Aspects of the criterion problem in small group research

Aspects of the criterion problem in small group research

North-Holkmd P,:tbli&ng Co., Amsterdam Acttl ~s~~~~~~~ 25 (1966) 101431; by pbatoprint or mkrofilm AS without written permission from the publisher...

3MB Sizes 1 Downloads 19 Views

North-Holkmd P,:tbli&ng Co., Amsterdam

Acttl ~s~~~~~~~ 25 (1966) 101431; by pbatoprint or mkrofilm

AS

without written permission from the publisher

OF THE CRITERION PROBLEM

GROUP RESEARCH AINS TO BE STUDIED

IRWIN ALTMAN * Institlttc, ~fffi~nul Naval Medical Center, Bethesda, u&and, ?XA

ne& ~~t~~~~~~ ~~~esti~~for any system, whether it be a small zation, or even an automated assembly line is: As any scientist will agree, whether he

ic laboratory environment or interested in finding solutions to practical problems, the criterion question is an im~rt~nt and difkult one. Unfortunately, the applied researcher to whom this problem is salient finds the press of time only allowing for a cont~~~~ raisin e problem, with no realistic; opportunity for solution. The basic researcher often doesn’t see the problem as a matter of real to him and therefore doesn’t address it. Thus, we are in the situ n where all accept the problem as important but, like the proverbial question of the weather, no one does anything about it. This aud the followin paper aim at proposals to do something about criterion problem. They will outline a program of 1 to the small group field and, between the harried applied terion issue is obviously more com@ic ed than that posed above, with many subtle issues cloaked in the neric question, as follows: * Prom Bureau of Medicine and &~rgcry,Navy Department,Research Task MROO5,12-2005.01. The opinions and statementscontainedherein are the private ones of the writer and are not to be construedas official or reflectingthe views of the Navy Departxnentor the Naval Service at large. The author is grateful

for ths criticsSma of earlier versions of the manuscript by Drs. William Haythorn, Kenneth MacCorqwodale,Joseph McGrath and Thorton Rohy. a01

102

IRWIN

ALTMAN

I. HOW and what is the group doing with respect to its behuvioq?

Obviously, a group functions by behaving and performing. The phrase “with respect to its ehavior” refers to the t:;hole dependent variable domain and is the heart of the criterion question. To know how a group is doing, one must know what aspects of behavior to record, watch, code, count and compare. This is a very wei decision, the import of which is not always appreciated. Unfortunately, smal; group researchers seem to spend considerable energy selectin independent variables to manipulate or predict from and too little energy examining ad choosing dependent variables, in spite laftheir raisoa d’&re being the understanding of dependent variables or behavior. It, almost seems as if we are building a complicated theoretical fabric, cloaking it in ternIs of sensitive laboratory contrived manipulations and then relying on a questiormaire response, single sociometric choice, or a tally of who-talks-to-who to evaluate our theories and hypotheses. Following are directions that seem worthy of pursuit with respect to what behaviors and performances need to be studied: (a) We need to study more than performance aspects of group functioning. It is ordinarily assumed that the criterion problem implies analysis of task-relevant behaviors alone and not social-emotionalinterpersonal behaviors. Thus, we hastily examine only performance errors, time to criterion, number of correct choices, number of attempted solutions, etc. This is too restrictive and only part of the criterion domain. In fact, task behaviors may be only the last point in a long line of behaviors culminating in errors and correct responses. If one agrees that interpersonal activities are important determinants of task pe::formance, then why not study them als not alone but Zinked to tiisk behaviors? (b) In addition to broadening the scope of study beyond the performance domain, stronger emphasis should be placed on studying the task performance process itself, i.e., the sequential chains of taskrelevant behaviors that precede final performance output. ese contributory behaviors surely cumulate and combine in various fashions to set limits oa Gmil performance and warrant further attention as intrinsic parts of the criterion question. (c) As implied above, one should not merely catalogue the variety of behaviors contributory to total effectiveness. Rather, their relationships to each other at a point in time and through time slrould also be

AINS IN SMALL

CROUP

RESEARCH

103

of the criterion problem is not just understanding otionai behaviors separately, but the causal and tween behavior classes. r addresses this general question. Subsequent

ups do not just “do”; they function in situations e rat in the maze, the hunran on a perceptual on a case study discussion, a communication nerie criterion question must add the qualifihat job, task or goal? I’+w will disagree with sical findings in terms of task and situations. cient merely to draw conclusi;Isns about the effects up cohesion on problem solving, for examiple, for the ball and occo case discussion task, etc., one at spiral task, for the Johnny a time, Reversion to the pr ctive womb of extreme operationalism issible, but not always fruitful. At least it becomes less so as is y of knowledge increases at an increasing rate, as is now true a in the small group field. What is needed t et leverage on this aspect of the criterion question is an unciersta g of task parameters so that tasks can be precisely compared and similarities and differences made apparent. Then, when the relationship bletween cohesion and rformance is studied, for example, we can ask about past research m terms of a total situation, not just the variables themselves. Then one could make a more systematic selection of situations to observe ables, then one could more readily synthesize knowledge and beyond a sin e empirical study. In short, without ati underroups are required to do in systematic terms we sssibk understand how the group is doing. 3. How

is the

up doing over time?

The issue of temporal processes is also a most critical feature of the criterion problem. Group activities do not occur all at once or at a single point in ey transpire &roug& time. Qne of the major problems in retime: search on small groups has been a general neglect of temporal factors. Without an appreciation and inclusion of time as a variable, the

104

IRWIN

ALTMAN

question of how a group is doing will receive only an incomplete answer. It is interesting that students of animal and human learning consistently employ time as a variable. Thley study learning through time, manipulate variables temporally in the form of d~rivat~~~ schedules and reinforcement schedules, Empirical research on therapeutic processes has also long been concerned with ch ;-&rat behavior through time. Yet, with few exceptions social psychologists interested in attitude change and with group functionin studied dependent variables and manipulated independent variables at onI!: a single point in time. It is our thesis that circumscribing of the criterion question mandates a greater appreciation of temporal aspects of group functioning. This aspect cf the cri.terion problem hss already been reviewed to some extent by TUCKMAN (1965) and MCCRAI-H and ALTMAN (1966). 11.

THE NATURE

OF GROUP BEHAVIOR

TO BE STUDIED

As indicated in the first question cited previously, G ~urral aspect of the criterion problem in small group research, and for that matter of any research concerned with human behavior is: “What behavior should be studied?” Our thesis is that small group research has not g&n suflicient attention to this problem. Past neglects in this area have several facets which will be discussed next. 1. IFhe jdlacy of performance measure as the only criterion indicators

The criterion question is most visible where research is done to help design, evaluate or improve a “real” system or group, e.g., an Army squad, a Navy crew, an Air Force bomber crew, a missile maintenance team, an industrial work group, etc. The goal addressed to the researcher by his patron can take any of several for one that will do this job; improve the one 6 have; tell me what’s wrong with this one.” The researcher’s eye is always on the sparrow of creating a system that will perform well, Le., generate a system end-product or sub-product. The researcher usually proceeds by only examining task performance since his mission is to predict or to maximize performance. The measures of performance selected are those that appear closest to ultimate performance, although at some later time, he hopes to validate intermediate criteria against actual job performance.

DBMAI[NS

IN SMALL

GROUP

RESEARCH

105

SU~I an ~ppr~~~h seems overly narrow and restrictive, Consider the n analysis of the mission will show rtain bombardier actions, but that ill surely affect the occursuch actions. The p!ae must be navigated ledge about environmental and , we might say that the ression, fear, conflicts, or it is true that such interier performance, doesn’t it of total system effectiveness also les? Thus, if one only relates crew not very much information is obtained, except by about why method A is better than method indirect inference or rtial prior hypathesis. One can thereby discover vered about the import

rformance should include more than measures of final system output,

all conceivable behavioral measures, but that the design of a study ious judgments about inclusion or exclusion of intervior measures that might be important. It is not the mere at is so serious, but the apparently unpremeditated exclusion and inclusion of measures without really addressing the ecision question. In some respects, historical precedent may have fostered this narrowness of viewpoint. The otions of Giemeinschaft and Gesellschaft roupsl JENNINGS(1950) distinction between psyche-groups and sociogroups etc., may have implied i ependent types of groups, although

formulations actually refer to different functions within groups. While most small group research focuses on one or the other class of behavior, with only infrequent attempts to relalte them to each other, there are some notable exceptions. BALES (1950); CARTERet al. (195 1 a, b, c); ST~X and THELEN(1958); THELEN(1959) and others study perfomance and interpersonal-social behaviors simultaneously. How can this lofty and comprehensive aim be achieved? It can in these;

106

IRWIN ALTMAN

theory, but not always in practice. Theoretically, it requires a concept of interlocked behavior dimensions in which all forms of behavior are identified. It cannot always be achieved in practice because of administrative reasons. An ideal system will ena.ble researchers to c make choices about what behavior to study. It could serve as an interrogatisn and synthesis device such that the
It must be admitted that the suggested approach will not solve the “ultimate” criterion issue; rather it emphrasizesa broader conception of the intermediate criterion domain and may provide a language of behavior which will allow the systematic linking together of different behaviors, both within and between research studies. Such an approach should reduce the bugaboo of the ultimate criterion problem because it will enable identification of more sources of variance within and between systems. Then, as ultimate criterion measures are obtained, even in crude form, it will be possible to study the link between intermediate and ultimate criterion classes, as well as relationships within the irtermediate criterion domain. 2. The fallacy of too few dependmt variable linkages Research on small groups has not only been too restrictive in the kinds of depeudent measures employed, but also in studying the relationship between such measures. Xnrecent work to inte group research information (ALTMAW aud MCGRATH, 1959 ALTMAN and MCGR’ATH,1960; ALTMA~N,FENDLETCIN a 1960; ALTMAN and TER~UDS, 1960; MCGRATH, 1962) some interesting actuarial statistics emerged. In a random sample of 250 small group studies, the median numlber of variables (defined in term!; of different operatioual definitions) emploiyed in the average study was four. One third of the studies contained three or fewer different variables and three fourths had six or fewer variables. These figures include both independent aad dependent variables. The feeling one gets from skimming such a sample is the propensity for complex factorial designs which use as many manipulated independent variables as possible at the expense of behavioral measures. Given administrative, time, and

AlNS IN SMALL GROUP RESEARCH

107

money restrictions,

s all group students have apparently selected trade-offs in the difcc ore manipulated independent variables e criticality of the issue becomes

ent that the criterion domain haviors linked to each other in a chronological off by independent variable nctioning. In fhis sense, omain are both independent Hews that the chances for havior are considerably d behaviors. Doesn’t our of the deference between the performance Lffectiveness for example, become more ng processes between size n’t we have to see if the s, more opportunity for tervene between the size condition and resultant performance? In short, shouldn’t we expend more ener understanding linkages between behaviors, along with assessing the gross i pact of antecedent co ditions? The ij.nswer is not “either-or”, but a balanced egy within resources and knowledge. out major sources of antecedent, nonhas been to sea ect behavior, e.g., group composition, onditions which ammunication patterns, size, etc. Many of these factors ugh results are not always unambiguous. Perhaps e timo to search out more intensively why theSe are imporof the chains of behaviors they set in motion and the ways in which those behaviors in turn dampen or elevate the probability of occurrence of subsequent classes of behavior. With more of such proccsscs, we can then return to the strategy of kIl0Wl manipulation of large numbers of independent variable conditions. As any position must be, the preceding criticisms are somewhat overdrawn. There are many in the small group field who have partly &opted the strategy suggested above. These derive mainly from the behavior observation systems of BALES (1950);CARTER et al. (1951 a, b, c); STOCK and THELEN (1958); ThELEN (1959) and others, and the search for group dimensions typified in the work of CARTER(195%

108

IR1NINALTMAN

HEMPBILL and COONS(1956); HEMPHILL and WESTLE (1950); CAT-USLL (1948); CATTELLand WISPE(1948); CATTELL(1953) and others. These approaches, although somewhat different ia philosophy and technique, have aimed at identifying the dependent aspects of group functioning which must be a?counted for in the conduct of small group research. 3. Behavior ubservation systems

Many behavior observation systems have been developed for the study of small group behavior, children’s social interactions and psychotherapcutic processes (HEYNSand LIPPITT, 1954). These systems vary in several technique aspects such as frame of reference of the observer, size of the recording unit, degree of observer inference, etc. More important for the present discussion are differences in types of behavior investigated and application to analysis of small group processes. One way in which systems vary is in tlheir differential emphasis on form versus substance (ofbehavior. At one extreme is Chapple’s Interaction Chronograph technique (1940, 1942), which focuses on temporal feature s such as length of interactions, periods of silence, pauses, etc. At the other end of the continuum are systems offered by CA~CTER et aI., (895 1 a, b, c) and students of children’s social behavior (ARRINGT~N,943). These contain specific:behaviors such as deference, negativism, calls for attention, asks for information, gives bald commands, etc. It is interesting to note that systems are differentially capable dsf reproducing what actually occurred during a group interaction, The Chapple system can only recreate temporal aspects of the interaction, unless further in*:c:ences are made from derived indices. F~URIB~S, et al., (1950) and STELNZ~R(1949) can only reproduce delimited aspects of interpersonal processes, BALES(1950) and CARTERet al., (1951 a, b, c) can recreate certain aspect of both task-relevant activities and interpersonal-social activities. In a sense, the Bales and Carter sysLems are most comprehensive because they are most able to directly reproduce more interaction, But even they fall considerably short of the mark, for without reversion to a sound tape or film, they cannlot answer the question-gives opinion of what, or proposes solution of what type, at what time, in

I. DOMAfNS

da

~r~~~~~~~ t~~Qr~ wht

fN WALL

GROUP

RESEARCH

109

mpect of the problem? This should not t as a way of seeing

ot focused on, thereby suggesting what the builders

serve in observing group functioning. P, it would seem useful for the aware of what parts of from reality, and what ored. Second, it seems ive as possible, selecting t as many aspects of toward becoming the all possible data that helps reproduce havior classification systems is in level ales” category shows antagonism covms atea other status, defends or asserts self. Once coded e to sort out how much each type of act contributed to stem (1951), there are ates, punishes, rejects (and more specific subcategories within each) which seem to be more c delineations of the Bales category slrows antagonism. This ales systems, with the feature also distinguishes the Carter an any more possible categ former having . Of course, specificity is neither neces rily desirable or undesirable; it all depends on what one is interested in studying. But a knotty question arises for the researcher who must decide what behavior to study, and at what level of speeificityagenerality. To help in this judgment, what is ideally needed is a ~~ass~~~ationsystem or language of interpersonal behavior which is relatively ~~~~reh~nsive and which also builds in alternative uestion, not previously noted, is the extent to ries are linked to one another in some conple, how do the behavilDr categories submit, ceptual fashion. For condemn self, and support in the FREEDMAN et al. system (195 1) differ from one allother and how are they alike? All that can be said is that they are i&rent, In many respects, the degree of distinction possible is analogous to differentiating people by proper name. The use of first names to separate Charlie from Joe is useful, but either incomplete and superficial, or a shorthand summary of a set oi proper-

110

IRWIN

ALTMAN

ties, e,g., sex, height, weight, eye color, personality traits. The latter dimensions allow for a more systematic description than nominal lalyels. Their use also provides a common kmguage of properties which enables specification of each person’s location in a multi-dimensional space and the determination of the “distance” between people. The ,majority of behavior observation Bystems are equivalent to the use of proper names which do not in and of themselves specify underlyin properties in any mtegrated sense. Bales’ system comes cl however, with behaviors tagged as being in task areas, positive or negative domains, etc., and e value on each such dimension. In short, what is sought is a framework of behavior which will specify behavioral events in a common language, and thus allow for precise description of how one behavior differs from anoaer, regardless of the setting in which the behavior occurs. That is, we seek a general language of behavior or an abstraction vocabulary oE behavior which parzllels, “length”, “weight” and other properties used to describe physical objects. It is not that we hope to be able to reduce all behavior to a common level of analysis, but only to provide a broad set of properties for conscious selection of a desired level of analysis. and to provide a broad perspective and basis for translating between behaviors. Let us expand t’urther on rhe reasons For seeking such a framework. First, it will enable researchers to see where they are in a multiIdimensional behavior space lnvhen they select variables and design a (study. If each possible form cjf behavior is placed in SW%a space, one should then be able to know immediately what sources of literature are relevant, i.e., those involving other behaviors “close” in the space. Or, if one is interested in relationships “‘CfistanP in the space (the variables differ considerably in their properties) one would then be signalied to understand intervening variables which may be mediators, relationship attenuator, etc. Moreover, if “distance” in multidimensional spacle between behaviors is valid, one could predict occurrence and intensity of relationships before the fact, i.e., probability of direct relationship is inversely related to “distanW. Some work along these lines by MCGRATHand ALTMAN (1959,1966);kkG~a~~(1962) indicate that this “distance” concept is useful for predictive purport. Equally irngartant, the location of a variable in such a space and analysis of existng knowledge within and around the space of interest could easily lead to id~entifrcation of gaps in knowledge, and ideas for additional or

~~~AV~~RA~ D0MArNSIN SMALLGROUPRESEARCN

111

research. A system which possesses the features fore serve as a powerful research vehicle., tial for applicability to any a case study discussio art&k! task. All could be guage. This possibility does example, TALLAND t i955) and ates that the Bales’ system is owps and community discussion groups to some problem. This is les’ system, since it was not meant for use in is felt that capability for wide application is a goal se all parts of such a system, ratdy select and reject portions of the total space. e various behavior observation systems currently available differ from each other in several respects. Tlleir reproducibility of events oc rring in group activity is diverse, ranging from limited to consider rtcreaticbn of interactions. With rqzt to reproducibility, some focus opdy on interpersonal activity, others on performance activities, and some on both. They also dil’fer in the extent servation categories are specific ok general, and their ability to link behaviors in a systematic fashion. Our thesis was that value would accrue if a behavior system or behavior language was developed that encompassed II broader range of behaviors than heretofore, reproduced more group activify, and allowed for the systematic linkage of various behaviors. 4. R

h on the dhemions sf groups

There have been several attempts in the past fifteen years to identify critical dimensions and properties of behavior in small groups. These 81%summarized in TMBAUTand KELLEY(1959); HARE(1962); VERBA (1961); O~LEMNEWSKI(1962); and S’JTOGDILL (1959). Typical approaches to the identification of group structural and behavioral dimensions appear in the factor analytic work of CATTELL(1945, 1953); CATTELLand WISPE (1948); BORGATTA and COTTRELL(1955); BORGATTA (1956), HEMPHILL(1956); HEMPFIILL and COONS(1956); HEMPHILLand WESTXE(1950); CARTER(3954), and in the more analytic work of DEUTSCH(1949); JENNINGS(1950); HOMANS(1950); and HARE (1962).

112

IRWII ALTMAN

In a sunmary of factor analyses conducted in five other studies, CARTER(1954) concludes that three factors can be employed to identify the behavior of indiGdua1 groups: (1) Indi:vi&d prmnirzmace and achievement. “. . . behaviors of the individual related to his efforts to stand out from others and individually achieve v:arious personal goals.” (2) Aiding attainment by the group. “. . . behaviors of the individual related to his efforts to assist the group in achieving goals toward which the group is oriented.” (3) Sociability. “. . . behaviors of the individual related to efforts to establish and maintain cordial and socially satisfying relations with other group membersI.” These factors were induced from t!he work of Couch and GARTER (1952); SAK~DA (1952);HEMPHILL and CROONS (195C); WHERRY (1950); and CLARK (1953), all of whom identified three general factors, Examination of the behaviors under each factor reveals some interesting features. It appears that individual prominence includes botIr task relevant activities, e.g., le adership, initiative, energy, physical ability,, and interpersonal interaction, e.g., aggressiveness, authoritarianism, timidity, and cheerfulness. The latter set can also be divided into those describing the person as an entity, and those describing forms of interpersonal behavior. The second factor a!lso has task and interpersonal behaviors, although the weighting is in the direction of task behavior. The third factor, sociability, only contains interpcisonal activities such as geniality, adaptibility. etc. In summarizing, several factor analytic studies of group behavior. BORGATTA, et al. (1956), induced five general factors which show some similarity to those identified by Carter. The five f:rctors, as labeled by STQGDILL(1959)are: (1) differentiated role structure (2) informal interaction (31) goal direction (4) group integ,ration (5;) within group interaction facilitatio Taking a more analytic approach, ~ENNINCS (19!Io); HaMANs (19%')): DEUT~CH (1949), and others hitve proposed similar, although grosser, dimensions to distinguish between groups and interaction processes witKin groups. Jennings distinguishes between sociogroups in which relations between members exist with respect to working together on

A’INS tN SMALL

GROUP

RESEAR’CH

113

some: task or problem, and psychegroups in which the major end of ssosiation. This is amount of each with the results of omans and Beutsch. directed toward those aimed at Homans refers

(b) indirectly related to task activities (e.g., sociability, friendship) (3) Individual roles, e.g., leaders ip, while may be complex derivatives of behaviors in the preceding areas, e mare features such as those listed above which can be incorpohavior system, the greater the probability that important up interaction will have been dealt with. In the next ased behavior classification or language IS discussed, which partially derives fronr the various behavioral observation systems and from studies identifying group dimensions. 111.

A

PROPOSED

BEHAVIOR

1. Gerwra! wumptions

CLASSIFICATION

SYSTEM

and characteristics

This section outlines a behavior language which aims at overcoming some of the limitations of small group research discussed earlier. As an interaction classification device the system is, at present, prelimi-

114

TRW IN ALTWPN

nary; is; has only been applied informally and will require further development. Therefore, the reader should focus marily on the general approach and secondarily on specific details. fore describing the system it may be useful to set forth some general features of the langua,ge. (1) The basic terms of the system derive from Newcomb’s conceptualization (195 3) that interpersonal interactions relaticnship between people (TIIELEN, 1959; ALTMANand 1959) and with respect to some event or object (X). ‘Fro Q.C~O~V in group activities and the form of their relaticnship to each other and to an object art the major dimensions of the proposed language. (2) The system analyzes interactions in a formal, almost grammatical sense. Each social interaction is seen as haviag a set of unique values on dimensions involving the parties to the interaction, t of their interaction, etc. Thus, each social act is not placed in a single category but in a set of categories indicating “how much” of each of several properties it possesses. To put it another way, each unit of interactive behavior is located in a multi-dimensional space, the bounds of which are defined by the dimensions and theii categories. (3) The behavior language allows for description atld analysis of any of several levels along a molar-molecular fXMhlUUXTi. Because it is multi-dimensional, certain dimensions can be eliminated without loss of sensitivity. (4) The classification scheme assumes that task relevant and interpersonal interactions are not from totally different universes of discourse but that they diRer only with respect to certain dimensional properties. Thus, interpersonal “hostility” and “task disagreement’” are similar in some respects and different in. other respects. However, most emphasis in the present discussion will be on the task-relevant aspects of interaction. (5) The language was designed for application to any form of social behavior. This required some sacrificing of the ability to completely reproduce an interaction and led to a focus on more generic properties of interaction. Thus, the language can reproduce an interaction such as ‘*John asked Mary for more information she possessed about the problem”, but cannot specify the exact substance of the information in question. (6) The system is meant to be much more than an interaction

ALL GROUP

RESEARCH

115

of as a language of interused to guide research, integrate alytic structure. Because it is a idic situation would, of course, of that situation. In guage as applied to ~~t~ra~tion~In a later paper, we shall suggest scribing small group tasks. f the ~lass~~~atio~system itse

The fuudam~ntal dimensio~~sof the behavior Iznguage are the i6thtca of the interaction 8r a.-~, the form of interaction, the focus or sb&? of the interaction ar the immediate recipient or r@ferezt. or ease of exposition, the discussion applies only to dyadic groups. situation, the actor or initiator can be (a) ACTOR. In a dya Ep4pr80fl 1 (PI ), person 2 as 2 whole. The latter OCCUSISwhen the dyad 01%~ a unified posi of some type or interacts with the environment as a group. In ps larger than a dyad, the number of possible actor,6 increases rapidly, since the possibility existsfor various subgroups tc$ (b) REFERENT, This dimersion identifies the person or thing the actor is inte~actiug with, or the immediate recipient of his interaction, tor) speaks to Many (referent). Categories which can obtain I), person 2 (PS), a prece of equipment (E) (if the ;roup task involves any). In larger groups, the referent can be any sized subgroup, To make the category more inclusive, intra- and extragroup referents should be employed, e.g., the actor may orient to 2 person who belon s to another group, to the experimenter, to subgroups of another Q another group as an entity. This dimension specifies the form of orientation between actor and referent with respect to an object. A single interaction can assume any one of the following forms: (1) Asks er s~ardws. A re ationship is interrogative when the actor asks the referent for something-information, an opinion,help, guid(c) INTERACTION

FORM,

116

IRWIN

ALTMAN

ante, etc. Such seeking behavio,r may obtain between any co oE actors and rcferents. A distinction will also be made between an original search and a repeated sea.rch (one previou (2) Informs or gives. This category occurs when tae referent with information which ordinarily is (see category 4 below), i.e., has no evaluative ~(~nnot~tion, examlple, “the book is blue’” is an informative or descriptive statement as compared to “the book is too blue”. Two subcategories of this tckmension seemed useful as the system was developed. I,@ Infoams of new informatiim, (I) Thk case occu when the actur provides the recipient with information that h not previously been exchanged. l(b) Repeats old information. (C’) In group interaction, there is often a repetition of information previously presented. Since a distinction between reitera,tion and introduction of new information may have import for understanding group developmental processes, this category was included. (3) Infers, presents intermediate conclusion. (I”) This category occurs when au actor gives more than a piece of inftirmarion but makes an inference of some type. The inference may take several forms including the statement of relationship between previously uninte rated bits of information, a suggested intermediate conclusion, a line of reasoning, e1.c. It involves the presentation of opinions and other forms of expression which go beyond raw descriptive facts. In many res category blends into the next major category. (4) Evahates or judges. (Ez?z) When Jn interaction involves a Ytatement of desirability, goodness ar poorness, correctness or iucorrectness,agreement or disagreement, etc., it is evaluative. definition is the notion of some standard or yardstick dnst which a thing, person, course of action, etc., is compared. This not only includes such obvious evaluations as above but covers cases such as a sonar operator reporting a sounding as bein “too close for comfort”. One of two possible values can oc r for an evaluation: positive ative evai~t~~~ evahation, (E+) as in agree, good, correct, etc., and (E--) ZISin disagreement, etc. The subcategories repeats evaluation (E-i-)‘, (IL-)’ can also occur when an evaluation ‘3 repeated. (!S) Tells or directs (!) This category applies only to cases where the actor is one person and the referent another and involves situations where the actor directs, orders, or demands modification in the

IN SMALL

GROUP

RESEARCH

117

irr sentence form S before, a given directing act ies, this involves en4 are not distinhitting, etc. lt is possible that this “Qdescribe both te~~~~i~nsor that a new dimension be intro-

carries thr~~u~h the remainder of the le, re~er~~~~sand s9xx3 r:tnge over a continuum of ne individual is the smallest comnt of a irrupt the su vidual embedded in one or mobs s-tits s, :~nd finally there is the group itself. The in an ~~~v~ronmentor surround. n or cumulation

ian from #SW& to a&v? is much like a Guttman scale concept. feature may have implic ations for group developmental processes,

is discussed later. THB ~~.IBCT OF

This dimension refers to the ~~~~~of the interactisn or thz thin of interest to the actor. The object is wht the interaction is at.xxlt; the referent is the vehicle through which the inteaaLon is mediated. ’ IYheobject may assume any one of several! values: I. 2Wti~ it member (P 1, P2) In a dyad the object can be either person regardless of who tbs.: actor is. For example, person 1 can ask person 2 about himself or about person 2. 2. The group (G) Either actor can ask the other person about the (d)

group as a whole, etc.

THE INTERACTION.

118

IRWIN

ALTMAN

3. The task (T) or some task feature, e.g., equipment, 4. The arzvironmnt (E) a. Nonhuman aspects of the context within which the group is embedded, e.g., weather, salinity of water, o animate features. b. Individuals who are not in the group, but w other groups. CL Cut,-ride groups as entities, cl. E..xp’I;rimenteror other specific individuals. F’or largzr groups this dimension can be expanded into s combinations of various types. The four first order dimensions are sunmarized diagramatical~~ in thi-!uplper poitions of fig. 1. Theoretically p sdl combinations of values on the dimerrsions are possible. There will of course be meanin combinations, but it does seem possible to -imagine instances of most behaviors, even where person 1 (actor) negatively evaluates (interaction form) himself (object) to himself (referent), e.g., he thinks out loud albout the error he made. As far as can be seen, any fo,rm of task-orienteld interactijon occurring in a dyad can be described in this framework. 3. Secund crder dimemions

There are two additional dimensions which provide a more detailed descriptiola of the interaction, object. These specify variations in properities of objects-their static-dynamic nature and their temporal locus. Thee introduction of these dimensions expands the system as shown in the middle section of fig. 1. (e)

C%IEC’T PROPERTY. STATE/ACTION.

Two general classes of object propertier; will be employerd: state (s) ard action (a). A stute describes ant: aspect of an object as an entity and provides a summary at a specified point in time, e.g., biographical, attitudinal, personality properties of individuals and groups. An action represants a dynamic property of an object. Actions de#t an event that occurs during a temporal interval, e.g., grsup or member performances, communications, inl:eractions. rence between state and action is analogous to the biologist’s distinction between the nerve (st ) and the nerve impulse (action), organ structure [state) and organ function {action), or tire

s

I

119

QUP RESEARCH

Fig. 1

Person Subgroup

extra group) extra group)

State/Action Temporal

is/a)

ILOCUS (0

State Action

Past Present Future

interaction form

dimensions

Person Subgroup Croup Equipment (intra and extra group)

Categori

Ask (2) Inform (I) Infer (I”) Evaluate (E) Tell/order (I) Act (a)

Sumnary Formulation

II

PII”E

R( ‘&,& t -------

-& !a -----3

0”)

1

marian’s distinction between the verb “to be” (state) and the verb “to do” (action). This distinction has been previously used in describing variables of the small group field (MCGRATH aud ALTMAN, 1966). States can be divided into several subcategories: a. Bio~raphical/demographic. Age, education, socioeconomic status, etc.

IRWIN ,iLTMAN

1120 b.

Cognitive. Intellectual,

belief,

general

values informational,

properties of objects. c.

Physictzl/mobor/operative. Physical or operating condition of objects such as motor

skill levels, equipment operati

dition, etc. d,

rties such as

r”MotivLrt~onrtl/personaEity. Object pr emotions, personality

characteristics, etc.

As with the states, actions can be subdivided vational/personality

and physical/motor/operative,

into cognitive,

moti-

but not

Cal. (r”pOBJECT PROPERTY: TEMPORAL LOCUS. or any state or action the actor can focus on a past9 present or Zuture property (1) Bust

of an object.

properties (pa).. This category specifies an object property

manifested in the past. While

the definition

of what is past, present

and future may vary depending on one’s purposes in analyzing interactions, the category past earlier interaction,

will be used when the object appeared in an

except that an immediately

preceding

interaction

would not be considered past. (2)

Presem properties @a). States or actions would

r”.spresent if they appealed in the act immediately

preceding

interaction. (3) Future

properties (fu).

Any

reference

Iproperty, even if manifes?.ed in the immediately

to

a future

following

object

unit would

be in this category. 4.

Thirci!order dimemiorw At a more rehned level, it is possible to further delineate

tioclgh

specifkation of dual-objects and the interaction

this time

to

describe

inter-object

relationships

rather

referent-object

relationships.

total behavior

system is shown in the last row of fi

(g) DUAL-OBJECT. This combination

The location

property

closely to the state-action dktinction.

than

actor-

of these dimensions in the

can theoretically

of values on the preceding

form dirn~~i~~,

dimensions,

occur

for

any

but ties most

The meaning of this dimension is

clearest in the case of actions, since it does not sec:m easily applicable to all states, When the object property

is an action it is possible to

may best be visualized as an object-object

matrix in which the actor

specify the object or focus of that action, i.e., the object’s object. This

AINS IN

A&L GROUP RESEARCH

121

object (rows of the matrix) with cification of the other r example, ‘Joe (actor) Sam’s (object) earliu e of eqc4aipment (dual

mber, group, etc.) can ot~v~~ion~lstates but may not always raphical or physical states, because the object as an entity with freenable objects. Actual ;~pplication al object is appropriORM. As shown in fig. 1, the nature

tween the object and the dual object can be xactly the same terms as the actor-referent relationship * For example, e asks ent of a piece equipam’s earlier imt3arecf a t interaction form dimenect did erroneously to the equipment as ension ,s most applicable to actions, but e seen that introduction of this dimension serves to further d4incate and reproduce the nature of interaction between group members. 5.

~mparimmwith other systems at this point is the way in which the proposed e: diEers from various social interaction classification ales, arter etc.).

G proposed system see nt of deswiption. bviously, distinction for ist~~~t~~~~sthan any now in use. ) but it is felt that the c is not necessarily a vi e extent to which this is additional di~er~nti~tions are important. true can, of course, be determined only from application of the system aflclcomparison of independent variable effects on behavior. An idea of the number of distinctions introduced by the system is suggested in fig. 2, which presents a few examples of dyadic interaction classes having to do with information seeking using first and second order

122

IRWIN ALTMAN

Fig, 2 EZxamplesof search behavior interactions using first a (ofthe classitkation system %I___ ‘ I Actor* hteractiopt Form Referen Person 1

ask, searches e .

Temporal Locus

Future

Person

State

What was your condition at T(x-l)?

Past

Present

.

I

second order dimensions

06 ect 2

about.

. .

lYma2

Action

What did you do at T(x-I)“!”

What is your condition now, TX? What will yam”condition ‘beat ‘I’(& I)?

What wiil you be doing at 7 I+- I)?

* These examples apply olsly to the category values indicated. Obviously, all other combination values on the dimensions could occur.

dimensions. #From this simple example it can be seen that tile total

number of possible interaction types is large, with each capable of being compared with all. others since they are described in the same terms. (b) Cmnpreiitr~ivenem.

The present liuqqtage seems to allow

coverage of as many, or more behaviors than existing interaction recording systems. For example, the range of objects used allows for the observatior of intergroup interactions; the employment of unit designations such as member, group, subgroup do not appear in other systems; the system builds in actor-refererat-objects sets as m integral part of the system,,whereas most other systems ignore these dimensions. Unlike some other systems howe:ver, it is narrower in that it ignores intervening processes such as internal feelings, attitudes, etc., and focuses only on overt acts. cc) Appl’icutim, The language was designed for application to a variety of group settings ranging from free discussion tasks to various man-group-machine task settings. This does not tend to be true of existing systems, e.g., the Bales and Carter methods were designed primarily for problem solving settings of a discussion type.

AINS IN SMALL

ROWP RESEARCH

123

oleculm continuum. As suggested , sel:ond, and third order dimensions, the on in detail of description. One can ignore all of them. is tends to be true of few d where feasible does not depend on cation of superordinate categories such ial emotions areas, problems of communication, cona built-in logical flexibility of the posed language defines all a common set of properties. aeh behavior in an n-dimensional space t occur on each dimension, not in one over-alt category. As discussed earlier, this has some d&nite advantages s of precisely identifying behavioral differences in more than notin terms, allowing more rapid integration of existing knowIedge, a interpretation, and having potential e virtues of an interlocked behavior e~ry, are discussed in the next section. 6.

hat

cm be clone with the general behavior language?

It is necessary at this point to deal with a provocative question, namely, “so what?” The system has been described, some of its features contrasted with those of various other interaction classSication d now the. reader is re y to adopt. it to his own needs. But what shall h do with it or what can he do with it? Following is a enexal discussion of some possible applications. (a) Andyds of i”ndr’vidud roles and gfiorcp strerctrrral dynamics. CCEW(:the system “tags” individual group members as actors, referents and objects, it is possible to describe the nature and drstributian of various roles and functions they assume. By pairing people and interaction form dimensions (asking, evaluat ng, informing, etc.1 it may be possible to develop more speci& empirical indices which escribe people’s roles as structurers, criticizers, information providers, etc, It may also be possible to identify and describe the nature of various cliques and coalitions depending on various combinative

124

IRWIN

AtTMAN

values on certain of the dimensions of the system. EQually interestmg, an analysis of changes in roles and structure can be observed as groups progress through a task, which could shed light on the consistency of individual and group behavior. (b) Interrela~L2n&/ps of behaviors. In current research on small groups, little is known about the interrelationship of various fortqls of behaviors. Are there behavior clusters such that the occurrence of one behavior in a cluster increases the probability of another from the same cluster being elicited? Is there a sequential ordering of behaviors within and between clusters? (e.g., do interpersonal affective interactions contribute to performance by tripping off information excha and facilitating introduction of new information, then allowi ences to be quickly made and tried out because trust exists , Tsmrugir zsmii~s of Meractions in terms of the proposed behavior language it may be possible to identify such behavioral linkages, not only in the chronological sense but in terms of concomitant associations. In fact, the language already implies the nature of such linkages in the inclusion property of the: interaction form and re:ferent dimensions as discussed earlier. Thus, it would that information interaction units are more apt to yield inherence units first, evaluation units to a lesser extent and action units to the least extent because they are more distant conceptually in the spank encompassed by the system. Similarly, persons as actors are mart, apt to be associated with other persons, then subgroups, then the group as actors. While it would not be possible to provide a final test of this hypothesis because of the role of situational preventing the “pure” case from ever appearing, this ty could be a useful vehicle for distinguishing between varying conditions. Certain conditions of group structure, co etc., might effectively bypass some of these hypothesized chains, others might not and to varying degrees, (c) Developnertial presses.. One major use of the system could be to describe a group as it progresses toward a goal---not in terms of chronological everIts but in terms of functional activity stages. There has been some; work on this problem by BALES and his coworkers (1951, 1953) but it is fairly sketchy and only describes broad developmental stages SUCI as orientation, control an tension release. Someothers have also begun to address this problem (SLXUTZ, 1959;

ALL GROUP RESEARCH

125

is is felt to be a most to which the present r a general discussion

roceeds through such a task, as

e total group interaction functional stage culmiinto a series of substages which e earliest such stage would f the inference were positively eva!uated, evaluation of that inference.

ithin each evaluat

is would be followed by

up would act, i.e., attempt a solution. stage there are substages defined in ence of information exrmation, eventually culminating in

dual transition from individual members as bodes will not necessarily manifest these stages overtly, there

soup structure, problem difficulty, etc. But it is such increase the potential value of this approach. For why is a cohesive group more effective than a non-cohesive t may be that the cohesive group, because of underlying rust, is more readily able to accept member inferences without extensive analysis of reasoning an thus shortens the time of such activities. The smaller group may be more effective than the larger

126

IRWIN

ALTMAN

;:roup because the available information is more concentrated, more OEit gets into the publife domain faster, inference logic may be more : meadilyspecified, etc.

(d) Asse#in,~ independmt

variable eflects. The precedin cussion has ber:n primarily on possible dependent as phenomena, wiih the proposed behavior language viewed for sharpening up the dependent variable or criterion domain. 1f this is done, the language obviously could then be used to more syste ally assess the impact of various antecedent conditions. These cover the range of interests now involved in research on smal from group structural factors, various input conditions, channels of communication, task differences, etc. This cou!d be done with respect to the previous dependent domains of member roles, developmental processes, group level effects, etc. This concludes the discussion of the proposed behavior classification language. It should be pointed out that the scheme only spells out dimensions and categories of the criterion dcmain. Et does not prejudg;: what behavior is good, correct or desirable from the point of view of a system designer or experimenter. It describes the group as a functioning entity without imposition of an outside judgment. In many respects such standards are not part of the group process but derive from the needs of the designer or researcher. This is not pointed out because it is an advantage or disadvantage of the proposed but merely to clarify the issue. The proposed behavior sys be used for a variety of designer needs, e.g., if his k&rests are to minimize inferences in a given task setting, then research could evaluate departure from this standard by differently composed But the notion of absolute standards of goodness, etc., are not automatically built into the behavior classification system anei we feel rightly so. At the present juncture, some major gaps remain to be filled in using the underlyinp behavior language as an interaction process analysis technique. In addition to further extensions to handle socialinterpersonal interactions, the system must be applied to determine reliability. It also needs to be compared more rigorously with existing

:systems to more precisely identify where as much ground is covered and where it is not, which distinctions in existing systems are lost and which are made more sensitive, etc. Moreover, more conceptualization

ALI GROUP

RESEARCH

127

es, group developmental is is possible will enable the data framework to a theory of age only suggests w various terms of one another in terms of more about the relative importance of r dimensions. Some will ior than others and some ore v~ri~~~~ thaw others. This needs to be explored to the dimensions and to the categories

~~~~~~t~~ .~~~~~~~t~~nimid experimentation.

A

most fascinating for uw Of the proposed behavior classification system is as a ~y~t~rn~~~~ lan for use in computer studies of small group ses. ecause of its multi~d~m~nsional description of all behavior B” in a common framework and because it is capable of fairly molecular levels of d~~~riptio~, it may be possible to e ploy the language in

computer stu e use of high-speed equipment to simulate group processes is ning to receive some attention in view of the complexity of group processes and the inability of existing laboratory techniques and experimental desi s to handle very many factors simultaneously. As a First step in this ~~~~~~,one could collect descriptive data about group e and then attempt to recreate them abilities and theoretical assumptions. production of actual interactions could be done, the way is various antecedent conditions an out on 8 comp~lter. his would be followed by empirical validation, pirically derived probability relationships and ch an approach wmdd tremndously accelerate our progress in the field, and a language which bridges between group processes and computer requirements is a critical link. The present approach may offer such a possibility.

128

IRWIN ALTMAN !SuMMARY

In this paper the general criterion question regarding small group behavior, namely, ‘How is the group doing?’ was expanded to include questions conce the total behavior of group members as they worked on a tas well as those behavio interpersonal behavior, their goal-contributory behavior directly related to task performance. Thus, the criterion question to include a very broad range of behaviors which conceptually and link to each other and which eveiltually interact and combine to group output. From such a starting point an attempt was made general1‘language’ of behavior which would include under its umbrella the vast numbers of types of behaviors possible, would link them to one another and which could be applied over a variety of situations. The advantages and limitations of such a general behavior class SJ*tern were discussed, spectically with respect to the description of social interaction in small group situations.

REFERENCES ALWAN, I., JS~SKINS, I. P. and MCGRATH, J. E., 1959. The translation of small group research information for computer analysis. Arlington, Va.: Human Sciences Research, Inc., October (AFOSR TN 59-l 194, ASTIA 4D No. 230 241). and MCGRATH, J. E., 1959. A lconceptual framework for the integration of small group researc,h information. Arlington, Va.: Human Sciences Research, Inc., February (AFOSR TN 59-252, ASTIA AD No. 212 252). , PENDLETON, CATHERINE A. and T~RAu=)s,ANXTA.1960. Annotations of small group research studies. Arlington, Vn.: Human Sciences Research, Inc., October, 1960 (AFOSR TN 60-1208, ANITA AD No. 248 440). and TERAUDS, ANITA.1960. Major variables of the small Arlington, Va.: Human Sciences Research, Inc., November, 1960 (AFOSR TN 60-1207, ASTIA AD No. 250 740). ARRINC~ION, RUTH E., 1943. Time sampling in studies of social critical review of techniques and results with research Psychol. Bull., 40, 81-124. BALES,R. F., 1950. Interaction process analysis: A method for the study of s,ma?l groups. Cambridge, Mass.: Addison-Weslell. ---, ‘1953. The equilibrium problem In small groups. In T. Parsons, R. F. Bales and EL A. Shils (Eds 1. Working papers in the theory of action. Glencoe, Ill.: Free Press, 11 l-261. --and ~RQDIIIBECK, F. L., 1951, Phases in group problem solving. I. Abnorm, LSoc. Psychol., 46, 485-495. BOR~ATFA, E. F., et al., 1956. On the dimensions of group behavior. Sociometry, %!@,223-240.

ORAL

ALL SROUP dESEARCN

129

S., 1955. On the classification of groups. nce of individuals ~er~n~Iit~?, and interest

ation of the criteria of leadership.

embers. .I. Abnorm. orizations and ratings in the observation elat., 4, 239-254. ncepts and metnod in the measurement of group of small groups. ity assessment and nd WEE% L. G., 194% The dimensions of syntality in small groups.

Relat., 6, 331-356. asuring human relations: An intr of individuals. Genet. Psychol. e measurement of interpersonal behavior. Trans. I’d. Y. he group structure of combat rifle squads. 952. A factorial study of the rated behavior of oup members. Paper read at Eastern Psychological Association, Beu’rson, ha., 1949a. A theory of cooperation and competition. Hum. Relat., , 129-152, . 949b. An experimental study of the affects of cooperation and tition upan group process. Wum. Relat., 2, 199-231. T., et al., 1950. Measurement of self-oriented needs in discussion Pou~~~o~, ups. J. Abnorm. SW. PsychoI., 45, 682-690. FREEDMAN,M. IL, et al,, 1951. The interpersonal dimension of personality. J. Pars,, 20, 143-161. Gan, C. A,, 1954. Leadership. In G. ILindzey (Erl.), Wandbook of social psychology, &mbrii:ge, Mass.: Addison-Wesley, 837-920. GOLEMBIEWSIU, R., 1962. The small group: An analysis of research concepts and operations. Chicago, Ill.: University of Chicago Press.

130

IRWIN

ALTMAN

P., 1962. Handbook of small group research, Glencoe, Ill.: The Press of Glencoe. HAYTHORN, W., 1953. The iufluence of individu of small groups. J. Abnorm. Sot. Psychol., HEMPHILL, J. K., 1956. Group dimensions: Columbus, Ohio: Ohio State Univers (Research Monograph No. 87). and COONS,A. E., 1956. Leader behavior desc~ip~~u. In I+‘: lvior: IQ Description and Measurement. Columbus, Oh State University, Bureau of Business Research (Rese,,rch No. 88). and WESTIK,C. M., 1950. The measurement of J. Psy;hol., 29, 325-342. HE’S, R. W. and LIPPP~T=,R., 1954. Systematic observational techniques. In G. Lindzey (Ed.), Handbook of social psychoilogy. Cambridge, Addison-Wesley, 370-404. HOMA,N~, G. C., 1950. The human poup. New York: Harcourt, Brace. JE~GS, VIXEN H., 1950. Leadership and isolation (2nd PA_.).New York: Longmans, Green. MCGRATH,J. E., 1962%.Systems of information in small group research studies. Arlington, Va.: Human Sciences Research, Iuc., April (AFOSR Document No. 2416). , 1962b. A summary of small group research stuaes. Arlington, Va.: Human Sciences Research, Inc., June [AFOSR Document No. 2709). and ALTMAN,I., 1966. Small group research syrithesis and critique of the field. New York: Holt, Rinehart & Winstan. NEWCO~B, T. M., 1953. An approach to the study of communicative acts. Psychol. Rev., 60, 393-404. ROBY,T. B. and LANZEITA,J. T., 1959. Considerations in the analysis of tasks. Psychal. Bull., 55, (2), 88-101, SAKODA,J. M., 1952. Factor analysis of OSS situational tests. J. Abnorm. Sot. Psychol. 47, 843-852. havior. *LUTZ, W. C, 1959. .EZRO: A. three-variable theory of interpersonal New York: Rinehart & Company, Inc. *AW, M. E., 1962. Scale andysis of group tasks Gainseville, lorida: University of Florida, October (Annual Report, Contract No. 580 (11)). of a measure of social , 1949b. The development and evaluation interaction. I: The development of reliability. Hum. Relat. 2, 103-121. SmK, DOROTHYand THELEN, H. A., 1958a. Emotional dynamics and group culture: Experimental studies of individual and group behavior. New York: New York University Press. and 1958b. Group culture and emotional dynamics. Chicago, Ill.: &iversity of Chicago Press, %cx;DnL, R. M., 1959. Individual behavior and group achievement. N&v- York: Oxford University Press. HARE,

A.

L

S fN

ALL

GROUP

l

RESEARCH

631

Some characteristics of c. Psychol., 50, 1R5-109. iences Research, Inc.,

sychology of groups.

rc’n Foundation,