Chapter 9 - Imposed Group
Structures
THIS CHAPTER WILL DISCUSS:
1.
Networks that are imposed on groups regarding who can communicate with whom.
2. How
these imposed networks affect group process and output.
3. The media forms that group members can use to communicate with one another.
4. How these communication media affect group process and output.
5. The effect that computer conferencing has on group process and output.
INTRODUCTION
Structural
Perspective
We have discussed
aspects of the structural perspective in earlier chapters, and we return to
that perspective now. This time, however, we approach it from a different
angle.
Earlier we discussed the
issues of conflict, power, and conformity/deviance in relation to this
perspective. In those cases, what we examined came out of the usual
structuralist view of norms and roles, which is that, as group members continue
to interact over time, various norms and roles emerge. Norms and roles, then,
are the structures that form within groups.
Many groups, however,
cannot let their roles and norms emerge naturally from interaction. Instead,
they find themselves working under imposed
structures. This happens frequently when people form groups within
organizations. Outside forces, such as the head of a company, control how the
people must interact in the group. Structures are often imposed on
organizational groups in this way.
We can organize these
imposed structures into two overall types: the structure that restricts who can
communicate with whom in a group, and the structure that restricts how group members can communicate with
one another. An example of the first type is a company group in which the
members must communicate with one another through one manager. They cannot give
information directly to one another. An example of the second type is a group
in which members cannot meet face-to-face but must make their decisions via
telephone or written messages.
In effect, someone can
impose a structure on a group by controlling the way members communicate.
Communication is the key and both reflects and determines structure.
Organizational groups are an example. Companies often impose lines of authority
and methods of message exchange on their workers. A structural theorist would
say that an organizational hierarchy, or chain of command, is a structure of
imposed roles and norms. This structure, in turn, helps to determine the
authority and influence patterns among the members of the organization.
Hence, structural
theorists are able to use their viewpoint to examine imposed structures. They
see norms and roles within the imposed interaction. Further, researchers treat
both types of imposed structures as input variables. By doing so, they are able
to examine the effect of the structures on group process and output variables.
The findings help them determine which structures best meet different
organizational needs.
Three Areas of
Research
In the following sections we will explore the research that scientists
have done on imposed structures. First, we will examine research on structures
that restrict who can communicate with whom in a group. Scientists have come to
call these structures "communication networks." Second, we shall
discuss studies that have looked at structures that control how group members
can communicate with one another. These studies compare communication that
takes place face-to-face with communication that utilizes various media,
including telephone, teletype, and closed-circuit television. Third, and last,
we will describe how the use of computer conferencing also acts as an imposed
group structure.
IMPOSED
COMMUNICATION NETWORKS
Early Work
Ideas about imposed
communication networks combine theoretical and practical concerns because
people in organizations frequently work with them. These concerns have led to a
rather large body of literature on the effect of imposed networks. Scientists
have been interested in how such structures affect group process and output. A
great deal of research was done, in particular, in the 1950s and early 1960s.
Problems plagued this
early research. Because few strong theoretical positions emerged from it, by
the mid-1960s theorists had lost interest in pursuing experiments about imposed
networks. Hence, the research about imposed networks dates from the 1950s and
early 1960s. Despite their weaknesses, these early efforts did produce some
interesting and useful findings. A study of these efforts provides a great deal
of information about how one can approach a topic, such as imposed networks,
and create appropriate experiments. In this way we help fulfill one goal of
this book, that of helping the reader to understand how a scientist works.
Possible Networks for Groups
We will begin by
describing possible networks for five-person groups. In Figure 9.1, we show
five possible structures. Each number represents a group member. The lines
represent possible channels for message exchange. In each case, these channels
allow two-way communication. The members who are linked by lines can send
messages directly to each other. Other group members are indirectly linked.
They can send messages only with the aid of intermediaries. Either directly or
indirectly, all members can communicate with one another.
FIGURE 9.1 
When evaluating these diagrams, we need to consider four important indices.
They relate to the imposed group network under study and are a way that
scientists can measure interaction and quantify a group's network of
communication.
Distance
The first index
is distance, which is the
number of links between any two members. The shortest distance is one. For
example, the member at the center of the wheel configuration needs only one
link to reach an outside member. The members communicate directly through one
line of message exchange. The longest distance for any five-member group
network is four, which occurs when members are in a chain structure. Four is
the number of links that each member at the end of the line must use to get a
message to the other end member.
Summed distance
The second index
is summed distance, the
number representing the group members on the diagrams. Remember that
"distance" is the number of links a member must use to communicate
with other members. To compute the summed distance, we add the distances it
takes for one member to reach each other member in the group. This is not a
group total. It refers only to each individual member. For example, the summed
distance for the central member of the wheel is four. It only takes four links
for that member to communicate with all other people in the group. The summed
distance for each of the other four members of the wheel configuration is
different, however. An outside member would have to use seven communication
events to communicate with all other group members. The figure "7"
applies because only one link is needed to talk to the central member, but two
links are needed for each of the other three. Thus, the wheel diagram shows a
"4" for the central member and a "7" for each of the other
members. The summed distance is important because the structure of a group is
based only on this criterion. The way in which we draw each group is arbitrary.
The diagrams may look like a "wheel" or a "circle" here,
but the actual group configurations might be very different. For example, if
it would still be a
"Y" group. The reason is that each configuration still has members
with the right summed distance for the "Y" group.
Another example is the
wheel group. The previous diagram makes the wheel look like a centralized
structure that differs from the "corporate ladder," or hierarchy,
approach. However, looking at the standard hierarchy we can see that they both
have one member with a summed distance of"4" and four members with
"7." The wheel and the hierarchy are the same structure.
Position centrality
The third index
is position centrality,
which rates how central or peripheral a group member's position is. It is based
on a comparison of summed distances of group members. In a wheel group, for
example, one member has a lower summed distance, "4," than the other
members, who have "7" each. The member with the lower summed distance
is central. This person is in the middle of message flow. The others are
peripheral. They are on the outside of the flow. Another way to think of it is
that a group has "levels" of centrality. The chain example has three
levels of centrality. Such a group has summed distances of "6,"
"7," and "10." The lowest number, "6," is the
most central. The "Y" group has four levels of centrality. In the
case of the circle and the comcon, all members have the same summed distance.
In these groups, position centrality is equal.
We can compare groups
according to whether members' position centrality varies. In centralized groups, such as the wheel,
"Y," and chain, some members are more central than others.
In diffused groups, members' centrality
does not vary. Groups such as the circle and the comcon fit this model. No one
member is more central than another. The communication links are diffused
evenly. We can make additional variations regarding how central or diffused a
group is, such as degrees of centrality. This idea applies to the concept of
network centrality.
Network centrality
The fourth index is network centrality. Scientists have
proposed mathematic methods for determining this index. We will develop an
intuitively simple way to create this index. We add all the members' position
centrality scores, which equal their summed distance numbers. We now have a
number that represents the total amount of "distance" involved if
each member were to send a message to each other member; in other words, the
number of messages the whole group would have to use for all the members to
communicate once. In Figure 9.1, this number is next to the name of the
structure. For example, the wheel has a total of "32."
In diffused groups, each
member's "contribution" to this total is equal. For a five-person
group, this contribution is 20 percent. For example, in a circle configuration
each member has a summed distance of "6" out of a total of
"30." "Six" is 20 percent of "30."
By contrast, members of
centralized groups need fewer links to send messages to each other member. They
have a lower summed distance than others. These members, therefore, contribute
a lower percentage to the total of the summed distances. By observation, we
would say that the wheel is the most centralized structure. Our index of
network centrality verifies this. We can see that the wheel has only two levels
of centrality. One member contributes only 12 1/2 percent to the total of
possible communication links. The four peripheral members contribute almost 22
percent each. The person with the smallest percentage of contribution is the
most central member. In a hierarchy, we would probably label this person the
most powerful. He or she needs to expend the least amount of energy and time to
communicate with all other group members.
The "Y"
structure has four levels of centrality; the chain has three levels. By
"level" we mean the numbers indicating each member's summed distance.
The amount of levels of centrality do not indicate how central a group is,
however. For example, between the "Y" and the chain, the chain is
more centralized. The "Y" configuration has a central member that
contributes a little less than 14 percent of the group's total. Its next most
central member contributes 16 2/3 percent. The chain, in contrast, has a
central member who contributes 15 percent and is flanked by two people who each
constitute 17 1/2 percent. This means that the chain has three members who have
low percentages in comparison with the others; the "Y" has only two
such clearly central members. We should remember that these percentages all
apply to five-member groups only and would change for groups of other sizes.
Further considerations
We have
two more considerations before we examine the research regarding imposed
networks.
First,
the five configurations we described in Figure 8.1 are not the only ones
possible. One can join the two peripheral members of a "Y," thus
making a "kite" network. Or, we could join some of the members of the
circle and form a network that is intermediate between the circle and the
comcon. It is also possible to restrict some or all of the circle or comcon
channels so that they are one-way only. This idea works only with the circle
and the comcon, however. The other configurations could not restrict channels
because then members could not send messages to all other members.
Second, we have not
considered group size. The larger the group, the more networks are possible. In
a three-person group, only two networks are possible if we don't use one-way
channels. For three people, the "Y" disappears, the chain and wheel
are identical, and the circle and comcon are also the same.
The diagrams and
considerations that we have explained so far may appear abstract; however, they
have concrete implications. Scientists use the ideas about distance and other
aspects of the structures as bases for their research into group functioning.
Research suggests that differences in position centrality lead to differences
in process and output variables. Similarly, among entire groups, network
centrality also affects group process and output.
In the following section
we will discuss research and the experimental findings concerning imposed group
networks. We will begin with a description of how the experiments were
performed.
Research
Methodology
Theoretical work by
Bavelas (1950) suggested the potential value of research on imposed group
networks. Shortly thereafter, scientists devised a method for controlling the
message linkages among group members. In this way, research on imposed networks
could begin.
Consider a five-person
group. In an experiment, each member is seated at one of five adjacent
cubicles. Walls with slots separate the cubicles; written messages can be sent
through the slots. An experimenter can impose any of the possible group
networks on the group's message exchange by varying which slots are open and
which are closed. Keeping all the slots open, for example, produces a comcon
network. If the researcher allows message exchange between only adjacent
members, a circle network results. Some studies have used messengers or
intercoms between group members in different rooms rather than slots between
cubicles. Any of these experimental situations creates an environment in which
groups work under the constraints of various structures. In this way, theorists
can compare the effects of different structures.
Early studies on imposed
networks employed a problem-solving task requiring input from all group
members. The experimenter gave each member of a five-person group a card. On
the card were five of the following six symbols:
The symbols were
combined on the cards so that each member's card had a different pattern. One
symbol, however, appeared on all five cards. The goal of the task was to
discover which single symbol the cards all shared. It became known as the
"common symbol" task. Once the symbol was discovered, a further goal
was to inform each group member of the correct answer.
The research design
placed each group member in a cubicle with a box that had six switches. Each
switch corresponded to a symbol. A group member that either discovered the
common symbol or was informed what it was could flip the switch corresponding
to that symbol. The experimenter sat nearby in a booth that had a bank of
lights connected to the group members' switches. When the group member flipped
a switch, a light showed the researcher which symbol the member had chosen.
This setup allowed the members to quickly reveal when they had finished their
work. The group completed the task when each member had switched on a light.
The researcher could thus note both the speed and the accuracy of the group's
performance.
Experimental Results
Leavitt's (1951)
experiment was the best known among these early experiments. Five-person
circles, wheels, "Y" groups, and chains performed 15 trials of the
common symbol task.
Network centrality
The effect of
network centrality on the groups was fairly consistent. It affected speed,
accuracy, and group process. In general, the more centralized the group, the
better it performed. The wheel, the most centralized configuration, was
slightly faster than the "Y" groups. The wheel averaged 32.0 seconds
and the Y group averaged 35.4 seconds. The circle, at 50.4 seconds, and the
chain, at 53.2 seconds, were far slower.
Group process followed a
similar pattern. Wheel group members sent the fewest messages, with the
"Y" and the chain coming next, and circle groups sending the most
messages. The wheel and the "Y" networks also made fewer errors than
the circle and the chain.
Position centrality
Position centrality also
led to consistent findings among the groups. The more centralized the group,
the more often and the more unanimously the members judged that the group had a
leader. In the chain, the most centralized member received the majority of
leadership "votes." The two adjacent members, who are more central
than peripheral, also obtained votes. With the "Y" configuration the
two most central members split the votes. The choice of the central member of
the wheel was almost unanimous. Also, the most central members sent the most
messages.
Maintenance output
The different group
networks also affected maintenance output during the experiments. Overall,
members of centralized groups were more satisfied with their groups'
performances than were members of less centralized configurations. Peripheral
group members of centralized networks, however, were not so satisfied with
their personal jobs in the group. This situation might mirror, for example, the
situation of an army private, Paula. She is satisfied with the way the army is
winning a war but not so pleased with the job that she has to do to help the
army complete its task.
Diffused group members
had a relatively high personal satisfaction level even though diffused group
members were not as satisfied with their group's performance as centralized
network members were. For example, Paula may not be satisfied with the
efficiency of a circle of her friends when they get together and plan a party,
but she is happy with the equal part she plays in the planning process.
Of all the groups,
however, central members of centralized networks liked their jobs the most. For
example, the general in Paula's army is probably happier on the job than the
soldiers. In the experiments, however, a centralized group's overall
satisfaction was lowered because peripheral members expressed a strong dislike
for their jobs. Their unhappiness offset the central member's satisfaction. The
differences in job satisfaction between central and peripheral members
increased as time went on during the game trials. For example, the general in a
centralized army would become happier over time, and the privates would grow
less content. Also, differences in the levels of satisfaction increased as a
group became more centralized. This might happen if a group went from a chain,
through a "Y," to a wheel.
Explanation of results
We can account
for the consistencies in these experimental findings by considering two
principles we have already discussed: (1) the demands of the task and (2) the
opportunity for group members to communicate. These two principles are
interrelated. For example, an easy task may not require as much communication
as a more difficult task. Both principles affect group process and output variables,
such as performance and level of satisfaction. As we discuss the experiments
and that behavior patterns are consistent with what an experimenter might
expect, we will keep in mind both these principles.
Task demands
The "common
symbol" problem is an accuracy task. As discussed in Chapter 2, this means
that the group's task performance is based on the performance of its most
competent member. In other words, when one person solves the problem, the whole
group is successful. Now, the task itself is simple. Anyone with sufficient
information about the symbols on all the cards can perform it. It is not a
problem that would involve analyzing complex data or making a difficult
judgment. Thus, the most competent member is simply the member who can most
easily accumulate the information about all the cards.
In a centralized group
network this competent person is the most central member. He or she can best
learn which symbol is on everyone's card. Even a diffused group can work
efficiently on this task if it chooses to exchange messages via a centralized
organization. Comcons can easily act as wheels, and circles can easily act as
chains, if group members ignore some of their message exchange channels. This
choice is not natural, however, and it seldom occurs spontaneously. Therefore,
truly centralized groups are more likely to organize themselves in the manner
most efficient for solving the problem of finding the common symbol on the
cards.
Communication "organization"
Leavitt's
observations on how his groups exchanged messages support this last claim. He
examined how organized his groups could become through their communication. By
"organized" we mean that they began to send messages in a consistent
pattern from trial to trial. The centralized wheel groups generally became
organized by the fourth of 15 trials. They also generally adopted the efficient
method of sending information to the central member. This member found the
common symbol and informed the other group members.
The "Y" group
members also generally organized themselves efficiently, although it took a
little longer for them. Chain group members usually, but not always, became
organized. Their organization often placed major responsibility on one of the
members adjacent to the central member. This pattern is sufficient but not
optimal for task performance. It may be at least partly responsible for the
chains' relatively poorer results.
Leavitt did not find any
consistent organization to develop in circle groups, the least centralized
configuration. Later researchers decided that such diffused groups did have
something of an organization. They saw a pattern naturally developing, and
called it "each-to-all." In this "organization" each member
sent information to all others. Such a pattern is definitely not efficient for
solving the common symbol problem.
Organization and efficiency
Let us study the
issue of efficiency more closely. A well-organized wheel, with five members,
requires eight messages to complete its task. Each peripheral member will send
one message to the central member and the central person will send four
messages back to the peripheral members. For example, after getting their
cards, the four members in cubicles encircling the central member will write
notes telling which symbols they have and put the notes through the connecting
slots. This makes four communication "moves." The central member then
looks at the cards finds the common symbol, and quickly tells each other member
the answer, in turn, through the slots. These four additional notes total eight
messages.
Centrally organized
"Y" groups and chains also require eight messages; however, the need
for using intermediaries for messages slows the groups slightly. Comcons that
are organized in "each-to-all" patterns require each member to
receive information from each other member. The members then must make
individual decisions on the common symbol. This requires 4 messages from each
member, or 20 messages in all. Such organization leads to slower work than
centralized configurations. The opportunity for error is also greater.
We can use these
considerations to examine group process and task output. As we have noted, some
tasks, such as the common symbol problem, are best performed by a centralized
group pattern. Wheel, chain, and "Y" groups have an imposed network
that predisposes them to adopt centralized patterns of message exchange. These
groups will become centralized more often and more easily than groups that do
not have such an imposed structure. They will then perform tasks such as
finding the common symbol more quickly and accurately than diffused groups.
In addition, an
organization that emphasizes its central member will need to send fewer
messages to complete the task than, for example, a group using an "each-to-all"
pattern. The central member will send the greatest proportion of messages. He
or she is also the sole problem-solver, and other members will judge such a
central person to be the leader. In the "Y" group and chain
configurations, one or more central member may share in or perhaps take over
the problem-solving responsibility. In that case, the other central members
will also receive some recognition for leadership.
As we can see, the
demands of the task bear directly on which group organizations function best.
Maintenance output and task
We need to
consider the impact of the task on organization again as we examine Leavitt's
results for maintenance output.
In a centralized
organization, the peripheral members have little opportunity to exchange messages
in comparison to the central members. They also have fewer opportunities than
members in an "each-to-all" group. With the "common symbol"
task, however, centralized groups are more successful than others. As we would
expect, peripheral members in a centralized configuration are more satisfied
with their group's performance than less centralized groups because the latter
do not perform as well. Also as we would expect, peripheral members in a
centralized organization are dissatisfied with their own role in the group.
They have little opportunity to communicate and are not directly involved in
solving the problem. They are the "soldiers" in an army hierarchy.
In contrast, central
members of centralized configurations and all members of diffused groups are
more personally satisfied. We would expect this because these members have a
greater opportunity to communicate. Central members, however, were even more
satisfied than diffused group participants, in part because of their greater
personal responsibility for the group's performance. In addition, central
people may be more content because their group has a greater overall
performance.
Organization and group performance
The manner in
which a group organizes itself appears crucial in determining the group's
performance. Several implications follow.
First, the communication
"organization" of groups improves as the group members become more
experienced at problem solving. For a problem such as the "common
symbol" task, groups develop over time so that they need to exchange fewer
messages. Thus, they can solve the problem faster. They also become more
satisfied with the group as the group does its job well. Further, there is
evidence that performance differences among imposed networks disappears a large
number of trials. Burgess (1969) asked four-member wheels and circles to
perform the "common-symbol" task for 10 hours. At the beginning,
circle groups were slowest, as in the Leavitt study. However, as time passed,
circle groups improved their performance more quickly than wheel groups. By the
tenth hour, the groups had solved the problem more than 500 times, and there
was no difference in speed between circles and wheels. This finding suggests
that any group can adopt a successful organization for its task. All the group
needs is experience.
Second, some evidence
indicates that groups can plan an efficient structure for themselves.
Researchers have shown that message exchange may be more efficient if a group
is given an opportunity to plan an organizational pattern before performing its
task. Such planning results in a relatively faster performance by groups with
diffused networks.
Third, groups try to
maintain their imposed networks if those structures are effective. If a group
has developed an efficient organization under the constraints of one imposed
network, the group will maintain that organization as much as it can under a
different imposed network. For example, an efficient centralized group can
maintain its organization if a comcon setup is imposed. Such a group will
actually tend to keep its original organization. In contrast, the change from a
comcon group to a centralized configuration may require some adaptation.
Further, the maintenance effects of this last change are predictable. Central
members have job satisfaction that rises, but at the expense of members who are
suddenly peripheral.
Additional Research
We have been
examining experiments using the "common symbol" problem. Other tasks
place different demands on a group.
As we have discussed,
task demands determine optimal communication "organization."
Organization, in turn, affects all phases of group functioning, such as the
level of satisfaction and performance. Consequently, tasks other than the
"common symbol" may entail different organizational patterns and
results for group functioning. For example, structures that were efficient for
the "common symbol" problem may not be efficient for other tasks.
Shaw (1954) performed a
study in which four-person groups solved simple arithmetic problems. As did
Leavitt, he imposed different structures on the groups. An example of the type
of mathematic problem that he used is:
How many trucks are
needed to move a company's office equipment if it owns
(a) 12 desks, (b) 48
chairs, (c) 12 typewriters, and (d) 15 filing cabinets and if
one truck can carry
either (w) 12 typewriters, (x) 3 desks, (y) 5 filing cabinets, or (z) 24 chairs?
Shaw gave each group member a general description of the problem; however, the
participant received only two pieces of the information necessary to solve the
problem. For example, the group member would see one item out of "a"
to "d" and one item out of "w" to "z." Further,
the two pieces of information that each participant received were never about
the same items of office equipment. Thus, if someone was told how many desks
the company owned, he or she would never be told how many desks a truck could
move. The latter piece of information would go to another member of the group.
The results of Shaw's
research were generally the reverse of the Leavitt study. The circle was
fastest and best at correcting errors. Group members in this network saw
themselves as the most cooperative and the best performers. They also enjoyed
the task the most. In contrast, the wheel ranked last in all these variables.
To account for the
differences between his groups and Leavitt's, Shaw emphasized the different
tasks. The demands of Shaw's arithmetic question were different from those in
Leavitt's "common symbol" problem.
As we have shown, Leavitt's
groups attempted to solve a problem requiring only that the members collate
information. For reasons we have discussed, this type of task is best performed
in a centralized group. In such a group the central member can assume primary
responsibility to collate the necessary information efficiently.
In contrast, Shaw's
groups performed a task with two requirements. Group members not only had to
collate information but also needed to perform arithmetic operations once they
had the information at hand. Shaw reasoned that this experimental task placed
too much responsibility on the central member of a centralized group to handle
the problem efficiently. Instead, group members can most efficiently solve the
mathematic problem by splitting it into parts with different people, or
subgroups, working on each part. The work becomes, in effect, a multiple stage
accuracy task. As you recall from Chapter 2, multiple stage accuracy tasks are
best performed when group members divide the stages among them. Diffused
structures most easily allow this division. Thus, circle group members perform
this task better than wheels. In addition, circle group members enjoy
themselves more because their diffused group organization allows them greater
opportunity to communicate.
Other studies have been
performed in order to compare diffused and centralized structures performing
"common-symbol" and arithmetic tasks. The results of this research is
consistent with our conclusions. In reviewing this work, Burgess (1969) found
13 studies of "common-symbol" tasks. Centralized networks were faster
in seven of these, diffused networks were faster in three, and there were no
differences among networks in the other three. Burgess also found ten studies
of arithmetic problems. In this case, diffused networks were quicker in six,
and centralized networks quicker in only one, with no differences in the other
three.
There are other types of
accuracy tasks that diffused groups perform well. For example, Davis and
Hornseth (1967) had individuals and five-person groups in wheel, circle, and
comcon networks work on "eureka" problems that required creativity
and insight. In this study, the comcon sent the most messages and was both
fastest and most accurate.
In addition, the
experimenters rated the groups. They took the results of the individuals and
used the Lorge/Solomon Model A (see Chapter 2) to estimate, without losses from
faulty process, the performance of the five-person groups. None of the groups
reached this estimate, but this usually occurs in disjunctive tasks. The comcon
groups, however, came closest to matching the estimates. The comcon appears
flexible enough to allow the group's most competent member to take
responsibility for solving the problem. The problem-solving member will then successfully
influence the rest of the group more often than not. Centralized networks, on
the other hand, lack this flexibility. In those groups, central members tend to
become most influential, regardless of whether they are the most competent at
the task.
These two studies
examined tasks that needed to create or manipulate information. We can conclude
that when a group is faced with such a task, in any fashion, the diffused
structure leads to better results than the centralized structure. Diffused
networks produce better task and maintenance group output in such cases.
Centralized networks, such as the wheel, appear to be of value only when their
function is to relay information quickly and efficiently. We can see networks
such as the wheel at work in the strict hierarchies officially imposed in the
military, for example. Based on the research above, we can infer that the
military's structure would lead the army to be better at transmitting data, for
example, than it would be at inventing a new tool. The army eventually could do
both tasks, but research suggests that it would take the army longer to invent
a tool, and perhaps with less satisfactory results, than a more diffused
network, such as a circle of inventors.
As always, however, we
must be careful not to overgeneralize. For example, conflict might make a
diffused network less effective. Comcons have a large amount of message
exchange, which increases the risk of conflict. In contrast, centralized
networks impose restraints that lower the odds that conflict will occur. If a
group wants to avoid conflict, a more open communication network may be too
risky. The benefits from the restraint of a centralized structure may outweigh
the advantages of a diffused network. (Recall the advantage of using intermediaries
in conflict negotiations. Intermediaries are analogous to central members of a
centralized group.)
In another example, a
group of people who have never met may be too uncomfortable to participate
freely in the give-and-take of a diffused organization. If the atmosphere is
tense, the group will have trouble accomplishing its task. To solve the
problem, group members could work through an intermediary, or central member,
who could make the group members feel more comfortable.
General Conclusion
The main
conclusion of these studies is that the group's goal is very important. The
choice concerning whether to restrict the structure of who-can-speak-to-whom
depends on the group's goal.
As we have seen, studies
have examined the effects of restrictions on who-can-speak-to-whom in groups.
We call such restrictions "imposed networks." We can find examples of
imposed networks throughout society, ranging from an official company
brainstorming circle that encourages all members to speak, to a military hierarchy
that places power in the hands of a few. Various networks affect group process
and output in different ways; however, the results always depend on the group's
task. For example, if a task mainly involves relaying information, the
centralized group is effective. Group members in such groups are satisfied with
their group's performance and complete the task well. By contrast, a task that
involves creative problem solving acts differently on a group. A diffused group
network generally leads to the best results with this kind of task.
MEDIATED GROUP
DISCUSSION
Now we shall examine
group discussions that members conduct through various forms of communication
media. During the 1970s, researchers performed many studies that compared this
kind of discussion with face-to-face group discussion.
Types of Channels
We can classify these
forms of media according to the communication channels they include. The
channels are visual and audio. In other words, these media differ in terms of
whether they let people see or hear one another. The four forms of mediated
group communication are:
1. Audio and visual channels. This form
combines the two channels. It includes discussions in which group members who
are in different rooms can both see and hear one another. Members usually use
close-circuit television to conduct this form of discussion.
2. Visual channel. In this case, group
members are in different rooms that are separated by glass partitions. Members
can see but not hear one another. They can use only written communication--handwritten,
typed, or teletyped messages. Researchers have even used arcane "remote
handwriting" systems in which a person writes with a pen in one room while
a mechanical pen duplicates their hand movements in another room.
3. Audio channel. Here, group members in
different rooms cannot see one another. However, they can speak with one
another through telephone or microphone and speaker systems.
4. Neither channel. In this case, group
members are in different rooms and can neither see nor hear one another. To
communicate, they can send written, typed, or teletyped messages.
"Social
Presence" Theory of Short et al.
According to Short,
Williams, and Christie (1976), each form of media should affect group process
and output differently, because the fewer channels of communication available,
the less information people can transmit. Let us examine their argument. To do
so, we must first distinguish between verbal
and nonverbal behavior.
Verbal behavior consists of words and sentences. Nonverbal behavior consists of
body movements, body positions, and facial expressions. Nonverbal behavior also
includes what are called "paralinguistic cues," vocal characteristics
that accompany speaking, such as the rate and pitch of speech.
People give meanings to
nonverbal behavior. When they look at nonverbal behaviors, they use what they
see to form judgments about how people relate to one another. Researchers have
done many studies to try to discover how people interpret nonverbal
interaction. The studies have investigated the sorts of judgments people make
when they see different amounts and types of nonverbal behaviors.
Mehrabian (1973)
reviewed these studies. Mehrabian concluded that scientists can classify
nonverbal behaviors according to three dimensions: immediacy, relaxation, and
responsiveness. Each dimension corresponds to a type of judgment. These
judgments concern how people relate to one another and are based on the
nonverbal behaviors that people display. Different behaviors help people judge
different aspects of a relationship. In essence, a researcher can look at a
nonverbal behavior and ask, "What type of judgment would people make about
this relationship based on what they have just seen?" For Mehrabian, the
answer would come from one of his three dimensions.
The first dimension
consists of behaviors that signal "immediacy," or liking. How much do
people enjoy being together? This type of judgment concerns the degree to which
people like each other as they interact. Many nonverbal behaviors indicate
immediacy. They include variations of the following: touching, interpersonal
distance, eye gaze, facial expressions (such as smiling and frowning), head
movements (such as nodding), forward lean, and body orientation toward one's
interpersonal partner.
Mehrabian's second
dimension includes actions that signal "relaxation." These behaviors
reveal the relative power that interacting people have. In Chapter 5 we
discussed the kinds of behaviors that show the levels of power that exist
between communicators. Of these, Mehrabian particularly stressed arm and leg
positions and sideways lean.
The third dimension
involves actions that show a person's "responsiveness," or attention,
to others in an interaction situation. Such actions signal how much attention a
person is paying to his or her interacting. Nonverbal behaviors that indicate
responsiveness include paralinguistic cues such as a person's speech rate and
the amount of variation in speech volume and pitch.
Thus, we use nonverbal
behaviors as clues. They help us judge how much people like one another, how
much power they have relative to one another, and what attentiveness level they
have as they talk.
Effects of Media on Nonverbal Interaction
All of this
changes, however, when people conduct their discussions through media rather
than face-to-face. When people don't talk face-to-face, nonverbal behaviors
cannot play the interactive functions we delineated above.
For instance, if people
use a medium that does not include an audio channel, they cannot hear the
paralinguistic cues that help them judge attentiveness. Similarly, when the
medium does not include a visual channel, people lose all the body movement and
position cues that help them judge levels of liking and power among people.
Even closed-circuit television may remove the most important indicators of
power if the picture includes only people's faces and upper bodies, rather than
their entire bodies.
"Social Presence"
Short, Williams,
and Christie (1976) examined what happens when group members communicate
through restricted channels rather than face-to-face. They concluded that,
overall, such restrictions lower "social presence" in groups.
"Social presence" is the feeling among group members that they are
communicating with people instead of with impersonal objects and that these
people have unique personalities and real emotions. When social presence is
high, each group member has a feeling of "joint involvement" with the
other members. This feeling is absent if social presence is low.
Effect on group maintenance
When social presence is
low in a group, the group fails to become cohesive and does not develop a
stable structure of roles for its members. The group thus finds it difficult to
perform maintenance functions.
Different media restrict
communication in different ways. The more severe the restrictions, the lower
social presence becomes. "Informationally poor" media, including
written messages, restrict communication rather severely. These media will
lower social presence and threaten group maintenance. "Informationally
rich" media, including closed-circuit television, on the other hand, place
relatively few restrictions on communication. These media heighten social
presence and preserve group maintenance.
If we were to call face-to-face
interaction a "medium," it would be the most informationally rich of
all. It best helps groups maintain themselves. We do not, however, label
face-to-face interaction as a medium. Hence, when we say "mediated
discussion," we mean that group members are not meeting face-to-face.
Effect on group task
One may think
that the absence of visual and audio channels would have little effect on how
well groups perform their tasks. After all, groups use verbal communication to
do their tasks. Even the most severe restrictions, which deny members the use
of audio and visual channels, should not impede them. As long as members can
send written or typed messages, they can successfully perform tasks. We saw
proof of this in our discussion of imposed communication networks. Group
members who had to work under severe restrictions solved common-symbol,
arithmetic, and eureka problems.
Perhaps we should not,
however, be quick to assume that groups can perform tasks even with severe
restrictions on communication. As we have seen, Short et al. hypothesized that
media that restrict communication lower social presence in groups. This, in
turn, lowers the cohesiveness level and creates group maintenance problems. In
Chapter 8 we discussed Bales's hypothesis of the "equilibrium
problem." He felt groups had to balance their maintenance and task needs
to be successful. If Bales was right, groups that have maintenance problems
will eventually have task problems. Therefore, groups may harm their ability to
do their tasks well if they use communication media for their discussions.
Let us keep in mind the
arguments behind social presence theory as we go on to look at research into
how media affect groups.
Research into the
Effects of Media on Task Performance
Is it true that
restrictive media can harm task performance? Scientists have performed many
studies to try to learn more about this issue. Research findings showed that
the answer depends on the type of task.
Tasks with Objectively Correct Answers
Some group tasks
require an objectively correct answer. Researchers can judge the accuracy of
the outcome. Studies have examined these kinds of tasks to see how they respond
to communication restrictions. For example, Chapanis, Ochsman, Parrish, and
Weeks (1972) formed two-person groups and controlled the communication
channels--face-to-face, audio only, teletype, or written notes. The groups had
two tasks: (1) to find the office of a physician closest to a certain home
address and (2) to assemble a trash can carrier. Thus, both tasks had
objectively correct answers that the researchers could measure.
The study found that
participants who could only write or type messages to each other took longer to
perform the tasks than the participants who could speak with each other. Most
people take five to ten times longer to type or write a message than to say it.
This time difference, however, did not affect the quality of the answers. In
the end, the media by which groups could communicate did not affect how well
they succeeded at their tasks. Other similar studies have found the same
results. When tasks require objectively correct answers, the medium a group
uses has no effect on the correctness of its answers. Groups can succeed at
these tasks even under heavy communication restrictions.
Tasks Without Objectively Correct Answers
What about other
types of tasks? Conflict and negotiation tasks, for instance, have no
objectively correct answers. Studies of these kinds of tasks lead to different
conclusions about the effects of communication media.
Wichman (1970) conducted
one of these studies in which participants played 78 trials of a Prisoner's
Dilemma Game. The communication channels they could use were (1) neither audio
nor visual, (2) audio only, (3) visual only, or (4) both audio and visual.
Wichman measured how often participants cooperated with one another in each
condition. The percentage of trials in which cooperation occurred in each of
the four conditions is shown in Table 9.1.
|
Table
9.1 |
|
Visual |
|
|
|
|
Yes |
No |
|
Audio |
Yes |
87% |
72% |
|
|
No |
48% |
41% |
As you can see,
cooperation was much higher when participants could hear one another than when
they could not. In contrast, cooperation was only slightly higher when
participants could see one another than when they could not. Therefore, vision
appeared to affect cooperation only a little bit; the effect of hearing was
very strong.
Morley and Stephenson
(1969, 1970) did another study of tasks that do not have objectively correct
answers. They asked two participants to role-play a session in which two sides,
union and management, were negotiating about wages. Each participant took one
or the other side. They negotiated either face-to-face or by telephone. The
participants had information about the "facts" of the situation,
which the researchers had slanted so that one or the other side had a more
reasonable case. Morley and Stephenson found that the side with the better case
was successful more often when the negotiation was over the telephone rather
than face-to-face. Apparently, on the telephone the negotiations focused more
on "facts" than on the needs of the negotiators.
Interpretations of Study Results
Let us review
the studies that we have just discussed. One group of research involved how
media affect accuracy tasks. The other examined the effects of media on
conflict and negotiation tasks. Let us compare the two sets of findings, while
keeping in mind social presence theory.
The social presence
theorist would say that the study results are understandable. Media affected
task performance according to whether the tasks required a high level of social
presence because media affect social presence.
For instance, the
studies found that media had little effect on how well groups completed tasks
that have objectively correct answers. A social presence theorist would say
that this is true because groups do not need a high level of social presence to
do accuracy tasks well. When a group works on such tasks, it does not need to
perform the sorts of maintenance interactions that communication media can
affect. Therefore, mediated discussion does not affect the accuracy of groups.
In contrast, research
found that media did affect tasks that do not have objectively correct answers.
The social presence theorist would say that this happens because the level of
social presence in groups affects how they perform these tasks. The more
"informationally poor" the media in use is, the more impersonally
people treat one another. When they behave impersonally, they are less
cooperative in Prisoner's Dilemma games, and they more often use the
"facts" of a case to decide negotiations.
Further, even the verbal
content of negotiations seems to be more impersonal when people use
informationally poor media. Stephenson, Ayling, and Rutter (1976) gathered
participants who had different views on management-labor relationships. They
then asked them to discuss union-management negotiation problems. The
discussions lasted 15 minutes and were either face-to-face or audio only. The
researchers found that audio-only discussions were more task-oriented and less
maintenance-oriented than face-to-face discussions.
Research into Other
Effects of Media
Emergent Structure
Another of
Bales's beliefs was that decision-making tasks require groups to form
structures of roles for their group members. How do groups form role
structures? As we will discuss more fully in Chapter 11, some members will tend
to dominate their group's discussion. As a consequence, they take on leadership
roles. The other members will be quieter and will take on different roles. In
essence, groups establish role relationships by a definite process. The process
requires members to watch one another's behavior to determine the role that
each person should have in the group.
Strickland, Guild,
Barefoot, and
When some members
clearly talk more than others, as in the face-to-face discussions, leaders
emerge more easily. This, in turn, makes it easier for the group to establish a
role structure. Bales believed that such a structure helps groups make
decisions. Hence, he would have predicted that the face-to-face groups in the
study would have performed the task more easily than the other groups.
Strickland et al., however, did not study which media made it easiest for
groups to make decisions.
Evaluations
Communication media seem
to affect the evaluations that people make about each other. Short, Williams,
and Christie (1976) discussed the results of an unpublished study by LaPlante.
In that study, a participant and a confederate played a 20-trial Prisoner's Dilemma
game. Three times during the game the confederate sent either friendly or
unfriendly messages to the participant. The confederate gave these messages
through different media: face-to-face, closed-circuit television, telephone, or
written. After the game, participants rated how they liked the confederate.
Table 9.2 shows the
averages of the rates of liking when the confederate sent friendly or
unfriendly messages in all conditions. The positive numbers show liking, and
the negative ones indicate dislike. In general, the richer the communication
media, the more extreme the feelings of liking and disliking. As the media
became informationally poorer, indifference of feeling was greater. In short,
the participants made less extreme evaluations when they got less information.
This finding supports the social presence idea that informationally poor media
restrict the kinds of information that people need to judge whether they like
someone.
|
Table
9.2 |
|
Communication
Medium |
|
|
|
Message
Content |
Face-to-Face |
TV |
Phone |
Written |
|
Friendly |
1.02 |
.80 |
.46 |
.34 |
|
Unfriendly |
-.93 |
-.96 |
-.70 |
-.03 |
"Inherent" Qualities in Media
One further
aspect of social presence theory is that media have inherent differences that
affect group process. In other words, a certain medium will, by its nature,
affect group process in a certain way.
When we look at
informationally poor media, we can see some support for this idea. Clearly, the
nature of these media causes them to restrict maintenance-oriented
communication in groups. Further, evidence indicates that people perceive the
differences between informationally rich and poor media. In essence, people may
see inherent differences that affect groups.
Research by Champness
(1973) revealed this evidence. In that study, dyads made decisions about choice
dilemmas by using various channels of communication: face-to-face, audio only,
or closed-circuit television. At the end of the study, the dyads made judgments
about the communication channels they had used. Ryan (1976) made a similar
study. In it, groups made decisions about a personnel problem using the same
media conditions as Champness's study. The groups also reviewed the
communication channels they had used and made judgments.
We can combine the
results of both studies. Altogether, the participants judged face-to-face
interaction as slightly more positive and aesthetically pleasing than
closed-circuit television. With the audio-only medium, however, they were more
critical on these issues. They felt that both face-to-face and closed-circuit television
were far more positive and aesthetically pleasing than the audio-only medium.
They judged that face-to-face interaction was no more "public" than
the audio-only interaction, and they felt that both face-to-face and audio-only
channels were less "public" than closed-circuit television. These
findings clearly suggest that people prefer meeting face-to-face more than
using media for their discussions.
Conclusions
The conclusion of the
social presence theorist, based on all we have discussed, is that
informationally poor communication media do affect group process. They restrict
the information that groups need to maintain themselves and, in contrast,
encourage task-oriented information. In this situation, liking and power
relationships are less likely to form.
When tasks have
objectively correct answers, group maintenance does not matter. Thus,
communication media do not affect the accuracy of groups that do the kinds of
tasks that objectively correct answers. Group maintenance does matter, however,
in tasks without objectively correct answers, and communication media can
affect how groups perform them.
We will discuss these
ideas again in the following section on computer conferencing.
COMPUTER
CONFERENCING
During the last decade,
organizations have come to rely more and more on computers for many tasks,
including conferencing. Computer conferencing is becoming more popular every
year.
Traits of Computer
Conferencing
Organizations can
connect computers through telephone lines, which allows computers to transmit
information quickly. Through these computer connections, people separated by
hundreds or even thousands of miles can form groups in order to make decisions.
Further, a computer system can store a great deal of information, which users
can later retrieve and read at their leisure.
In face-to-face
discussions and in mediated forms of discussion, members must take turns
sending messages. Computer users, however, can type messages to each other
simultaneously. The computer can store all these messages until group members
are ready to read them. Earlier, we discussed how certain forms of mediated
group discussion generally take longer than face-to-face discussions because
people write and type much slower than they speak. Sending messages becomes
much faster, however, if group members can type at the same time. In theory, a
group of about seven or eight people ought to be able to make a decision as
quickly when they are typing simultaneously as when they take turns speaking.
Types of Computer
Conferencing
Different types of
computer conferencing can be classified according to whether or not the group
discussion occurs at the same place and the same time. Let us start with the
circumstance in which a small group meets in a specially-equipped "decision
room." Each member of the group has a keyboard. At the front of the room
is a large public screen that displays what the group is working on at any one
time. Special "jointwork" word processing programs allow each member
to make changes in the display. In a complex decision room, members can work on
either the central public screen or individual computers at their seat. In this
case, people can send public messages to everybody or they can send private
notes to specific group members. Members can easily combine face-to-face
discussion with computer conferencing in a decision room.
The "decision
network" type of computer conferencing links individuals and subgroups
that are in different locations. One way a decision network can be used is to
allow people to work together at the same time, rapidly exchanging messages via
electronic mail. The members may be in the same building, in different
buildings in the same complex, or many miles apart. This is called
"synchronous" message exchange. A synchronous decision network is
similar in many ways to circumstances that we have already described in this
chapter, when people can interact by writing but cannot see one another. In
this case, most nonverbal cues are missing from their interaction. Examples we
have mentioned are the network study groups and some of the mediated discussion
groups.
Decision networks can
also be used when people are hundreds or even thousands of miles apart. For
example, a group might include members in
There are both strengths
and weaknesses to asynchronous message exchange when compared to synchronous.
On the positive side, asynchronous exchange allows for people far away from one
another to work together. It also provides people with the time to reflect on
their decision and send thoughtful messages to one another. On the negative
side, asynchronous message exchange will lead to a far slower decision process
than synchronous. It may also be more difficult for members to coordinate their
work when they are not exchanging messages directly. McLeod (1996) has hypothesized
the types of circumstances in which each is likely to be better than the other.
One may also ask why a
group that can meet face-to-face would want to use a "decision room."
There are in fact several potential advantages to having computers aid in group
decision making. First, the use of "jointware" can simplify any
necessary word processing. Second, using a computer easily allows for a record
of the discussion to be kept. Third, people can use computers to control the
actual process of group discussion. We will discuss this issue in Chapter 13,
"Formal Procedures for Group Decision Making."
Computer
Conferencing and Social Presence Theory
Researchers have begun
to explore how computer conferencing affects decision process and output.
Scientists have done much of this research on groups in which people are
interacting through a decision network. The group members are unable to see or
to hear one another.
Members are in a
situation similar to the ones we discussed in the last section, in which groups
used "informationally poor" media. What can we imply from this
similarity? Initially we can hypothesize that this form of computer
conferencing would be low in "social presence." We can follow that
idea with several hypotheses based on the studies we examined in the last
section.
One hypothesis is that
group discussion should be high in task-oriented content and low in
maintenance-oriented content. Another is that the amount of participation
should be relatively equal among members, which should lead the group to have
problems forming a stable role structure. All these factors should not affect
groups' performances when they work on problems that have objectively correct
answers. It may, however, lead to difficulties with other types of tasks.
Further, group members should be relatively less satisfied with decisions made
through computer.
Research into
Computer Conferencing
Are these hypotheses
correct? Does this type of computer conference so closely resemble groups that
use informationally poor media? Particularly during the 1990s, a great deal of
research has been performed to evaluate these and other ideas. However, the
answers to some of these questions is still unknown. For example, some
researchers have found accuracy and quality of decisions to be higher in
face-to-face groups, others have found accuracy and quality to be higher in
computerized groups, and still others have found no differences. As McLeod
(1996) discussed, different researchers have been evaluating task performance
using different tasks and different types of computer programs. Without more
consistency across studies, it may be impossible to definitively answer this
question..
In contrast, there are
other hypotheses in which research findings are more consistent. We will
describe some of them next. (See Benbaset and Lim, 1993, and McLeod, 1996, for
more detailed discussions.)
Participation
Many studies
have shown that the amount of participation in discussion tends to be more
equal among members of computerized groups than among members of face-to-face
groups. This finding is consistent with results for mediated discussion, as we
described earlier in the chapter. However, this conclusion presumes that groups
do not have preexisting
social structures. New groups have yet to develop a social structure, and so
participation will be fairly equal among members. "Standing" groups,
with a history of working together, are likely to have already developed a
"pecking order," and those members higher in status will probably
participate more. Consistent with this idea, Benbaset and Lim's (1993) review
of literature showed that equality of participation in computer conferences
were greater for new groups than for standing groups.
Even new groups can have
unequal member participation if their members are aware of preexisting status
differences among them. Weisband, Schneider, and Connolly (1995) formed groups
consisting of two M.B.A. students and one undergraduate. The groups were asked
to evaluate, either face-to-face or by computer, the conduct of a computer
professional in an ethical dilemma. The members knew one another's status in
both conditions. The results showed the M.B.A. students to communicate more
than the undergraduate, whether via computer or in person.
Influence
If participation
is more equal among members in computerized groups than in face-to-face groups,
it follows that influence among members should also be more equal on computer.
Smith and
As with participation,
however, differences in influence can easily in introduced into computerized
groups. In the study just described, Weisband et al. also found the M.B.A.
students to be more influential than the undergraduate both face-to-face and on
the computer.
Minority influence is also affected in
some interesting ways by computerization. McLeod, Baron, Marti, and Yoon (1997)
reasoned that minority opinions were more likely to be expressed during
anonymous computerized discussion, because members would be less concerned with
the negative responses of others. However, they also hypothesized that other
group members would respond more positively to minority opinions during
face-to-face discussion. This is because those offering the opinion would have
more social presence, so that other members would pay more attention to them. McLeod
et al. formed four-member face-to-face, anonymous computerized, and
non-anonymous computerized groups to evaluate three companies available for
acquisition by an imaginary investment company, based on information the
researchers supplied the participants. One member of each group was purposely
given information that differed from the other three group members. Consistent
with their thinking, the researchers found the anonymous computerized groups to
result in the most minority comments but also the most negative reactions to
those comments by other group members. The highest number of positive reactions
to minority comments and, probably as a consequence, the greatest individual
opinion change occurred in the face-to-face groups.
Consensus
In general,
computerized groups find it harder to reach a consensus than face-to-face
groups. This might be as a result of the more equal participation in
computerized groups. As we discussed earlier, equal participation inhibits a
group's ability to form a stable role structure. Without a clearly dominant
member or two directing the group, consensus is harder to reach.
Again, however, there
are complications. As discussed earlier, social presence theory implies that
the effects of computerized message exchange should be less for tasks with
objectively correct answers than for other types of tasks. Hiltz, Johnson, and
Turoff (1986) asked five-member groups to decide either a survival-game
accuracy problem or a human relations quality task. When working on the accuracy
task, consensus was almost as great on computer as it was face-to-face. When
working on the quality task, all eight of the face-to-face groups reached
consensus, whereas only one of the eight computerized groups did.
Task and Maintenance Activity
It is probably
the case that, at least for groups that are learning how to use computerized
media, discussion is more task-oriented and less maintenance-oriented. In the
Hiltz et al. study just mentioned, the researchers used Bales's coding scheme
to content analyze the messages that members sent one another. They found the
face-to-face groups to make about twice as many positive maintenance-oriented
messages as the computer groups did.
Another area in which
computer groups seem to behave differently than face-to-face groups is negative
maintenance behaviors. Social presence theory would hypothesize that both
positive and negative maintenance behaviors would be lower in computer groups
than in face-to-face meetings. Hiltz et al.'s findings for positive maintenance
behaviors are consistent with this prediction. It is clear, however, that
social presence theory is wrong about negative maintenance behaviors. People in
computerized groups are more negative than people in face-to-face
circumstances. A study that showed this came from Siegal, Dubrovsky, Kiesler,
and McGuire (1986). In the study, three-person groups worked on choice
dilemmas, either face-to-face or via computers. The researchers found that the
computer groups made many more negative statements than the face-to-face
groups. Their negative behaviors included swearing at and insulting one
another. This type of behavior has come to be called flaming.
It seems that flaming is
likely to occur when group members are anonymous, so that the person who
receives their flames has no way of knowing who sent them. A finding by Jessup,
Connolly, and Tansik (1990) supports this possibility. Their research again
compared anonymous computer conferencing with signed conferencing. They
discovered that, when compared with the other participants, members in the
anonymous groups made many more critical comments.
One must not, however,
overemphasize the importance of these findings. Even in computerized groups,
flaming comprises no more than two or three percent of the total group
interaction.
Changes Over Time
One reason that
computerized groups are different than face-to-face groups is that computerized
groups are often unfamiliar with their communication medium. This implies that
as group members become more experienced in using computers to interact, their
process and output should eventually become more equivalent to that of
face-to-face groups.
To test this idea,
Hollingshead, McGrath, and O'Connor (1993) placed participants in three-or
four-member groups that performed a variety of tasks across a 13-week semester,
one task per week. For the first two weeks, face-to-face groups performed
better than computerized groups, but for the next month there was no difference
between communication media. During the seventh week, groups traded media; the
previous face-to-face groups started using computers and the previous computer
groups started working face-to-face. Once again, the newly-computerized groups
performed worse for two weeks, after which there was no difference. Group
member satisfaction paralleled task performance; lower for beginning computer
groups, but no lower for experienced computer groups.
Another area in which
computer and face-to-face groups become more similar over time is in the
proportion of task and maintenance comments. In several studies and essays,
Walther has claimed that, even on computer, people have as great a desire to
exchange maintenance information as they do face-to-face. However, people are
unable to satisfy this desire at the beginning of their experience as members
of computerized groups. Instead, members of new computerized groups concentrate
on task information. These people often unfamiliar with computer conferencing.
Further, it takes more time to type than to speak. For both of these reasons,
members of new computer groups seem to feel that they need to concentrate their
effort on task work. However, if the group stays together long enough, the rate
of maintenance-relevant discussion will eventually increase to the level of
face-to-face interaction. Walther's research results are consistent with these
ideas, and he has found further support in reviews of other researchers' work
(see Walther, Anderson, & Park, 1994).
Emotional Content
of Electronic Communication
Walther's research has
important implications for the examination of the emotional content of
computerized communication. Social presence theory leads to the idea that
people will be less emotional when communicating on computer than when
interacting face-to-face. The implication is that something inherent in
computers causes this. They behave like an "informationally poor"
medium. Emotional messages are absent because that is the nature of the medium.
It may be true that
members of decision making groups tend to be more task- and less
maintenance-oriented when using computer conferencing. If this is the case,
however, it may be limited to the decision-making situation. When researchers
look at other circumstances in which people communicate by computer, the
findings change. For instance, when people use computer networks to exchange
personal "electronic mail," maintenance information is as great as
face-to-face. Rice and Love (1987) examined six weeks of electronic mail
messages that passed through a nationwide public system. Using Bales's coding
scheme to analyze these messages, they discovered that 28 percent of the
messages fit into Bales's categories as positive maintenance statements. This
percentage is as high as Bales found when he analyzed face-to-face
decision-making groups.
The implication of these
findings is that nothing inherent in computers leads to an absence of emotion
in computer mediated groups. Clearly, one can communicate emotions
electronically. In fact, people often go out of their way to add emotional
information to electronic mail. Nonverbal codes for emotional information have
become standardized among electronic mail users. For example, WRITING IN
CAPITAL LETTERS means that the computer user is angry. Often, people who
receive messages in capital letters will admonish the sender for yelling. Other
nonverbal codes in common usage include a colon, a dash, and a closed
parenthesis to symbolize happiness, as the smiling face below shows:
:-)
In short, when people
exchange personal messages, they will satisfy their maintenance needs even when
they have to use "informationally poor" media.
SUMMARY
Structures are often
imposed on groups that form within organizations. One type of imposed structure
occurs when who can communicate with whom in a group is restricted. We call
these structures "imposed group networks."
Scientists have found
that group members have different roles in the flow of communication. They may
be in the middle of the communication network, as central members, or they might be toward the outside, as peripheral participants.
Groups as a whole are
affected by how the communication flow is arranged. Groups with an unequal flow
of communication, in which only a few members have greatest control over
communication, are called centralized
groups. An army might be an example of such a network. The army gives only a
few people the right to talk to anyone, any time. In contrast, diffused groups allow their members
more equal positions. For example, a circle of friends talking together
probably would let all members talk equally.
Experiments have shown
that centralized groups perform better than diffused groups when they work on
tasks that require only the exchange of information among members. An example
might be a task that requires group members to seek a common symbol.
Centralized groups, however, do not do as well with tasks that also require the
manipulation of information (for example, working together on an arithmetic
problem). In these cases, diffused groups perform better.
In the maintenance
realm, centralized groups tend to have a clearer power structure than diffused
groups. The more central members act as leaders, and other members perceive
them that way. Also, central members of centralized groups are the most
satisfied of all group members in both networks. Diffused group members,
however, on the average, tend to be more satisfied with their individual jobs
than centralized group members.
Research has shown that
all these differences between centralized and diffused groups diminish over
many trials.
In a second type of
structure, group members exchange messages through communication media rather
than face-to-face when they make decisions. Different media place different
restrictions on the number of communication channels that group members can use
when they send messages. For instance, closed-circuit television allows members
both to see and to hear one another. Telephone systems, on the other hand,
allow them to hear only each other. Media that use written or typed messages
allow for circumstances in which group members can only see one another or in
which they can neither see nor hear one another. These restrictions take away
nonverbal information from group members. Nonverbal information can signal
members' degree of liking, power, and attentiveness.
Theorists have looked at
what happens to groups when media restrict their communication channels. They
have proposed that the overall effect is to lower "social presence."
Social presence is the feeling among group members that they are communicating
with people rather than with impersonal objects. The way that social presence
affects group process and output seems to depend on the type of task that a
group performs. For accuracy tasks, group maintenance is relatively
unimportant. For these kinds of problems a group's amount of social presence
apparently has little effect on group process and output. In other tasks,
however, group maintenance is important, such as in conflict and negotiation.
Social presence does seem to affect these types of tasks. When groups with low
social presence work on these tasks, participation becomes more equal,
discussion content becomes more task-oriented, and cooperation lowers, relative
to other groups.
More and more groups are
using computer conferencing to make decisions. When people in separate
locations make decisions through computers, they can neither see nor hear one
another. This implies that groups that use computer conferencing should have
low social presence. Many findings on this topic are consistent with social
presence theory. For example, the use of computer conferencing generally leads
to more equal participation and influence among group members and more
difficulty reaching group consensus. In addition, group discussion via the
computer conference is probably low in maintenance statements and high in task
focus. However, when people exchange personal messages by computer, their
statements are often as maintenance-oriented as face-to-face interaction would
be. This implies that social presence theory is wrong to presume that
electronic media inherently leads to task-oriented behavior. Further, when
groups experienced at computer conferencing, their performance becomes similar
to face-to-face groups.