Group Processes
Psychology 457/557
Observation in the Small Group Setting
Many problems of small group research are common to other research in which the human observer is the measuring instrument. This exercise will focus on two of these problems. The first concerns obtaining agreement among observers, while the second deals with utilizing reliable human observations to test hypotheses. We will use a small group setting to show one way of increasing reliability and also one way in which this kind of data can be used.
In addition to illustrating problems of method, the purpose of this session is to acquaint the student with what has become a major area of research in the behavioral sciences. Investigators considering substantive topics such as interpersonal relationships, attitude change, group structure, leadership, and power relations, have tested many of their hypotheses in small group situations.
There are a number of difficulties with observation in a field setting in which we are unfamiliar with the structure of the situation. Under such circumstances, disagreement among observers is to be expected. Two of the sources of this unreliability in field observation are: 1) lack of a common frame of reference for the observers and 2) the occurrence of distracting events which are extraneous to what is being observed.
One way to provide observers with a common viewpoint is to develop a category system. A system for categorizing human interaction is a set of rules for deciding what behaviors are relevant and in which category each item of behavior should be classified. Some of these systems are quite complex (e.g., SYMLOG, MICS); others are quite simple. In general, the complexity of a category scheme increases with the complexity of the situation being studies, so that field observation often requires a larger number of categories than does the observation of a small discussion group.
Accordingly, one of the major advantages of the small group setting is the control which the investigator can exercise over the behavior that is being studied. For example, the investigator may:
1) Limit the population from which the group members are drawn -- thus making the groups very homogeneous in age, gender, education, social class, etc.
2) Fix the size of the group.
3) Decide on the group's task.
4) Require the group members to remain in view of the observers.
5) Fix the length of time the group is in operation.
6) Limit external noises, interruption of the group by outsiders, and other distractions that might cause the group to deviate from its assigned task.
7) Use microphones and recording devices to ensure that all observers hear what is said.
8) Use one-way mirrors to lessen the effect of the observers on the group.
By using techniques like these, the investigator can greatly simplify the problems of observation and measurement. However, whenever investigators interfere with the operation of a group, they run the risk of altering the group's behavior in important ways and limiting the generalizability of the conclusions of the study. Laboratory studies can be used to study the central features of problems and to discover important relations which the extraneous features of real life situations might obscure. However, wise investigators combine both laboratory and field research in the quest for understanding human behavior.
Introduction
In small groups, communication patterns and status hierarchies usually develop quickly (e.g., Berger, Cohen, & Zelditch, 1972). Bales, Strodtbeck, Mills, & Roseborough (1951) suggest that the amount of communication is a good rough measure of this type of group structure. In particular, Bales et al (1951) have proposed that after a small discussion group has been in existence for a short while, the communication patterns will tend to stabilize and the amount of communication initiated by a group member will closely map the amount of communication directed to that person. This hypothesis is tested in this exercise.
Method
Communication can take many forms. Researchers typically classify communication as either verbal or nonverbal. In this exercise, only verbal communication will be studied.
To test Bales et al's (1951) hypothesis, one must obtain a measure of the amount of communication. In this exercise, the communication between the group members was divided into "speech acts" (see below). The number of speech acts initiated and received by a group member then provides a measure of the amount of communication from and to that group member.
A "speech act" is defined as the single continuous speech of one individual. A speech was considered continuous, regardless of pauses, if it was not interrupted by the speech of another individual. A single work or a single phrase was considered an act only if the word or phrase expressed a complete thought. Thus, "What?", "Why?", "Yes", and "No" are considered acts but "er...", "Mmmmm", and "We..." are not. Details of the scoring procedure are included in the appendix (see appendix).
Because of the small number of speech acts observed and the uncertain statistical distribution of these acts, ordinal statistics were used to analyze these data. The group members were ranked twice, once in order of number of acts initiated and once in order of number of acts received. If Bales et al's (1951) hypothesis is correct then:
When the participants in the small discussion groups are ranked by the total number of acts they initiate, they will also tend to have the same rank for the number of acts they receive.
To test this hypothesis, a small group of xx individuals was observed by two observers and the number of speech acts initiated and received were counted. The reliability of this coding system was evaluated by comparing the rankings given by the two observers. Then, their data were combined and used to test the hypothesis.
Results
Reliability. Before the hypothesis can be meaningfully evaluated, the reliability of the communication measure must be determined. The correlation between the rank orders given the group members by two observers were calculated (see appendix). This provided two measures of reliability, one for rank order of acts initiated and one for rank order of acts received.
The rank order correlation for acts initiated was xxx. For acts received the rank order correlation was xxx. This indicates that xxx.
Hypothesis Test. The data for the two observers According to the hypothesis, the participants in the small discussion groups should receive the same ranking for both acts initiated and acts received. In this study, the correlation between acts initiated and acts received was xxx. This indicates that xxx.
Discussion
In discussing this exercise, first summarize the results in English and state their implications for the hypothesis. Then consider the following questions:
1. In studying the amount of participation, what assumptions do we make by defining an act as a single speech?
2. On the basis of your experience in this and other exercises, what are some of the problems involved in using the human observer as a measuring instrument?
3. Were there any differences in reliability for acts initiated and acts received? If so, why do you think these differences occurred?
4. Are there any other conventions we might have adopted to improve inter-observer reliability?
5. Whether or not the hypothesis supported by your data, under what conditions would you expect this hypothesis to hold, and under what conditions would you expect it not to hold?
6. Was there a task leader in the group? Who was it? Do you have any guesses as to why this individual was the task leader?
7. Was there a socio-emotional leader in the group? Who was it? Do you have any guesses as to why this individual was the socio-emotional leader?
References
Bales, R.F. (1950). Interaction Process Analysis. Cambridge MA: Addison-Wesley.
Bales, R.F., Strodtbeck, F.L., Mills, T.M., & Roseborough, M.E. (1951). Channels of communication in small groups. American Sociological Review, 16, 461-468.
Bales, R. F. & Cohen, S. P. (1979). SYMLOG: A system for the multiple level observation of groups. New York, NY: Free Press.
Heyns, R.W. & Zander, A.F. (1953). Observation of group behavior. In Festinger, L. & Katz, D. (Eds.), Research Methods in the Behavioral Sciences, New York: Dryden, 381-416.
Appendix
A simple system for observing interaction in groups
I. What to observe
A. This decision depends largely on the problem one is studying. Considering the hypothesis of this exercise, and to simplify the observer's task, only verbal interaction will be considered. Thus gestures, incoherent mutterings and other forms of non-verbal behavior will not be considered. While some of this non-verbal behavior might be important for the group task, (for example a thumbs-down gesture), the difficulty of training observers precludes recording the more subtle forms of interaction.
II. Speech Acts
A. A "speech act" is defined as the single continuous speech of one individual. A speech is considered continuous, regardless of pauses, if it is not interrupted by the speech of another individual. A single work or a single phrase is considered an act only if the word or phrase expresses a complete thought. Thus, "What?", "Why?", "Yes", and "No" are considered acts but "er...", "Mmmmm", and "We..." are not.
B. To facilitate scoring, it is necessary to establish some conventions to cover ambiguous cases:
1. If a person starts to speak, but does not complete a single thought before being interrupted, he or she is not recorded as having acted.
2. If a person begins a thought, is interrupted, and then continues with the same thought, his or her speech is considered a single act. For example:
Person A: "What if ..."
Person B: "But..."
Person A: "...we did X?"
A is scored as having acted once and B is not scored.
3. Although pauses generally are not considered to break up single acts, an excessively long pause (greater than 15 seconds) may be scored as two acts. If during this pause the speaker or other group members engaged in some other activity, such as reading case material, then two acts should be scored.
III. Aspects of each act to be classified:
A. Who initiated the act. Each member of the group will have a number in front of him or her for coding convenience. Use the identification number of the group member in your scoring. Classifying each act according to its initiator will not present any serious problem if you use the following conventions:
1. An act is only initiated by an individual. The "group as a whole" cannot initiate. If two people are talking simultaneously and each completes a thought, score both acts. The order in which you score simultaneous acts in not important.
2. As soon as a speaker completes a thought, record him or her as initiating. Do not wait until he or she has finished the speech. To code the receiver, wait until the end of the speech. Because you must make a more difficult decision in determining who receives the act, you should make the simpler decision as soon as you can.
B. Who received the act. Any individual member of the group or the group as a whole can be the receiver of an act. Record the member's identification number, using "0" for the "group as a whole" next to the number identifying the initiator of the act.
Scoring the receiver of an act presents serious problems. The observer must infer the intent of the speaker. If the speaker mentions the receiver by name, then the inference is clear. Most of the time, however, the observer will have to use more subtle cues. Among these cues are: (1) the person at whom the speaker looks, (2) the content of the previous speech, and (3) physical gestures such as pointing. To help decide who is the receiver of an act, the observer should pay attention to the eye movements of the speaker.
Because the speaker may change the intended receiver in the middle of the speech, you should wait until near the end of a speech before recording its receiver. The following conventions will improve reliability:
1. If an initiator speaks first to one person and then shifts to another, score the person who was the object during more than half of the duration of the speech. If you are unable to make this decision, score the act to "the group as a whole."
2. Insofar as possible, try to score individual members as receivers rather than "the group as a whole." That is, err in the direction of too few scores to the "group."
3. If you miss an act, forget it. If you spend too much time worrying over a missed act, you will make mistakes on many later acts.
C. Note that it would be easier to decide who the receiver is if we used a finer unit of interaction. If, for example, an act was defined as a simple sentence, there would be fewer problems of shift in receiver. However, it would be much more difficult for observers to agree on what constitutes an act, and it would require many hours of training to use such a definition.
IV. A more detailed classification system
A. If you have had previous experience with behavior coding or would like to try an added challenge, try to classify each speech act after it is completed into one of the following categories (from Bales, 1980 Interaction Process Analysis system):
P. Positive Seems friendly
Dramatizes
Agrees
N. Negative Seems unfriendly
Shows tension
Disagrees
Q. Questions Asks for information
Asks for opinion
Asks for suggestion
A. Attempted Answers Gives information
Gives opinion
Gives suggestion
If the content of a speech act can be classified into more than one category, write down both categories. Don't miss acts trying to classify previous acts. Make a decision and go on. Leave the classification blank if you are having trouble keeping up with the discussion or deciding between categories.
V. Examples
A. An example of
an observer This indicates that:
protocol is:
12 Member 1 spoke to member 2.
20 Member 2 spoke to the group as a whole.
32 Member 3 spoke to member 2.
41 Member 4 spoke to member 1.
.
.
B. Example: An example of an observer protocol using the IPA classifications is:
12 Q
20 A
32 N
41 P
.
.
This could have come from a discussion in which:
Member 1 asked member 2 a question.
Member 2 answered by speaking to the group as a whole.
Member 3 spoke to member 2 and disagreed.
Member 4 spoke to member 1 encouragingly.
VI. Results
A. Tally the number of acts initiated by number of acts received on the tally sheet. Count up the total number of acts initiated by each subject and place this number in the column labeled Total. Count up the number of acts received by each subject and place this number in the row labeled Total.
B. Transfer the total number of acts initiated by each member of the group of subjects to the column under "Self" entitled "Raw number of acts" in Table 1. In the column labeled "Rank" assign each of the raw numbers a rank, giving the largest number rank 1. Split ranks if necessary. For example, if the subjects in order initiated 25, 10, 25, and 30 acts, subject 4 would be ranked 1, subjects 1 and 3 would be ranked 2.5 [(rank 2 + rank 3)/2], and subject 2 would be ranked 4.
C. Copy your partner's data in the appropriate columns of Table 1.
D. Subtract your partner's ranks from your ranks and enter the result in the column marked "d".
E. Calculate the rank order correlation (rho) between your rank order for acts initiated and your partner's rank orders. To
calculate the rank order correlation coefficient (rho) use the following formula:
= 1-6 d2 d = difference between rankings
N(N2-1) N = total number of subjects ranked
This will provide a test of the reliability of your scoring of acts initiated.
F. Repeat steps A-E for "acts received" entering the data in Table 2. Omit from this analysis acts received by the "group as a whole" (acts received by 0). This will provide a test of the reliability of your scoring of acts received.
G. In Table 1, compute the average number of acts initiated by each member of the group by averaging your raw data with that of your partner. Rank order these averages. Enter these ranks in Table 3.
H. In Table 2, compute the average number of acts received by each member of the group by averaging your raw data with that of your partner. Rank order these averages. Enter the result in Table 3.
I. Calculate the rank order correlation between the rank order of acts initiated and the rank order of acts received. This is the test of the hypothesis.
Speech Acts Initiated by Speech Acts Received
Receiver
Initiator 1 2 3 4 5 6 0 Total
Initiated
1
2
3
4
5
6
Total
Received
Table 1. Acts Initiated
Subject Self Partner d d2 7 8
(c4-c2) Average Rank
1 2 3 4
Raw # Rank Raw # Rank
1
2
3
4
5
Sum of d2=
Table 2. Acts Received
Subject Self Partner d d2 7 8
(c4-c2) Average Rank
1 2 3 4
Raw # Rank Raw # Rank
1
2
3
4
5
Sum of d2=
Table 3. Comparison of Acts Initiated & Acts Received
Subject 1 2 3 4
Acts Initiated Acts Received d d2
Rank Rank (c4-c2
Sum of d2=