Chapter 8 - Group Process

THIS CHAPTER WILL DISCUSS:
1. What group discussion "functions" are.

2. How interactional researchers study group process.
3. Whether group process relates to group output.
4. Whether group discussion consists of a series of sequential states.

INTRODUCTION

In Chapter 1 we examined the concept of "perspectives." As we explained, scientists approach an object they wish to study with a particular viewpoint, or perspective. Each perspective suggests distinct questions for the scientist.

Further, with small-group research, it is best to approach different areas of study from the viewpoints of particular perspectives. For example, we can study conflict from any perspective; but to examine conflict in the most profitable way, we based our discussion in this book on the relational perspective. Similarly, it is perhaps best to approach topics such as conformity and deviance from the structural viewpoint. Equally, much of what we covered in our chapter on cohesiveness came from the two psychological perspectives.

In this chapter, our topic is the process of communication in decision-making groups. How can we best approach this topic? The theories and research studies that we will describe relate to two viewpoints that are variants of perspectives that we defined in Chapter 1.

The first viewpoint, the functional approach, is a variant of the structural perspective. We will use this approach when we discuss theories and studies that relate to the functions that communication plays in group discussion.

The second viewpoint, the interactional approach, is a variant of the relational perspective. We will use this approach when we examine theories and research that concern the way group discussion becomes patterned in repetitive sequences over time.

In the following sections, we will discuss many examples of scientific work on group process, both from the functional and the interactional approaches.

THE FUNCTIONAL APPROACH TO GROUP PROCESS

The Concept of "Function"

Before we discuss how communication functions during group discussion, let us look again at the functional approach. In Chapter 1 we discussed the general claims of functional theorists. We review those claims again now.

According to functional theorists, any social system has a set of goals toward which it directs its actions. Its primary goal is to survive and maintain itself. Beyond this, each type of social system also has other goals particular to its type.

Decision-making groups are a type of social system. As with any social system, a group has a primary goal of maintenance. In addition, groups have the particular goal of performing a task. These two goals are of utmost importance to groups. Their importance has led to the central distinction, in this book, between task and maintenance variables.

To reach its goals, a social system must perform a set of necessary actions that theorists have labeled "functions." In essence, the functional perspective concerns the behaviors of people in a social system.

What about a group? How does it perform the functions it needs to reach its goals? A group uses communicative statements. By talking with one another, group members perform the necessary behaviors. In other words, a successful group discussion is what takes a group to its goals.

For example, a group is planning a party. The members need to use communicative statements to perform their task. Statements such as "Should we hire a band?", "We ought to serve pretzels and popcorn," and "So we have decided to string ribbons around the room" all help the members plan their party. Each time a person makes such a statement, he or she does an action that helps the group toward its goal. In this way, members perform the functions the group needs.

Positive and Negative Functions
The above example statements help the group. Not all statements, however, serve positive functions. Members can also make statements that serve negative functions and that do not help. For example, the party-planning group may be plagued by one negative member who keeps complaining about all the work he did for the last party. He makes statements, such as, "Nobody ever thanks me for my effort," "Who was it that made all those trips to the store?" and "No way I'm gonna do all that again." These statements also play a function in the group situation, a negative function. Negative functions do not help the group and may prevent the group from reaching its goals.

Types of Functions

It is not enough simply to conclude that statements serve positive and negative functions during group discussion. Scientists also want to be able to describe these functions. In other words, a statement is not merely positive or negative; it has other qualities. Does it summarize the discussion? Does it show agreement? These kinds of questions help scientists find out what types of functions exist.

More generally, scientists want to be able to identify the kinds of positive and negative functions that statements can have in a group. They want to categorize the behaviors to understand them better. A scientist who has this knowledge can observe a group's discussion and see if the members perform the positive functions they need and avoid the negative functions they do not need. Based on this observation, the scientist can predict whether the group will reach its goals. Such a researcher could also recommend to group members the types of statements they should and should not make to reach their goals.

Hence, scientists are interested in the discussion functions that are possible in groups. Several theorists have proposed lists of group discussion functions. We can consider each of these as a theoretical view. The theory concerns the types of functions groups need to perform to achieve their goals.

In the following sections, we will describe three of these proposals.

Benne and Sheats's Approach to Discussion Functions

"Functional Roles"

Benne and Sheats (1948) formulated their list as a result of what they had observed in groups. They did not present a list of discussion functions per se but proposed a list of what they called "functional roles." Benne and Sheats believed that they had seen members play these "functional roles" in groups they had observed. In essence, each member had taken on a kind of persona within the group that related to the kinds of statements that person would tend to make. Benne and Sheats labeled each functional role they observed. For example, they might label a person who consistently tries to lessen discord in a group as the "harmonizer."

Benne and Sheats divided their list into three categories: (1) group task roles, (2) group maintenance roles, and (3) individual roles. Group task roles focus on the task at hand and include, for example, roles such as the "elaborator," the "coordinator," and the "orienter." Group maintenance roles focus on keeping the group together and include roles such as the "encourager," the "harmonizer," and the "compromiser." Both task and maintenance roles help the group pursue its goals.

Individual roles are negative functional roles and do not help the group move toward its goals. People playing these roles attempt to satisfy their own needs and desires and work against the group as it tries to achieve its goals. Some examples of individual roles include the "aggressor," the "dominator," and the "recognition seeker."

Accompanying Functions

Accompanying each role is an analogous function. The name that Benne and Sheats gave each role describes the kind of function that the person playing that role would perform. For example, the "coordinator" would make statements that provide the "coordination" function for the group.

As we have stated, Benne and Sheats originally listed functional roles. Some scientists and practitioners who have elaborated on their original work have not focused on the roles that group members can play, but on the functions themselves.

There have been several such elaborations. Tables 8.1 and 8.2 show one example. These lists are based on contributions from Professor Erma Jones of Temple University and from the Friends Peace Committee and Life Center in Philadelphia, Pennsylvania. As you can note, the lists include only task and maintenance functions. These help the group move toward its goals. The lists do not cover the individual functions that Benne and Sheats believed harmed the group's ability to achieve its goals.

Applications of Benne and Sheats's Approach

Which of the task and maintenance roles on Benne and Sheats's list are most critical for a group that wants to reach its goals? The scientists make no overall claims in answer to this question nor do they make any assertions about how often group members should play these roles.

We can perhaps implicitly assume from their work that it is good for a group when the members frequently take on task and maintenance roles. The more members play roles that fall into these two categories, the more successful the group will be. Benne and Sheats, however, do not discuss this explicitly. They make few recommendations on how a group could successfully utilize functional roles.

Table 8.1	Group Task Functions
Function	Purpose	Behaviors
Initiating	Give direction and purpose to group	Proposing tasks and goals, defining problems, suggesting procedures and solutions
Information seeking	Make group aware of need for information	Requesting relevant facts, clarification
Information giving	Provide group with information relevant to its work	Offering relevant facts, avoiding reliance on opinion when facts needed
Opinion seeking	Test for consensus, find out group opinion	Asking for feelings or opinions
Opinion giving	Provide basis for group decision	Stating feelings or beliefs, evaluating a suggestion
Clarifying	Eliminate confusion	Defining terms, interpreting ideas, indicating issues and alternatives
Elaborating	Reduce ambiguity, show consequences of plans and positions	Giving examples, developing meanings, explaining
Coordinating	Adjust issues or harmonize issues that may conflict	Suggesting ways to handle issues
Procedure developing	Establish an order to the meeting	Suggesting agenda, order of business
Summarizing	Show how ideas are related, draw ideas together	Pulling together related issues, showing contradictions, restating suggestions, offering conclusions
Philosophizing	Show that a particular issue is not unique	Drawing general statements from specific ones, critically examining assumptions and ideas
Evaluating	Keep group in line with goals, provide sense of progress	Measuring accomplishments against goals, noting progress and blocks
Testing agreement	Find out how close group is to agreement	Stating ideas of agreement, asking if agreement is possible
Energizing	Keep group working on problem	Prodding the members to action
Orienting	Keep group in direction set by agenda	Guiding discussion by keeping group on track, moving discussion along
Mediating	Reconcile disagreements	Conciliating differences, offering promises

Table 8.2	Group Maintenance Functions
Function	Purpose	Behaviors
Encouraging	Bring out others' opinions and give others recognition	Speaking positively to and of others, accepting others' contributions
Expressing feelings	Call group attention to own and others' ideas and reactions	Expressing own feelings and restating others' feelings and opinions
Harmonizing	Reduce tension, allow group to express feelings	Joking, clowning, breaks and task-irrelevant statements
Compromising	Maintain group cohesion	Offering or accepting compromises, yielding status, admitting error
Facilitating	Maintain open discussion, keep channels open	Drawing out silent members, suggesting procedures for discussions
Standard setting	Make group aware of direction and progress	Expressing group concern, suggesting tasks, stating standards for group to achieve
Interpreting	Explain what someone has said	Paraphrasing initial speaker
Observing	Making group aware of its feelings	Evaluating the mood of the group
Following	Provide stimulation and support for speakers	Accepting others' ideas, going along with group

This does not mean that Benne and Sheats make no observations that relate to this topic. They have some relevant thoughts. One is that a group would probably require different roles during different stages of its discussion. For example, a group that is beginning to talk about its task would probably have little need for an "evaluator." On the other hand, a group

that has progressed toward its decision would probably need someone to play that role.

Benne and Sheats point out that group members should not limit the range of roles they can play. In a group with a rigid role structure, each member plays only one or two roles during the whole discussion. Such a group will probably not be as successful as a group that has a more flexible structure. A flexible structure would allow everyone to play many roles. This kind of system helps a group effectively utilize the talents of its members.

Bales's Approach to Discussion Functions

The "Equilibrium Problem"

Unlike Benne and Sheats, Bales (1950, 1953) used a well-developed theory of group process to develop his list of discussion functions. The theory he used is a good example of the functional perspective.

Two Conflicting Goals

As you will remember, functionalists theorize that groups try to achieve two major goals: maintain the group and perform the group's task. To achieve each goal, the group must perform certain actions.

For example, a group has the task of cleaning a room quickly. The demands of the task equire certain actions. For instance, the group must be strictly organized and work hard. On the other hand, the group cannot deny its other goal, maintaining itself as a group. This goal also requires certain actions. For instance, members may need to joke with one another and to spend time forming friendships. As you can see, the cleaning group has task and maintenance demands that are in conflict. In fact, the demands oppose each other.

Bales believed that it is inevitable that task and maintenance demands oppose each other, which leads to a major problem: Task performance usually results in maintenance problems. For example, if the group cleans the room quickly, it probably doesn't give itself enough time for the necessary social interaction. As the group tries to deal with these maintenance difficulties, more task dilemmas arise. If the cleaning group takes time to allow members to form friendships, it probably is not cleaning the room as quickly as it could.

Task Changes and Maintenance Problems

Further, changes in a group's relationship with its task will cause maintenance problems. For example, a group is reassigned from a coordination task to an accuracy task. This change requires adjusting the group's structure. Changes will include altering its division of labor and its power hierarchy. As we have seen, faced with a coordination task, a group relies equally on all members to perform their part of the task competently. Faced with an accuracy task, however, the group should rely on only those members most competent to perform the task. Hence, the group probably will change from one in which members share responsibility and power to one in which the most competent members hold responsibility and power.

The change to a power hierarchy will mean improved task performance. The group's maintenance structure is disturbed, however. We would expect that the members who suddenly find themselves peripheral to the task will lose satisfaction in their jobs. This will lead to losses in cohesiveness, which could threaten the group's survival. When the group realizes that it has lower morale, it will try to rebuild cohesiveness by changing its social structure. Once again it will give more responsibility to the peripheral members, leading to a redefinition of task roles. Ultimately, the group might return to a shared power structure. If it did, it would not be able to perform the accuracy task as well as it could using a power hierarchy.

The danger is that a group will swing back and forth between task and maintenance crises. Bales called this the "equilibrium problem," because a group can manage the problem by developing properties that serve to balance these extremes. It can do so by repeating sequences of actions that alternately dampen task and maintenance disturbances. By carefully tracking both sets of problems, the group can maintain its balance.

Sequential Task Problems

As we have seen, the first goal of a group is to maintain itself. The second goal is to perform its task. Functionalists believe that communication performs a special role in task performance. Van Lear (1996) has described this belief:

1 - Group decisions do not automatically occur. They must be developed.

2 - Communication and its associated functions are necessary for a developed group decision.

3 - Decisions tend to develop through a series of stages or phases. This is because certain communicative functions are needed to move the group from one stage to the next.

The task of the theorist is to determine what the phases of group task performance are, and how communication functions to help the group complete each phase and move on to the next.

In Bales's view, successful task performance rests on the group's ability to solve three problems: orientation, evaluation, and control. Orientation involves the members coming to a common understanding and definition of the task. For example, members in the cleaning group should all understand that their task is to neaten a room.

Evaluation involves members' developing common values regarding what a good solution must accomplish. For example, should the room be so clean that it sparkles when the group is finished, or will a relatively clean room be a good solution? Control involves the members' finding the best solution for the task, using the power and influence relationships among its members. Looking at the cleaning group, who should sweep the room and who is good at polishing?

As you can see, the group must solve the three problems sequentially. Each requires successfully solving the previous problem. For example, if the group decides that the room doesn't need to really sparkle, it is not necessary for the best-qualified people to do the jobs that they do. well. Any group member could probably do a fair job; thus, the problem of control changes.

Bales's Research Methodology

As you can recall, according to functional theory, each group has a necessary set of functions--actions that a group must perform to reach its goals and solve its problems. In turn, communicative statements are the actions that groups can use to perform these necessary functions. It follows from this that Bales would measure group communication to evaluate his claims.

How did Bales measure this communication? To create a methodology he proposed 12 functions. Bales believed that these functions are critical for groups as they work to maintain equilibrium and solve their orientation, evaluation, and control problems. Bales used his list of functions to perform content analyses of the communication that occurred during the group meetings he studied. Observers "coded" the group's communication in terms of the functions ir performed. Hence, we will use the term "coding scheme" to describe a list such as he created.

Table 8.3 illustrates his coding scheme.

Table 8.3	Bales's Interaction Process Analysis coding scheme
Categories	Specific Examples
Positive Reactions
1. Shows solidarity	Jokes, gives help, rewards others
2. Shows tension release	Laughs, shows satisfaction
3. Shows agreement	Understands, concurs, complies, passively accepts
Attempted Answers
4. Gives suggestion	Directs, proposes, controls
5. Gives opinion	Evaluates, analyzes, expresses feelings or wishes
6. Gives information	Orients, repeats, clarifies, confirms
Questions
7. Asks for information	Requests orientation, repetition, confirmation, clarification
8. Asks for opinion	Requests evaluation, analysis, expression of feeling or wishes
9. Asks for suggestion	Requests directions, proposals
Negative Reactions
10. Shows disagreement	Passively rejects, resorts to formality, withholds help
11. Shows tension	Asks for help, withdraws
12. Shows antagonism	Deflates others, defends or asserts self

As you can see, the last six functions are "mirror images" of the first six. The first three and the last three functions relate directly to one another. All have to do with group maintenance. Functions 1, 2, and 3 are "positive reactions." Bales believed a group needs these functions to maintain its cohesiveness. Functions 10, 11, and 12 mirror them and are their opposites. They are "negative reactions" that endanger group cohesiveness. For instance, "shows solidarity" is a positive reaction; "shows antagonism" is a negative reaction.

The other six functions relate to the group's goal of completing its task. They still mirror one another, however. They are, in a sense, complementary. One set of behaviors asks questions; the other set attempts answers. The "questions" in functions 7, 8, and 9 mirror the "attempted answers" that are functions 4, 5, and 6.

In general, Bales presented a wealth of findings from his studies of groups. As you can recall, in Chapter 2 we discussed his finding that the differences in group members' talk time increases as group size increases. More specifically, Bales used his coding scheme to test his hypotheses about group equilibrium and about the problems of group orientation, evaluation, and control. Later in this chapter we will discuss these findings in some detail.

For the present, it is sufficient to refer to one of Bales's overall claims that group functioning is not so much a result of having many "good" functions as it is a result of using those "good" functions properly. Benne and Sheats may have proposed that a successful group has a high quantity of task and maintenance functions. Bales asserted that this was not quite the case. He claimed instead it is more important that groups properly balance task and maintenance statements. In other words, Bales believed that groups seem to do well when these statements are in a certain proportion and occur in a certain sequence.

Hirokawa's Approach to Discussion Functions

Essential Functions

Hirokawa (1982) claimed that scholars should relate group task performance to communicative functions that are essential for the group's task. Essential functions are those that groups must perform to complete their tasks successfully.

Hirokawa argued that not all earlier functional approaches had focused on these essential communicative functions. Hirokawa singled out Bales's theory for criticism. As you can recall, Bales's coding scheme includes functions such as "asks for opinions," a phrase that does not specify an essential function. Hirokawa said that, under this heading, group members can ask all kinds of questions that might or might not relate to the group's task.

For instance, a member can ask a question such as, "What opinions do you have regarding the proposals we've made so far as we've looked at the decision we have to make?" This action is relevant to the group's task. As such, it is an essential function. In contrast, another member could interrupt the group's task work to ask, "Where should we go to dinner after our meeting?" This action is not relevant to the group's task. Therefore, according to Hirokawa, it is not essential. Nevertheless, Bales would count both questions in his coding scheme and say that both perform equally well as functions under the heading of "asks for opinions."

With all this in mind, Hirokawa created his own list of functions that he believed clearly related to groups' tasks. Hirokawa looked at earlier work on this topic and concluded that four functions meet his requirements:

1. Establishment of operating procedures

2. Analysis of possible solutions

3. Generation of possible solutions

4. Evaluation of possible solutions

Similarity to Other Lists

Unfortunately, Hirokawa's list has the same problem that he believed Bales's list had. The categories can contain a statement that does not relate to a group's task. Members can perform a function from this list that is irrelevant to a task just as easily as they can perform a function that is relevant.

We can go back to Hirokawa's own example to see how this could happen. A group member could easily "generate a solution" to the non-task issue of where the group members should go to dinner after the meeting. Hence, Hirokawa was not able successfully to isolate the essential task functions. This problem does not, however, in any way invalidate his list.

Research Questions
At the time of his essay, in 1982, Hirokawa was not sure exactly how best to relate group task performance to the functions on his list. Is the number of functions or the sequence in which they occur more important? Which relates more to how well a group does?

As we will discuss later in this chapter, Hirokawa went on to study these questions.

The Three Levels of Discussion Functions

Many Proposals to Utilize

We have just discussed three proposals that look at communicative statements within group discussion. Each attempted to describe the types of functions that those statements can perform.

What can we do with all these proposals? When we see a statement, how should we analyze it? Do we have to pick and choose from among the proposals, or can we combine them?

For example, someone makes the following statement during a group discussion: "Several corporations have adopted the second proposal." How would this statement fit into each proposal we have described? Benne and Sheats might have decided that it serves the function of "elaborating." On the other hand, Bales might have considered that it performed the function of "giving information." In contrast, Hirokawa could have claimed that the statement fulfilled the function of "evaluating a proposal."

In addition, each theory is specific to the relationship between discussion functions and a group's performance. Which formula is most likely to ensure success? What might make a group fail? Benne and Sheats would probably have claimed that how well a group does depends on the number of functions its members perform. Bales, instead, would have said that a group's performance depends on either the proportion or the sequence of functions that its members act out.

Method to Blend Proposals

Can all of these ideas be correct? Can one statement, such as our example above, perform many types of functions? If yes, can these functions relate to one another in any way? Finally, can functions interact with group performance in all of these ways?

The answer to all three questions is yes. One method can utilize all these proposals simultaneously.

Based on Philosophical Studies of Actions

To understand this method, let us offer an example. You are unhappily watching a comedian perform at a supper club. You decide you want to see the comedian leave the stage. You believe that the comedian may leave if you register displeasure. To show this displeasure, you decide to throw a dinner roll at the comedian. You know how to do this concrete action. All you need to do is to pick up the roll, move your arm forward quickly, and let go of the roll at the proper moment.

Philosophers interested in human action have studied such situations. Essentially, someone watching you would only see you throw a dinner roll. Philosophers would say that in reality you have simultaneously performed at least four action. All four actions relate to one another. How?

Levels of Abstraction

The actions relate through levels of abstraction. Each is on a different level. Some are more abstract; others are less so. In essence, the more abstract the action, the more involved the thinking is at that level of action.

For instance, the action of "throwing a dinner roll" is more abstract than the action of "moving your arm forward." Someone who saw your arm move could say with some certainty that you moved your arm forward. If the person wants to know more about why you moved your arm forward, the situation is more complex. If the person moves to a higher level of abstraction, he or she could discover that the action of deciding to throw a dinner roll led you to move your arm.

Relationships Between Levels

Further, the levels of abstraction relate in two ways: the "in-order-to" relationship and the "by-means-of" relationship. It is as if the levels were on a ladder. As you move up each rung of the ladder, you seek the motivating action behind the action on your current rung. As you move down the ladder, you seek the actions behind the motivating action on your rung. In other words, you move up to answer the question of why, and you move down to answer the question of how. If someone asks why, you can begin your answer with "In order to ..." If someone asks how, you can answer with "By means of...."

When philosophers look at one action and move up to the more abstract action related to it, they see an "in-order-to" relationship. They see an action that is rather concrete, such as an arm that moves forward. To discover the less concrete action behind it, they can ask themselves, "Someone performed this action in order to do what?" For example, you moved your arm forward "in order to" throw the dinner roll.

When philosophers want to move the other way, from more abstract to less abstract actions, they see a "by-means-of" relationship. People act out their less concrete actions by means of more concrete ones. For example, you wanted to register your displeasure. This desire was a rather abstract action. You acted out this abstract action "by means of" throwing a dinner roll.

The "act-tree" in Figure 8.1 diagrams the relationships among these actions (Goldman, 1970).

FIGURE 8.1

Thus, we can "describe" the same physical action in many ways. In this case, the physical action is the forward movement of an arm. Figure 8.1 shows that the arm movement did four things. It moved forward, threw a dinner roll, registered displeasure, and convinced a comedian to leave a stage. Each action is on a different level of abstraction, but all are related. Each level connects to the one above it through an "in-order-to" relationship and to the one below it through a "by-means-of" relationship.

The relationships among levels are flexible. A person can register displeasure "by-means-of" many lower level actions other than throwing a dinner roll. Similarly, someone can throw a dinner roll "in-order-to" perform many higher level actions that have nothing to do with registering displeasure.

Application of Method to Communication Functions

Now that we have examined the ideas behind the "act-tree," we can discuss how we can use this method to look at communication functions.

Functions Are Actions

Communication functions are a type of action. Therefore, we can analyze them in the same way that we analyze other actions. As can any other action, a statement that a group member makes can simultaneously serve several functions. In essence, there are many answers to the question of why a group member said something, in the same way that there were many reasons you moved your arm as you sat in a dinner theater.

Similarly, each function of a statement can reside on different levels of abstraction. Every level relates to a more abstract level via "in-order-to" relationships and to less abstract levels via "by-means-of" relationships.

The "Act-Tree" and the Three Proposals

Now let us go back to our example statement and analyze it using the "act-tree." We see that the sentence, "Several corporations have adopted the second proposal," can simultaneously serve many functions. It can simultaneously serve all the functions that we said the three proposals claimed. We find that the three proposals--from Benne and Sheats, Bales, and Hirokawa--all present claims that can work together.

We propose that as a representation of what takes place, the act-tree in Figure 8.2 may be defensible:

FIGURE 8.2

As we can see, the statement "Several corporations have adopted..." simultaneously serves functions at different levels. At the highest level of this diagram, the statement evaluates a proposal. At the lowest level, it gives information, and so on.

The functions themselves remain on their levels. For instance, elaborating will always be a higher level function than simply giving information. The relationships among these functions are flexible, however. Many functions are on each level. For example, a person can give information "in order to" perform a middle-level function other than elaborating. Similarly, someone can elaborate "by means of" a low-level function that is not giving information.

The "Act-Tree" and Group Performance

Because of this flexibility, functions on different levels can relate to group performance in different ways. For example, in one group many statements may serve the function of elaborating, but they do so by means of several lower level functions. Hence, if we code the functions that relate to how well that group does, we would find several lower level functions at work, but only one middle-level one. We could find an infinite number of such combinations.

With this in mind, we can examine again the theories of how communicative functions relate to group performance. We can now look at the functions on different levels. It is conceivable that we can link low-level functions to group performance through their proportion, which is what Bales believed. It is also possible that middle-level functions relate to group performance through sheer number, as Benne and Sheats's approach may imply. Hence, both theories may be valid at the same time because the functions exist on different levels.

Conclusions

The moral of the story is that the levels of functions are important. Theorists who advocate the functional approach to group discussion need to remember this. When they make claims about group discussion, they must take care not to mix levels of abstraction. Further, they cannot assume that the relationship between functions on different levels is inflexible. Finally, these theorists must not believe they can relate group performance to all functions in the same way. They need to be careful about which level of the functions they wish to relate to group performance. Functions from different levels perform differently, and any theory about group performance must account for this.

Later in this chapter we will return to the functional approach. We will discuss the results of functionalist research into the process of group discussion. In the meantime, however, we will turn to the second approach--how interactional theorists look at group process.

THE INTERACTIONAL APPROACH TO GROUP PROCESS

As we mentioned in Chapter 1, and again earlier in this chapter, advocates of the interactional approach look at group patterns that form over time. They see groups as aggregates in which, over time, repetitive communication sequences become patterned. As a consequence, group process is of utmost importance to them. They wish to discover the processes that lead to patterns. Scientists who have adopted the interactional approach focus their efforts on the study of group process.

The theoretical underpinnings to account for this focus are in an essay by B. A. Fisher and Hawes (1971). They distinguished between two approaches to studying small groups--the Human Systems Model, or "HSM," and the Interact Systems Model, or "ISM."

A "system" exists, technically speaking, when a number of objects are interrelated. The relationship of the parts is such that the action of one object affects the action of all other objects in the system. Both the relational and structural perspectives, including all their variants, view a small group as such a system. This idea differentiates them from the psychological perspective. A scientist from the structural or the relational perspective is more concerned with the system as a whole than with its individual parts. The interactionalists agree with this view.

Human Systems Model (HSM)

Most variants assume that the Human Systems Model (HSM) applies to groups. This model holds that the "objects" in a system are "people." HSM theorists characterize the systems they study by the relationships among the people. These relationships include variables such as the degree of cohesiveness and the power relations among group members.

Interact Systems Model (ISM)

The ISM (Interact Systems Model) derives from the extreme interactional view. In contrast to the HSM, this model holds that "units of communication" are the objects that interrelate in a system. Theorists using the ISM call the units of communication "acts." They characterize the systems they study by referring to recurring patterns of these acts. In so doing, they examine which pattern of interaction leads to another pattern, and so on.

The ISM extreme interactionalists believe that the basic unit of analysis is not the unit of a dyad but rather a system of two "acts," the interact. The acts are conversational utterances. To be part of an interact, the acts must be contiguous, that is side-by-side in time, one occurring immediately after the other. The acts must also be probabilistically related.

What does it mean to be probabilistically related? Let us consider Bales's coding scheme as an example. As we described above, this coding scheme includes categories such as "Gives suggestion." Now let us imagine that you were to perform a content analysis of a group discussion using Bales's coding scheme. How often would group members give suggestions? Based on Bales's (1953) research, we might imagine that about five percent of members' utterances would fit in this category.

An interactionalist would not be satisfied with this information. An interactionalist would want to know the circumstances under which a group member is likely to give a suggestion. Once again looking at Bales's data, we find that when one member makes an utterance that is coded "Asks for suggestion," the odds are 36 percent that another person will respond by giving one. Looking at these two pieces of data together, it seems that giving a suggestion is a relatively rare event overall, but if someone has just asked for one, giving a suggestion is fairly likely to occur. Asking for and giving suggestions are then probabilistically related.

Example of Interactional Perspective Research

The goal of interactionalists when they research group process is to determine whether the utterances in group discussion are probabilistically related. If so, then we can say that group discussion has a "sequential structure." Baird's (1974) research is a good example. Baird compared the interaction of eight groups of four or five members each. Baird gave the first four groups "cooperative" instructions. He told members that he would grade their performance as a group. The second four groups received "competitive" instructions. Baird told participants that their grades would be based on their relative individual performances.

Results showed that competitive group members tended to initiate and disagree more than cooperative group members. Also, persons in the competitive groups informed one another and agreed less than the members of the other four groups. We can see the results in Table 8.4 with numbers representing the proportions of acts in each category.

Table 8.4	Baird's Data
Category	Cooperative Groups	Competitive Groups
1 - gives information	.540	.466
2 - expresses agreement	.177	.148
3 - asks for information	.132	.113
4 - initiates and develops theme	.085	.163
5 - expresses disagreement+	.066	.110

The differences between cooperative and competitive groups also affected the sequential structure of the interaction of the groups. Before we discuss these results, we offer a note to help in the interpretation of the data. Remember from our example above that suggestions occur about five percent overall but more than a third of the time given that someone has just asked for one. We need to interpret this data the same way. For example, if cooperative groups give information 54 percent of the time, then there is sequential structure in the group discussion if giving information follows another communication act more or less than 54 percent of the time. If giving information follows another act more than 54 percent of the time, we will say that it occurs more often "than expected." Similarly, if giving information follows another act less than 54 percent of the time, we will say that it occurs less often "than expected."

The diagrams in Figures 8.3 and 8.4 use the categories from Baird's research that we listed in Table 8.4. They show Baird's most significant findings regarding how often interacts in his groups varied from what we would expect.

FIGURE 8.3

FIGURE 8.4

Note that some interacts occurred more often than acts in both cooperative and competitive groups. For example, "initiates and develops theme" was followed by "asks for information" more often than expected. Similarly, "asks for information" was followed by "gives information" more often than expected. We can conclude that these interacts play an important function in any group discussion.

Other interacts differed between the groups. Some of the disparities are unexpected. For example, "initiates and develops theme" both followed and is followed by "expresses agreement" more often in competitive situations than in cooperative ones. Some differences are more predictable. In competitive groups, "expresses disagreement" was followed by "expresses agreement" less often than by chance, a sign that conflict was occurring. In contrast, in cooperative groups, "expresses agreement" was followed by "expresses disagreement" less often than by chance, a sign that conflict was not occurring. In other words, conflict existed more often in competitive situations than in cooperative settings.

Conclusions

During the 1970s, ISM modelers produced a great deal of research into group process. However, as a whole we did not learn a lot from these studies. There are several reasons for this failure. One of these reasons was that most ISM scientists believed that by just looking at their data, they could eventually come up with theories. Researchers outside the ISM realm, such as Bales, often work differently. They prefer to start with a theory that, in effect, tell them what to look for in interaction. They then test their theory to see if the research supports it.

This approach is not perfect. A theory can, for example, keep researchers from seeing patterns in their data that may be important but that are irrelevant to their theory. It is also true, however, that without a theory researchers may not be able to find any patterns. ISM modelers appeared to have been merely looking at their data. If instead they had looked for something in particular, they might have found it.

Second, there is a problem even if ISM researchers had been able to produce theories about group interaction. These theories would not have told us anything particularly useful. The ISM modelers were willing to study how HSM input variables affect group process. They were uninterested in describing the effects of process, in turn, on HSM output variables. However, the process-output relationship is just as important as the input-process linkage. Task groups do not talk simply in order to talk. They talk so that they can perform a task, leading to output variables. By ignoring the impact of process on output, ISM researchers made it impossible to learn how process can be used to help groups perform their task or get along with one another better from their perspective. How can this problem be solved? By studying the relationship between group interaction and group output. This is what researchers such as Bales and Hirokawa have done.

For the remainder of this chapter we will discuss group-process research related to potentially useful issues. These studies have attempted to answer two general questions: First, does group process relate to group output? And second, can scientists divide the process of group discussion into a series of phases? We will discuss each question in turn. The studies we will describe refer to ideas already introduced in this chapter.

THE RELATIONSHIP BETWEEN GROUP PROCESS AND GROUP OUTPUT

Bales

Earlier in this chapter, we described Bales's belief that groups have an inherent problem balancing two opposing goals: group maintenance and task performance. Each goal places opposing demands on a group. Thus, groups have an "equilibrium problem."

Bales believed that groups use communication in different ways to perform this balancing act. Group output depends upon how well they do it. Therefore, he theorized that a researcher should examine a group's communication to see the functions it plays during discussion. By doing this, the researcher can see how successfully the group solves its "equilibrium problem." Bales proposed a coding scheme with 12 categories to use when he content analyzed group discussion. We discussed this coding scheme earlier in this chapter.

During the late 1940s and early 1950s, Bales performed a great deal of research on group discussion. We will now look closely at this research, starting with a summary of how he performed his studies.

Method

Bales (1953) brought together groups of three to eight people. The groups met to discuss what Bales called "human relations cases." The researchers gave members five-page summaries relating the difficulties of a fictional administrator in dealing with subordinates. The group's decision-making task was to consider why these difficulties arose. In addition, they were to decide what could be done to solve them. The researchers gave the groups 40 minutes to perform the task. When they finished, members rated their reactions to the discussion. They also noted their level of satisfaction with the group, their views of the other group members, and their opinions of the group as a whole.

As the groups worked, observers behind one-way mirrors classified every communicative act according to three criteria: who said it, to whom it was said (for example, to an individual member or to the group as a whole), and the act's function.

Research Results

Acts. Overall, in Bales's groups, an average of 25 percent of a group's statements were positive reactions. Only 11 percent were negative. Bales believed that negative reactions indicated maintenance disturbances. Positive statements needed to outnumber the negative to restore equilibrium. For example, a person might need to give two or three compliments to make up for one harsh word.

Indeed, in one example of a satisfied group, positive reactions were 34 percent of the total; negative reactions were only 5 percent. In contrast, one dissatisfied group had only 16 percent positive statements and 17 percent negative. As we can see, the more satisfied group made a far greater proportion of positive statements as compared with negative statements than did the less contented group. Also, overall the more satisfied group performed more maintenance acts, or statements classified as negative or positive.

In task-oriented discussion, the groups had an average of 57 percent attempted answers. Only 7 percent were questions. Satisfaction levels were unrelated to the number of attempted answers. Dissatisfied groups, however, did ask relatively more questions than contented groups (10 percent versus 4 percent). Overall, in dissatisfied groups the number of questions increased at the expense of maintenance, negative or positive, statements. Less satisfied group members seemed to spend more time trying to understand their tasks than they did trying to get to know one another.

Interacts. Groups deal with the problem of equilibrium at the level of interacts. Remember, Bales hypothesized that a group needs to balance the opposing pressures of task and maintenance needs. He believed that groups could do this by alternating between task statements and maintenance statements. Several research results speak to this hypothesis.

For example, Bales saw that, when a group member reacts positively or negatively, he or she is likely to continue with an attempted answer. Any further continuation is also likely to be an attempted answer. The diagram in Figure 8.5 illustrates this relationship between contiguous communication acts by the same person.

People dampen their own maintenance reactions by using attempted answers. They try to tell why they feel as they do. For example, a group participant might express the negative reaction, "It seems as if we don't use our time very well," and then continue with an attempted answer that helps to soften the negative words, such as, "Maybe this is because we aren't as organized as we could be." After saying this, the person will probably yield the floor to another member, which also helps to mitigate the negative statement.

FIGURE 8.5

FIGURE 8.6

Contiguous statements can also come from different people. In this case, positive and negative reactions are also balanced by attempted answers, as the diagram in Figure 8.6 shows.

As you can see, attempted answers in this scenario lead to statements of agreement. For example, Amy says, "I think we need to focus more on our work." Bill continues the conversation by attempting to answer Ann's negative statement. He says, "Maybe we need an outline." Ken agrees, saying, "An outline is a good idea, Bill. I've worked in other groups who did that." Such a conversation pattern is consistent with theory. Theory holds that task-oriented activity seems to build up tensions that are broken by positive maintenance acts. When the tension started to build as Bill talked about an outline, Ken's positive words helped to ease the tension.

In the diagrams above we omitted links that were either smaller than .1 or irrelevant to the equilibrium hypothesis. One strong regularity that we left out is worth mentioning, however. Attempted answers tend to follow questions. This is not surprising, particularly when contiguous statements come from different group members. Bill might ask, "Does anyone know how to make an outline?" and Ken may continue with the attempted answer of, "I think I do. Let's get some paper."

Conclusions

In summary, Bales was able to show through his research that groups use communication to solve the equilibrium problem. A group builds up tensions when it performs task functions. Positive reactions dissipate these tensions. The process that groups use to dissipate tensions takes place on two levels: acts and interacts.

On the level of acts, Bales claimed that satisfied groups make many more positive reactions than negative ones. In these groups, members use positive reactions to bleed off the tension that comes through task work and negative reactions. The positive messages help restore balance. In contrast, dissatisfied groups do not have such a preponderance of positive reactions. Without these positive reactions they are unable to bleed off the tension successfully.

On the level of interacts, Bales was able to see in detail the process groups use to dissipate tension. In a satisfied group, the process on this level often follows a certain pattern. First, tension builds when members make negative reactions. Second, to begin to lessen the tension, members often use an attempted answer. Finally, when one member attempts an answer, another member will often respond with a positive reaction. This circle of messages completes the dampening process. Afterward, the group is ready to return to task work, using questions or attempted answers.

Thus, both Bales's theory and his research findings imply that a successful group's output is based on the proper proportion and sequence of communicative functions. Group members need to know how to balance and coordinate their statements if they want their group to do well.

Of course, keep in mind that Bales worked with functions that are on a low level of abstraction. His findings support his ideas only as far as those particular functions are concerned. What if one were to examine discussion functions at a different level of abstraction? Such an examination could reveal a different relationship between group process and group performance.

Hirokawa

As we discussed earlier, Hirokawa (1982) focused on communicative functions on a high level of abstraction. His belief was that scientists should relate group performance to these high-level functions. In the decade after that proposal, Hirokawa performed a series of studies to find out which functions are essential for successful groups. In other words, which high-level communicative functions must groups perform to reach high-quality decisions? In Hirokawa's studies, he asked groups to make decisions about policy issues including topics such as traffic safety and plagiarism. He then brought in experts on these topics to judge the quality of each group's decision. Finally, he analyzed the group discussions to see which high-level communicative functions each group performed. He wanted to find out the relationship between these functions and the quality of each group's decision.

Method

Hirokawa used several methods to analyze group discussion. Before we discuss his results, we need to understand the shortcomings of one of his methods. In this method, Hirokawa trained people to observe groups and look for sets of possible functions. The observers watched videotapes of group discussions and then judged the extent to which the groups performed each set of functions. A serious problem arises because of halo and horns effects.

As you can recall, in Chapter 3 we discussed halo and horns effects. These effects take place, for example, when a research participant sees a picture of an attractive man and, without knowing anything more about him, judges him intelligent and exciting as well as attractive. This happens because many people believe that these three attributes--attractive, intelligent, and exciting--usually "go together." Such a belief affects the participant's judgment of the attractive person.

This technique of Hirokawa's is problematic because it relies on judgments from observers. We have good reason to believe that observers, despite their training, are vulnerable to the halo and horns effects. For instance, observers are likely to believe that a "good" group will perform "good" functions. Similarly, they probably think that a "bad" group will not perform good functions. The process could start when they see a group do a good function. After they see this, they probably will label that group as "good." Once that happens, they are likely to expect the group to do well, and they will judge that it performs all of the good functions.

Studies indicate that this happens. For example, in one study (Hirokawa, 1988, Study 1) observers strongly tended to look at the functions they were judging and to decide that groups had performed all these functions to about the same extent. Thus, a method that relies on observers' ratings does not result in trustworthy findings.

How then should scientists measure which functions occur in group discussion? The most valid method is the one that Bales used. Coders should observe group discussion and, using a coding scheme, note when functions actually occur.

Hirokawa did use this technique in some of his studies. For example, he often used a coding scheme when he studied the actual discussions of groups. His coding scheme consisted of high-level communicative functions (for example, Hirokawa, 1988, Study 2). The studies that used a coding scheme have more trustworthy findings than the studies that used observers' ratings.

Conclusions

Hirokawa concluded from his research that groups need a small set of "critical functions" to make successful decisions. Hirokawa has presented various lists in his essays. The following five functions have recurred regularly:

1. The group must come to understand the nature of the dilemma they face.

2. The group must agree on the requirements for an acceptable solution.

3. The group must come up with a range of realistic alternative proposals for solutions.

4. The group needs to thoroughly and accurately assess the positive consequences of each alternative proposal.

5. The group must thoroughly and accurately assess the negative consequences of each alternative proposal.

Hirokawa's research has led him to conclude that when communicative statements help groups meet these requirements, they bring about a high-quality decision.

Hewes's Challenge

Hewes (1986) challenged the conclusions of scientists such as Bales and Hirokawa. He did not agree that group discussion affects group output.

Implications of the Input-Process-Output Model

To present his challenge, Hewes utilized the input-process-output model of group discussion. He considered some implications of the model to help him with his work. As we discussed in Chapter 1, the input-process-output model diagrams what we should think about when we study group discussion.

As the model shows, a scientist needs to keep a number of things in mind. For instance, to understand group discussion, a scientist needs to think about the variables that precede it and to look at how such variables affect the process of group discussion itself. These input variables include the size of the group, the group's task, and the amount of cohesiveness among its members. A scientist then also needs to consider how the process of discussion affects group output variables. A researcher will understand how group discussion works only if he or she simultaneously considers all these variables and their relationships.

Using this model, Hewes approached the question of how group process affects group output by focusing on input variables. He proposed that these variables are the most important factor as far as group output is concerned. Therefore, group process loses its significance. Hewes argued that, if a scientist wants to show that communication truly affects group output, he or she must show that the effects of communication go over and above the effects of input variables.

The general argument is that the group members' initial qualities, not what they discuss when they come together, most affect their group's output. This general argument has already come up several times in this book. One significant example appeared in Chapter 2, and another was in Chapter 7.

Examples of General Argument

In Chapter 2 we asked if people should perform tasks individually or in groups. We often showed that the answer depended on what affects decision accuracy. Did interaction among group members lead to the output, or was output instead an aggregation of members' work? The Lorge/Solomon Model A looked at accuracy tasks. It was based on the assumption that interaction has no effect on accuracy when groups work on these kinds of tasks. Instead, Model A predicts a group's accuracy by looking at the members individually. The prediction is based on the odds that members can solve the problem individually. Thus, Model A directly relates an input variable to an output variable. The input variable consists of the group members' individual knowledge and skill. The output variable is the group's accuracy.

Social decision schemes work the same way for quality tasks. They begin with each group member's prediscussional opinion. They then predict how these opinions will be "combined" into the group's decision. It likewise presumes that interaction has no impact on this combination.

In Chapter 7 we discussed this general argument again when we looked at social influence. We examined five positions concerning how social influence works in groups. In the first four, scientists believed that they could predict a group's decision by looking at the members' prediscussional ideas. Prediscussional ideas comprise an input variable, and the group's decision is its output.

Position 1 is based on social decision schemes. Under Position 2, we examined social comparison theory, which explains the group polarization effect by hypothesizing that members affect one another by sharing their prediscussional opinions. The persuasive arguments theory from Position 3 differs a bit from the first two positions; however, it also emphasizes prediscussional ideas. Someone can predict a group's decision by looking at the arguments that members have before they talk with one another. In every case, as you can see, these proposals hypothesize that an input variable sufficiently accounts for group output. Only Position 5 differs. Its structurational approach claims that group discussion affects group output and that this effect is over and above the effect of group input.

Hence, many theories incorporate Hewes's general idea. His specific proposal challenges researchers. He believes that a scientist who wishes to claim that group process really affects group output must show that the effects of group process are independent of the effects of input variables. Group process must do something special that input variables alone do not do. When Hewes wrote this essay, either Bales nor Hirokawa had considered this issue.

Possible inadequacies of earlier research

For example, Hirokawa claimed that when a group performs critical functions, it reaches high-quality decisions. Hirokawa, however, ignored the possibility that input variables might actually account for his findings.

We can consider some scenarios. For example, in one group, members are particularly knowledgeable about how groups should make decisions. They believe they know the correct steps. They realize they must understand both the dilemma itself and the requirements for an acceptable solution to this dilemma. Further, they understand that they should propose several realistic alternatives and assess both the positive and the negative consequences of each alternative. In short, the members know that groups should make decisions using the techniques that Hirokawa found at work in good groups. During their discussion, each member of this group thinks individually and performs all these "critical functions." Essentially, the members work on their own. They all come up with the same high-quality decision because they have individually followed the same steps. In this case, group discussion would have had little effect on the group's good decision.

In contrast, the members of another group do not know much about how groups should make decisions. Again, the members think through things individually. This time, during group discussion, the members on their own each come up with the same poor-quality decision. As in our first scenario, group discussion has had little effect on decision quality. The poor decision comes out of the individual processes of group members, not out of the processes that the group goes through as a whole.

Hewes's Proposals

The importance of individual aspects. These two scenarios make up part of what Hewes claimed actually happens during group discussion. He suggested that the members' individual aspects determine the quality of their group's decision. These individual aspects include the members' personalities, motivations, knowledge, and decision-making skills. Hewes believed that what happens during group discussion has little impact on the quality of a group's decision. In essence, Hewes believed that, if group members individually make poor decisions, putting them together in a group will not help them to make a better decision. Similarly, if people know how to make good decisions, a group discussion will not hinder their abilities and lead them to a poor decision.

"Thinking aloud." If Hewes is right, why have researchers such as Hirokawa found a relationship between what happens during a group's discussion and the quality of the decision that a group makes? Hewes has a ready answer.

In Hewes's view, group members face two tasks when they engage in group decision making: (1) participating in a discussion and (2) formulating their own thoughts about the group decision. To complete the first task, members need to make relevant comments in response to others. To complete the second task, they must concentrate on the decision. Hewes's idea was that when these two tasks come together they create some problems for group members. He felt that people cannot simultaneously perform both these tasks well. They have to let one fall by the wayside. Therefore, they concentrate on the second task and focus their attention on thinking hard about their decisions.

This does not mean that people forget about their first task. They know they are supposed to be talking with one another. To help themselves out of this dilemma, they vocalize what they are thinking. In essence, their minds are filled with their own thoughts, but they need to talk; so they end up thinking aloud. The consequence is that group members speak but do not listen. Hewes believed that group discussion consists of members thinking out loud but not listening to one another.

"Vacuous acknowledgments." If Hewes is correct, why does it seem to an outsider that group members react to one another's statements? Hewes claimed that members say things that make it seem as if they are seriously listening to one another. In actuality, Hewes said, they are not. Members' statements are "vacuous acknowledgments." Through the use of these statements, members sound as if they are reacting to comments when in truth they are thinking aloud. Hewes presented the following example of how a member could use a vacuous acknowledgment (1986, p. 282):

Claude: How about building an underpass for bikers on Sierra?

Steve: I can see your point, but how about more money being put into traffic lights?

Steve's statement "I can see your point, but" is a vacuous acknowledgment. The acknowledgment makes it seem that Steve's comment is in response to Claude; yet the actual content of Steve's statement has nothing to do with Claude's idea. The proposals are completely different. In Hewes's view, Steve is not really listening to Claude. Instead, he prefaces his own "thinking out loud" with an acknowledgment that makes it appear that he is responding to Claude. In reality, he is not. Hewes claimed that an observer should not be surprised to see this pattern at work. Group members say things to each other such as, "Yeah, and we could also . . ." or "Right, but I've also been thinking about . . ." and then proceed with comments that have nothing to do with what they supposedly heard. Whenever members behave this way, they are doing what Hewes claimed group members normally do.

How "good" groups appear "good." How could Hirokawa have seen "good" groups if the members were mainly thinking aloud? Hewes's idea was that the individual behavior of members can make a group appear "good." For instance, group members have the same beliefs about how groups should make decisions. They all seek to follow the same pattern. The members of this group are likely to think of the same sorts of things at the same time. If they voice these thoughts, they would appear to respond to one another when in actuality they are not.

Similarly, imagine that the members individually are knowledgeable and skillful. Their group is likely to make a good decision; however, group discussion is not an important ingredient in that decision. The group would come out of the discussion with a good decision because the members went into the meeting knowing how to make good decisions on their own.

Actual challenge

We have said that Hewes challenged the scientists who have studied group discussion. In essence, he challenged them to prove his idea wrong. To do so, scientists must show that group discussion consists of more than vacuous acknowledgments and thinking out loud.

How can researchers accomplish this? They need to demonstrate that during group discussion members make comments that actually respond to the content of other members' statements. Only in this way can they show that group members are really working together and interacting, rather than ignoring one another's ideas and merely thinking out loud. Further, they must show that the effect of group discussion on group output is over and above the effect of input variables.

At present, no study has succeeded in showing these two things. No study has successfully isolated group discussion and made its effects independent of group input. Therefore, as yet no study unambiguously demonstrates that group discussion, independently, actually affects group output.

Several studies have, however, tried to answer Hewes's challenge.

Answers to Hewes's Challenge

During the 1990s, Hirokawa has taken Hewes' challenge seriously. He has acknowledged the possibility that group process may not have a significant impact on group output. Hirokawa has not, however, come out totally in favor of Hewes's standpoint. Rather, he has come to believe that there are situations in which group process has no effect on output. In contrast, there are other situations in which process is critical for group performance. The job of the researcher is to determine when process is and is not important.

Hirokawa and his students have written several essays in which they have proposed various input factors that affect the importance of group process in determining group output. They have also performed some studies to evaluate their proposals. Thus far, all of these studies have problems, and none have succeeded in answering Hewes's challenge. We will consider perhaps the best of these, performed by Hirokawa's student Salazar.

Salazar Study

Salazar proposed two input variables that he thought might impact the process/output relationship. One of these was group homogeneity. Salazar hypothesized that when group members have similar beliefs or preferences, they find it relatively easy to reach consensus, and this consensus will reflect those beliefs or preferences. As a result, it will be those beliefs and not group process that will have the largest impact on the group's decision. In contrast, when group members' prediscussional preferences are dissimilar, then they need to communicate quite a bit to make a decision, and one cannot predict what the decision will be based on those preferences. In this case, it is group process and not prediscussional preferences that will impact the group's decision.

The second input variable Salazar considered was task complexity. When tasks are simple, group members know how to perform them, and need not spend much time discussing the issue. Thus group process will not have a large effect on their performance. However, when tasks are difficult, group members must interact to figure out how to do it. In this circumstance, process takes a more critical role.

Method

In his study, Salazar used the "Lost on the Moon" accuracy task that we discussed in Chapter 2. As you recall, participants are asked to rank-order 15 items in terms of their survival value for a space crew stranded two hundred miles from their mother ship. Participant rankings are then compared to those made by survival experts. In the "low task complexity" conditions, participants were asked to rank-order five of the items that experts had ranked as very dissimilar in their survival value (those ranked 3, 6, 9, 12, and 15). The researcher reasoned that group members would find it easy to judge the relative value of each for the lost crew. In the "high task complexity conditions," participants were asked to rank five items that were seen by the experts to be very similar (11, 12, 13, 14, and 15). In this case, Salazar thought that the participants would have a hard time distinguishing which of the items were more or less important for survival.

Salazar then formed groups that were either in prediscussional agreement (high member homogeneity) or disagreement (low member homogeneity) to perform either the high or low complex version of the "Lost on the Moon" task. He content analyzed their communication using Hirokawa's (1982) coding scheme, in order to see which conditions led to more or less of the "critical discussion functions." To measure group output, he also evaluated the accuracy of each group's decision.

Findings

First, Salazar analyzed the relationship between his input and process variables. He found that groups with low member homogeneity performed more discussion of the problem, the requirements of a good solution, and the positive and negative consequences of alternative solutions than groups with high member homogeneity. This supports his idea that a difficult task would lead to more relevant discussion than a simple task. Task complexity, however, had very little impact on critical discussion functions.

Salazar then attempted to study the relationship between process and output. He found that the accuracy of the member's prediscussional preferences accounted fully for group accuracy when the groups were high in homogeneity, no matter how complex the task. In these cases, communication had no effect. In contrast, when the groups were low in homogeneity and the task was complex, the accuracy of member prediscussional preferences had no impact on group accuracy. This time, communication was important. Finally, when groups were low in homogeneity and the task was simple, both prediscussional preferences and communication were important. However, each condition only included about ten groups. This is not nearly enough for an accurate analysis, and these results cannot be trusted.

Conclusion

Setting aside the small number of groups, can Salazar's study be considered an answer to Hewes's challenge? Not really. Hewes insisted that researchers have to study the sequential structure of group discussion. If there is no sequential structure, then there is no reason to believe that group members are really responding to one another as they make their decision. We have no way of knowing whether Salazar's participants are really working with one another or whether they are just thinking aloud in one another's presence.

Imagine the following possibility: A non-homogeneous group is performing the complex version of "Lost on the Moon." During the group meeting, each member is thinking through the problem on their own out loud, not really listening to one another. As they do so, they come to a better individual understanding of it, and change their minds about the rank-ordering of the items. Their final individual rankings are more accurate than their prediscussional preferences, and so they "combine" these rankings into an accurate group judgment. Their communication reflects their individual thought processes, which makes it appear as if communication has led them to a more accurate group answer. However, as they were never truly interacting, communication is not really having any impact. As unlikely as this possibility might appear, Salazar's findings cannot rule it out.

Despite its methodological problems, and despite this latter possibility, there is no question that Salazar's study is a serious attempt to show that group discussion affects group output. We hope that future research overcomes these difficulties and comes closer to a definitive answer to the question.

POSSIBLE PHASES IN GROUP DISCUSSION

We now turn to the second general question about group process that has intrigued scholars. This question concerns the manner in which group members conduct their discussion.

Bales and the Phase Hypothesis

At times in this chapter we have discussed Bales's contributions to group discussion theory and research. Now we shall examine his phase hypothesis. To introduce this hypothesis, let us review what we have already discussed about Bales's work.

Earlier we explained Bales's theoretical views of the equilibrium problem. We also described his research into the question of how groups try to solve this dilemma. We then discussed Bales's argument that groups must solve three additional problems before it can perform its task.

The first problem is that members must come to a common understanding, or orientation, toward the group's task. Second, the group needs to create a common set of values, or evaluation, about an ideal solution. Third, group members must use their influence and power relationships, or control, to come up with a good solution. In essence, a group needs to find its own orientation, evaluation, and control techniques before it can perform its task.

Further, a group should solve these three problems in the order given. A group must come to a common understanding before it can find a method of evaluation, and it must have a common method of evaluation before it can choose the best solution.

As we stated earlier, Bales argued that groups use discussion to solve the equilibrium problem. He also believed that groups solve these three additional problems through discussion. It follows then that different parts, or phases, of the discussion should involve specific problems. The communication during "phase one" should be concerned with "orientation," for example, and so on. The group's discussion should reflect these relationships.

This line of reasoning is known as the "phase hypothesis" of group discussion. To test this theory, Bales divided his group's discussions into thirds. He then compared the relative amounts of acts in each third. His results are shown in Table 8.5.

Table 8.5		Test of Linear Phase Hypothesis
Rank	Segment 1		Segment 2		Segment 3
	Function	Percent	Function	Percent	Function	Percent
1	information	35	opinion	38	opinion	35
2	opinion	31	information	25	pos. reaction	30
3	pos. reaction	20	pos. reaction	22	information	19
4	suggestion	6	neg. reaction	8	suggestion	9
5	neg. reaction	6	suggestion	7	neg. reaction	8

Bales theorized that the group dealt with the problems of orientation, evaluation, and control, in that order. The findings bear him out to some extent.

The problem of orientation involves gaining a mutual understanding of the issue the group is examining. It requires that members exchange relevant information primarily during the first phase of discussion. As these results indicate, Bales found that the first segment did contain a greater proportion of asking for and giving information than did the second and third segments. This finding was consistent with Bales's expectations.

To solve the problem of evaluation, group members need to form mutual values for evaluating proposed solutions. For example, if group members decide to make an outline, they also need to agree on how they will know when the outline is good. This requires that members share opinions concerning these values. As evaluation is the second of the three problems, asking for and giving opinions should be proportionately higher in that section than in any other. The data show that opinions were highest in the third section, but not by very much.

The problem of control requires that the group reach a consensus on the best solution. Group participants suggest solutions to solve the problem. Thus asking for and giving suggestions should be at their highest proportional level during the third segment. Again, this was the case, but only by a very tiny margin.

As we can also see in Table 8.5, the total number of positive and negative maintenance messages went up from segment to segment. Bales concluded from this finding that as the group progresses on its task, its tension level goes up and more negative maintenance statements occur. To counteract these, the group responds with more positive maintenance. Although it is not shown in Table 8.5, Bales also found more subtle evidence concerning negative and positive messages in the control phase. He found that, during the first half of the third segment, groups exhibited a larger proportion of negative reactions than they did during the second half of the third segment. The proportions of positive reactions fell and rose accordingly. This finding implies that the group has a great deal of conflict during the first half of the control phase, and resolves the conflict during the second half. It warrants dividing the control phase into two subphases, negative and positive.

Additional Research

Bales's contributions are among the largest of any scholar in the history of small-group research. First and foremost should have been his theoretical work. The field has largely ignored his accomplishments in that area, however. Instead, Bales's research methods and findings, divorced from the theory they intended to evaluate, have proved far more influential.

Bales's phase hypothesis has been the most well known part of his work. His coding scheme, however, has had almost as much impact on research. In the 1950s and early 1960s scores of studies retested the phase hypothesis. Many employed Bales's coding scheme despite the fact that he designed it specifically to test his theory. Its relevance to other studies would therefore be questionable.

Tuckman (1965) reviewed this additional research. He concluded that groups pass through four phases, with each phase containing both maintenance and task phases that run roughly concurrently. Hence, four maintenance phases and four task phases occur together. The phases are as follows:

Phase One--Forming

Maintenance - "Testing and dependence"-discovery of the limits of acceptable behaviors during interaction.

Task - "Orientation"-attempts to define the task and the manner in which the group will accomplish it.

Phase Two--Storming

Maintenance - "Development of intragroup conflicts"-polarization around interpersonal issues, resistance to emerging structure of interpersonal relationships.

Task - "Emotional response to task demands"-resistance to attempts at influence, conflict between individual orientations and task demands.

Phase Three--Norming

Maintenance - "Development of group cohesion"-acceptance of a structure for interpersonal relationships.

Task - "Open exchange of relevant interpretations"-expression of opinions dealing with the problem.

Phase Four--Performing

Maintenance - "Functional role-relatedness"-use of the now-agreed-upon relationship structure to solve interpersonal problems.

Task - "Emergence of solutions"-constructive attempts to complete the task.

Tuckman's analysis implies that a group's social structure is important for the task. The degree to which a group's social structure has evolved determines the group's ability to perform its task at each phase. The social structure serves as a foundation for the performance of tasks.

The Linear Phase Model of Group Process

Bales and his immediate successors left the legacy of the "phase hypothesis" for future researchers. We have shown how this hypothesis relates to the process of group discussion. The theory has come to be known as the "linear phase" model of group discussion because it implies that groups handle problems sequentially. A group will go through each prerequisite stage before starting subsequent stages. It also will never go back to earlier stages.

The linear phase model may be valid in the area of concern that scientists such as Bales and Tuckman researched. They were interested in the function of interaction in the development of a group's social and task structures. The linear model is deficient as a complete hypothesis, however. Later theorists turned their attention to the development of specific ideas under discussion in a group. As they did so, it became apparent that the linear phase model was insufficient for characterizing group process. Nevertheless, keep Bales's work in mind as we continue to examine theories concerning group process.

Scheidel and Crowell

A coding scheme for interaction analysis is like a scientific perspective. Both point out certain features of the "real world" but hide others. We can see this in Bales's coding. His coding scheme points out the functions of statements in group discussion but hides the manner in which specific ideas develop over the course of a discussion.

Scheidel and Crowell (1964) were among the first scientists to escape the domination of Bales's influence. They proposed the following coding scheme.

1 - Initiation of new idea	7P - Positive modification
2 - Restatement	7C - Negative modification
3 - Clarification	8A - Stated agreement
4 - Substantiation	8B - Stated disagreement
5 - Extension of idea	9 - Synthesis
6 - Minor response	10 - Summary
Q - Question

As you can see, Scheidel and Crowell included only task functions. The maintenance functions of acts are absent from the chart. Thus, the equilibrium hypothesis is invisible from the view this scheme implies. In addition, Scheidel and Crowell instructed the coders to note the proposal that each act, or statement, covered. This allowed them to trace how proposals developed during the discussion. Bales did not do this.

Research results

In their major research, Scheidel and Crowell used their coding scheme to analyze the discussion of five groups. Each group consisted of four to seven members.

At the first level, the level of acts, approximately one-fourth of the conversation involved initiating, extending, modifying, or synthesizing proposals. In other words, one-fourth was creative work. The second fourth dealt with clarifying and substantiating proposals. A third fourth consisted of evaluating proposals. The researchers did not mention the final fourth. We presume that it was maintenance oriented and thus irrelevant to the scientists' analysis for this study.

More important, researchers discovered much about the sequential, or interact, structure of the discussions. Statements that clarified/substantiated or evaluated usually followed communication acts that "created" proposals. These researchers called this tendency "reach-testing." Group members first followed the process of "reaching out," in which they initiated a new proposal or developed an old proposal. Groups then seemed to need to "test" their ability to accept the proposal by clarifying it or evaluating it. If the result of the test was affirmative, the proposal became an "anchor." The group could then go through another "reaching out" by using the "anchor" as a base.

If this process continued, one member's initiation became progressively modified until slowly it represented the entire group's point of view. In many cases, however, the idea received a negative evaluation. The group then returned to a previous anchor and reached out in a different direction.

Spiral model

In many instances groups reintroduce proposals that they had previously discussed. Scheidel and Crowell emphasized this finding in their discussion. They characterized this finding as indicative of a "spiral model" of group process. This model opposed the linear model in two ways. First, the linear model maintains that groups never look back. Groups should move in a linear pattern and follow a continued progression through stages of discussion. The spiral model allows the group to look backward.

This difference is not really relevant, however. Scheidel and Crowell found spiraling in the ways that groups developed their proposals. Bales and Tuckman saw a linear progression in the ways that groups developed their social and task structures. As we can see, the researchers were looking at two different features of group discussion. The findings do not necessarily conflict.

Bales's second interact diagram, showing contiguous statements by different speakers, illustrates how groups use statements of agreement following task work. Bales considered statements of agreement maintenance acts. In this way, his diagram shows a tendency that is consistent with the equilibrium hypothesis. Scheidel and Crowell, on the other hand, thought that statements of agreement were task acts in themselves. For them, the statements of agreement showed reach-testing. Thus, both research programs found similar results. They both saw that statements of agreement followed statements that were task oriented. They interpreted these results differently, however.

Second, the linear models of Bales and of Tuckman limit the time when the group works on the problem. Bales said that such work takes place in later stages, after the group has gone through orientation and evaluation stages. Scheidel and Crowell's results show that the group works on its problem throughout the discussion.

In this case, Scheidel and Crowell's interpretation is justified. We can reexamine Bales's data on phases. The amounts of each type of statement across the three segments is important. These amounts are too similar to support the conclusion that task work is not being performed throughout the discussion.

Fisher

We have shown two models for group process from the functionalist perspective: (1) the linear model of group structural development and (2) the spiral model of group idea development. These models address different group functions, and each has merits and deficiencies. The potential exists, however, to synthesize a model that contains elements of both hypotheses. B. A. Fisher (1974) proposed a synthesis.

As did Scheidel and Crowell, Fisher used a coding scheme that focused on the task functions of communicative acts (see Appendix for a list of its categories). Fisher's scheme classifies acts according to function: interpretation, substantiation, clarification, modification, and evaluation. It also classifies them as favorable, unfavorable, or ambiguous toward the idea under consideration. Further, it is important, as with Scheidel and Crowell, that coders trace the progression of specific ideas during the discussion.

Fisher's own research supported the synthesis of linear and spiral models. He examined groups of four to seven people who met for 25 minutes to 30 hours. His work suggested a four-phase sequence. Positive statements were greatest in all phases. Differences occurred, however, in the relative number of ambiguous and negative acts.

During the first phase, the "orientation" segment, ambiguous statements were at their relative highest, almost as plentiful as the positive acts. During the second phase, the "conflict" segment, negativity was at its maximum point, and ambiguous statements fell to their minimum level. Further, analysis of the interacts showed that in this segment the tendency for conflict among the four phases was greatest. By conflict we mean periods when statements for and against an idea alternated.

During the third segment, the "emergence" phase, ambiguity reemerged, although it was still not as high as during the orientation phase. In addition, negativity dropped. One proposal seemed to win. Dissenters no longer objected to proposals they didn't like. Instead, their responses became noncommittal. In the fourth phase, the "reinforcement" segment, positive statements rose to their highest levels. They outnumbered ambiguous and negative acts by the largest proportion of all phases. The groups had achieved consensus; however, rather than immediately adjourning, the groups took time to express support for the chosen proposal. (Note the similarity of Fisher's findings with those of Hoffman in Chapter 7.)

Thus, as we can see, Fisher's work supported a linear model of idea evaluation. Scientists could apply the linear concept to ideas as well as to task acts. Fisher found that spiral "reach-testing" also occurred in the groups. He saw this as he traced specific ideas through the conversation. Even when a group seemed to support an idea, the group often dropped it. The group would then turn to another proposal for a while. Then, once again it would reintroduce the idea it seemed to support and discuss it further, often in a modified form.

In the end, Fisher's work supported aspects of both the linear and the spiral models. In this way he synthesized the two.

Poole

The work of Poole is the most significant recent advance relevant to the phase hypothesis.

Criticisms of the four-phase model

Poole's first article (1981) described two ways in which the research that supported the four-phase model was flawed.

First, the research had made arbitrary "segments." Many earlier studies had divided group interaction into segments that corresponded to the theoretical stages of the model. This action limited the number of "stages" that the researcher could find. For example, if a researcher divides interaction into four segments, he or she can find only four or fewer phases. The researcher cannot find five or more stages even if they exist. The segmentation dictates the research from the beginning. Researchers could not find more subtle changes in interaction than they had proposed because they initially divided their data into four stages.

Second, most studies combined the data from many groups before they analyzed it. Groups may be so different that researchers should not do this. If scientists look at the groups only as a whole, they may be unable to make valid generalizations about any one group. Any differences among groups might cancel out when researchers combine data from all of them. This situation is similar to when we talk about the "average person"--differences among people are ignored. All scientists, however, know that the "average person" concept is only a convenient fiction, not representing even one real person. Individual groups may be as different as individual people. If so, then researchers should keep this in mind. If they insist on combining data from different groups, then they must speak of the "average group." They should admit that the "average group" is a fiction just like the "average person." It may be a convenient concept, but it may not represent any group in reality.

Research

Poole (1981) performed content analyses of the discussion of five groups of students and five groups of physicians. He used both Bales's and Fisher's coding schemes as he examined the data. He found that his groups varied widely in the number and order of phases in their discussions, however, in contrast to earlier research. In further analysis (1983a), Poole found that five of his groups did approximate Fisher's model, going through orientation, conflict, emergence, and reinforcement phases.

In more recent research, Poole and Roth (1989a, 1989b) studied twenty-nine groups of different types as they went through forty-seven decision making discussions. They found that eleven of the discussions approximated the linear phase model and progressed from an analysis of the problem to an attempt to find solutions to it. In twenty-two of the other groups, the members had discussions that cycled back and forth, as many as seven times, from an analysis of the problem to a discussion of possible solutions. During the remaining fourteen discussions, the group members did not analyze the problem. Instead, they only discussed possible solutions.

Further, the researchers found that groups that most often followed a pattern like the linear phase model had certain traits. These groups tended to be relatively large, with six or more members, and have low cohesiveness and a diffused power structure. In addition, they often performed tasks that had no clear goals and many possible solutions. In contrast, the groups that most often ignored problem analysis entirely had different characteristics. Such groups tended to be relatively small, with three to five members, and have high cohesiveness and a centralized power structure. Further, they often performed tasks that had clear goals and only a few possible solutions.

These findings support the idea that group members have a concept of how groups "should" make decisions and that their concept is like the linear phase model. It also appears that members believe that their group needs to follow this model only when it will be difficult for the group to reach a decision. In essence, the members try to follow the model when they feel they need it, but not otherwise. For instance, it might be hard to reach a decision when a group is uncoordinated or its task is difficult. Members of this group will probably want to use their concept of "ideal" discussion. On the other hand, it is not so difficult to reach a decision when a group is coordinated or the task is easy. People in this kind of group feel that it is unnecessary to analyze the problem. They believe they can go right into discussing possible solutions, and they do not follow their concept of "ideal" discussion.

"Tracks"

Poole (1983b) also felt that it is misleading to propose that groups follow a series of phases. Instead, he believed that groups simultaneously progress along a series of tracks, with each track representing a feature of group process. At least three tracks correspond to earlier research models. One track for task process is similar to Tuckman's "task" model. A second track is for maintenance process and is similar to Tuckman's "social" model. Another track is for topic process and corresponds to Hoffman's "valence" model (see Chapter 7).

As you will remember, Tuckman believed that the development in social process serves as the foundation for the growth in task process. Poole's ideas contrast with Tuckman's concept. Poole envisioned that the tracks evolve independently and that one track need not necessarily be a substructure for another. Also, Poole believed that researchers must analyze each track separately. The researcher should not expect parallel development. Each track may have its own structural requirement, but growth rates might differ.

Tracks may, however, develop at the same rate. Such parallel evolution gives the illusion of consistent phases and approximates earlier models of group process. This convergence of parallel growth may represent an "ideal" type of group process. Perhaps groups should attempt to meet it. This idea awaits further theoretical and research advances.

Pavitt

As we have discussed, Poole's findings supported the idea that group members have a conception of how groups "should" make decisions. Some of our own research (Pavitt, 1992) attempts to learn about these conceptions, to see if they have any influence on the process of group discussion.

To find out how people think groups "should" work, the study focuses on the linear and spiral models of group discussion. The research looks at how the two models relate and asks a specific question. We will come to this question by first reviewing what we know about the models.

The linear model suggests that a group will perform an entire phase of the process before it moves to the next phase. The group leaves no stage unfinished to come back to later. Thus, group members ought to finish the task of proposing all possible solutions before they move on to the phase in which they discuss them. Similarly, they should discuss all solutions before they evaluate any of them. The linear model may come closest to the "ideal" method for making a decision. Nevertheless, Poole's research showed that group discussion comes close to the linear model only under certain conditions. If the conditions are not right, groups tend not to work in linear phases.

Research by Scheidel and Crowell and Fisher showed that groups often perform "reach-testing" sequences as they discuss proposals. In other words, group members often suggest, discuss, and evaluate one proposal before suggesting another. They go through each "phase" several times, each time they look at a proposal. They do not move in a steady progression during a meeting.

Thus, the specific research question of the study attempts to discover why groups "spiral" so often when scientists think that a linear process is better.

To answer this question, the study sought to find out what conceptions people have about ideal group procedure. What models appealed to the group members themselves? The participants' task was to imagine that they were members of a group that is making an important decision. Members would make and discuss three proposals during their discussion. Each participant then received 15 cards. On each card was a description of a possible step in the group discussion process. Participants received the cards in a random order. The task of each participant was to place the cards in the order in which they would want their group to step through the discussion process.

We had expectations about some of the cards. Two cards read "Define Problem" and "Establish Group Goal." We believed that most participants would place the "Define Problem" cards first and the "Establish Group Goal" card second, which they did in almost every case. Thus, these two cards represent the "beginning" of the discussion.

Other cards fit the study's expectations: "Eliminate Bad Solutions," "Improve Solutions," "Choose Best Solution," and "Improve Best Solution." We believed that these four would be the last cards in the participants' orderings. Although the specific ordering of these cards differed substantially, far more often than not, they generally were the last four. Hence, these four cards represent the "end" of the discussion.

The other nine cards read as follows:

Propose First Solution	Clarify First Solution	Evaluate First Solution
Propose Second Solution	Clarify Second Solution	Evaluate Second Solution
Propose Third Solution	Clarify Third Solution	Evaluate Third Solution

If a study participant thinks that groups ought to follow a linear process when making a decision, his or her ordering should reflect that. The order of cards would be: the two "beginning" cards, all three "propose" cards, all three "clarify" cards, all three "evaluate" cards, and then the four "end" cards. The cards will be in a different order if, instead, a participant thinks that groups ought to follow a spiral process. In such a case, the ordering would be: the two "beginning" cards, all three cards (propose, clarify, and evaluate) relating to the first solution, all three cards that relate to the second solution, all three cards that relate to the third solution, and finally the four "end" cards.

These two example orderings are extremely faithful to their models, the linear phase model and the spiral model. People can however "compromise." For example, a participant could believe a group should "spiral" between possible solutions as they propose and clarify them but wait to evaluate them all together in one stage. On the other hand is a participant who feels that groups should propose all possible solutions in one stage but then spiral between the solutions when clarifying and evaluating them.

The study results showed that about one-half of the participants ordered their cards to reflect the spiral model. We called these participants "reach-testers." Only one-fifth ordered their cards to reflect the linear phase model. We named these "linears." The remaining three-tenths created one or another compromise. That so few participants chose to order their cards in linear phases might explain why actual group discussion so often fails to meet this model.

This study is continuing and is now looking at the way the participants' group discussions progress. The question is whether each discussion actually follows the card orderings of the members of that group. Do the groups behave as the participating members felt they "ought" to? If so, we would expect that groups consisting of linears would use a linear process and that groups consisting of reach-testers would use a spiral process. We also want to find out what happens in even-sized groups that contain half linears and half reach-testers. What process will these groups use? Will the linears and the reach-testers overtly conflict about the correct process? Will the way they "choose" a process reveal anything interesting about power and social influence? We hope the research will reveal preliminary answers to these and other questions about group process.

SUMMARY

Researchers have studied group process from the viewpoints of two perspectives: the functional and the interactional.

Functionalists are concerned with the ways groups complete their tasks and maintain their cohesiveness. Functional researchers explore the aspects of group process that help or hinder these two aspects of groups. Scientists have proposed several functional approaches toward group process. We can view these approaches as existing on levels of abstraction. As a result, they are not necessarily inconsistent.

Interactional researchers, in contrast, are concerned with discussion patterns. They look at the types of statements that occur during group discussion and attempt to describe their sequential structure.

Overall, research in group process has been concerned with two major areas: the relationship between group process and output, and the way groups utilize types of statements during group discussion.

Bales has done extensive research in how group process relates to group output. His findings suggest that groups need to do certain things if they wish simultaneously to maintain their cohesiveness and solve their task problems. Bales looked at statements on a low level of abstraction that perform maintenance and task functions. He found that groups need to balance the proportion and sequence of these statements if they wish to do well. In contrast, Hirokawa found a set of "critical functions" on a high level of abstraction. He discovered that decision quality is associated with the performance of these functions. Hewes challenged the findings of both Bales and Hirokawa. Hewes argued that neither study showed conclusive evidence that group process, and not group input variables, relate to group output. Research has begun to answer Hewes's challenge.

In the second major area, concerning how group members use statements, studies of group process have examined how discussion progresses. The question has been whether groups change the proportion of types of statements as they go along. Scientists call this idea the "linear phase model." Bales hypothesized that groups must go through a series of three linear stages. Each phase has different proportions of varying types of statements.

Later research has found that the linear phase model is too simple an explanation. For example, cyclical processes occur as a group discusses specific proposals. This process is explained in the "spiral model." Groups may abandon an idea but then go back to it. Such action is not linear. Other research also found large disparities in the number and order of phases that individual groups go through during discussion. Research continues in an attempt to discover when group discussion will and will not follow the pattern of the linear phase model.