS IN THE FOUR CONDITIONS (IN MINUTES)-. .-.-. -. – – .- — – — .- –..- — – — – .— — .. …- – .. . Agenc Presentation LOW — – HIGH Relatlve Group LOW’ HIGH LOW HIGH R ~ K (53) (5%) (Cr) (5) (96) — T~me in 9-‘ 155′ 115′ 42′ 409’ Interxrlon (40) (65) (48) (1s) (43I

The incongruent status conditions have the higher proportion of time in interaction, but the difference between the high and the low agents in the low groups is not great. The high agents in the low groups were in interaction 48 per cent of the time while the low agents in low groups were in interaction 40 per cent of the possible time. The most striking difference is between the agents in high and low presentation in the high meetings, where the low agents were in interaction 65 per cent of the possible time and the high agents 18 per cent of the possible time. The Kruskal-Wallis rest on the ranks of the individual group’s interaction times by treatment is not significant at the 5 per cent level (H = 6.36, .10>P>.05). However, substantively, we judge differences of this magnitude to be important. The interaction content between the members and the agents usually involved a number of the following topics: 1) giving information about:a. A.A. b. the members way of working the program, c. their alcoholic backgrounds, and d. their non-A.A. statuses; 2) giving suggestions about: a. working the program, and b. the attitude the newcomer should take toward his alcoholism; and 3) seeking information about: a. the newcomer’s familiarity with A.A., b. his alcoholic condition, and c. his non-A.A. statuses. On the content form, the number of times that most specific content

items were reported as occurring in each condition was roughly proportional to the number of persons interacting in each condition. We had expected to do an analysis of the interaction content of the members by experimental condition to ascertain any class differences by group or agent presentation, but the N turned out to be small and the difficulties in measuring accuracy were such that we have little confidence in most of the differences. Two of the items, on which accuracy was excellent due to their salience, were judged to be particularly important as indicators of initial activity out comes. They were, the member giving the agent his phone number and inviting the agent out for coffee after meeting (agent politely declined). These are actions that commit members beyond the open meetings. on the basis of substantive knowledge of the norms of A.A., all members who gave phone numbers and/or asked the agent out for coffee are considered committed. Thirteen per cent of the total number of members committed themselves (Table 4 ) .

TABLE 4 COMMITMENT: NUMBER OF A.A MEMBERS INVITIKGAGENTS OVT FOR COFFEEOR GIVINGA PHONE NCMBER IN THE FOURCONDITIONS – — — – — – -.— — – PA– — . — – -. – Agent Prrsentation LOW HIGH — — —– -. — – Rtlative LOW HIGH LOW HIGH Group Rank No. (5) No. ( % ) No. ( No. ( % i – Members 2 9 3 1 15 Committed (7) (24) (8) (7) (1’1 — Members Not 26 29 35 13 103 Committed (93) (76) (92) (93) (S-1 Total 28 38 38 14 118 Members (100) (100) (100) (100) (100) _/

The two presentations in the low groups and the high presentation in the high groups have about the same proportion of members committing themselves to the agents. Compared to approximately 7 per cent in the high group had 24 per cent of the members commit themselves. Although the relationship is the same as on the other measures, the difference is ~ o tstatistically significant at the five per cent level by the X test (x’ = 6.09, .2O>P.10). Overall, the data indicate consistent differences in the initial activity outcomes of A.A. groups as a function of the social class level of the group and the class level of the newcomers’ presentations. Although any general statement must be very tentative, for the present, we state that initial activity outcomes seem to be highest where the A.A. group is relatively high and the newcomers relatively low in social class. The outcome

differences that exist between the response of the low groups to high and low newcomers are very small, but consistently the outcomes for the high presentation newcomers are higher. The lowest activity outcomes occurred in the high groups where the newcomer is in high presentation. DISCUSSION We offer a postfacturn explanation as to why our hypothesis was not supported. First, all newcomers who are incongruent with the class level of the groups they attend are physically more visible. We suspect that lower class newcomers do not normally appear very often in the high groups, so that visibility is especially high for them. Further, the high presentation newcomers in the high groups have low visibility, lower even than the low newcomers in the low groups, by virtue of the confounding effect of the larger size of high groups. Most of the high groups had attendances of over one hundred persons while most low groups had attendance well under one hundred.* Operating together, these factors probably reduce the initial activity outcomes for the high groups in relation to the high presentation newcomers. Secondly, for the case of the low presentation newcomers in the high groups, we would say that the normative obligation on A.A. members to help other alcoholics is evoked with greater strength here because of the tendency to associate lower class presentation with the more acute need to be helped. In addition, it could be that some subtle cues still exist in the agents’ presentations that conveyed to the members in the high groups that the agents were of a higher social class origin. Upper-middle class training and many years of higher education is indeed difficult to completely cover. However, we do not doubt that the agents were perceived as being presently of a lower class. We understand that among members of high groups there is a particular concern with reclaiming alcoholics who have fallen in class rank. Third, for the case of the high presentation newcomers in the low groups, we suspect that members who are of a lower class derive some sense of importance from introducing a member of a higher social class to A.A. A.A. presents a system reference where the lower class members can, at least initially, be of higher rank by virtue of his length of time in the system. In closing, we must note that only initial activity outcomes were dealt with in this experiment. We do not know the values of activity outcomes over time for different types of meetings. It may be that activity outcomes are quite different in later stages. On the newcomers’ side of the process, we do not know how given newcomers react to given activity outcomes. Trice reports (5) that high activity outcomes contribute to effective socialization, but it also may be true that some newcomers are alienated by high activity, and prefer to be left alone – at least initially. We pose these problems as topics for further research.

*This fact was controlled for in comparing effect of presentation by the equal treatment of all groups by all presentations.

REFERENCES 1. Barber, Bernard. Social Stratification. (New York: Harcourt Brace, 1957.) 2. Chapin, F. Stuart. The Measurement of Social Status. (Minneapolis: University of Minnesota Press, 1933.) 3. Form, William H. and Gregory P. Stone. “Urbanism, Anonymity and Status Symbolism,” American Journal of Sociology, LXII (March, 1957), 504-514. 4. Goffman, Erving. “Symbols of Class Status,” British Journal of Sociology, 2 (December, 1951), 294-304. 5. Trice, Harrison M. “The Process of Affiliation with Alcoholics Anonym~us,~’ Quart. 3 . Stud. Alc., 18 (March 1957), 39-54. 6. —————– . “Alcoholism: Group Factors in Etiology and Cure,” Human Organization, 15 (Summer, 1957), 33-40. 7. —————– . “Alcoholics Anonymous,” The Annals of the American Academy of Political and Social Science, 315 (January, 1958), 313-320. 8. —————– . “The Affiliation Motive and Readiness to Join Alcoholics Anonymous,” Quart. J. Stud. Alc., 20 (June, 1959), 313-320.