The aim of the present experiment was to demonstrate that group discussion modifies personal opinion, and that people alter their initial judgment in the direction of a group consensus. Participants in the experiment were asked to individually review information about eight contentious topics in an opinion questionnaire. They were then asked to record their opinion about each topic using a Likert scale of one to six (1 = total agreement with the topic, 6 = total disagreement). For example, if the subject completely agreed with the legalisation of marijuana, they would rate the topic as one. Complete disagreement with the legalisation of marijuana would score six. Participants were then grouped and discussed four of the eight topics. Following discussion, participants reappraised their initial individual ratings for those topics that had been discussed in their group.
The measured variable was the mean shift score for each topic. A positive shift score would indicate that the participant had changed their opinion rating following group discussion, towards group opinion. A negative shift score would show disagreement with the group and a move away from group opinion.
The manipulated variable for the present experiment was whether the groups discussed the topic or not.
It was hypothesized that the topics discussed by groups would have a positive shift score, indicating that subjects were influenced by members of the group, and that those topics not discussed would have a mean shift score and standard deviation of zero, representing no change in attitude.
Method
Participants
Participants in the present experiment were 20 undergraduate Psychology students from the University of Otago. All participants were required to participate as part of the terms requirements for their Psychology course. There were 16 females and four males with a median age of 19 years. It was assumed that no subject had any disabilities or impairments that would affect results
Apparatus
Supplied to each subject at the beginning of the experimental phase was a booklet containing an outline of eight controversial issues. Arguments both for and against each issue were described.
The system used for rating each topic was a Likert scale from 1 – 6. If the subject were strongly in favour of a given position on the topic, they would rate the topic as a one (e.g., complete agreement with the legalisation of marijuana). If they felt strongly negative toward the position on the topic, then it would be scored as six (e.g., complete opposition to the legalisation of marijuana). These ratings were recorded on paper separately for collection and collation later in the experiment.
IBM computers, connected to the Department of Psychology Local Area Network (LAN) were used by the subjects to input their ratings for each phase of the experiment. The LAN was also used to collect data and calculate the average shift score and standard deviation for each topic.
Procedure
The procedure involved three phases. In phase I, each subject was given a booklet, and was instructed to read about each of the eight topics. Once read, participants were asked to record how strongly they felt about the topic based on the Likert scale.
On completion of the grading process (beginning of phase II) the subjects were randomly assigned into groups of five. In these groups, subjects were allocated four of the eight topics, and then asked to spend ten minutes discussing each of the four topics amongst themselves. At the end of each ten minute block, participants were asked to reassess their opinion for that topic. If they felt more or less strongly about the issue, they were asked to change their allocated mark to what they felt was more appropriate.
At the end of the group session (beginning of phase III) subjects were again required to individually re-read the supplied literature, and assign each a final grade. Each subject then input all three sets of results into a computer connected to the LAN. All data was collated to find the average shift score and standard deviation for each topic. Shift scores were the difference between marks allocated at phase I and those recorded at the completion of phase III.
Design
The design of this experiment was within subjects, as all participants experienced the same manipulations; i.e., were exposed to the same stimuli, and expected to perform the same task. The study also made use of counter-balancing, in that each topic was discussed by two groups and therefore equally. The manipulated (independent) variable for the current experiment was whether or not the topic was discussed (i.e., it had two levels; discussed or not discussed). The measured (dependent) variable was the average shift score and standard deviation for each topic.
Results
The measured (dependent) variable was the mean shift score for each discussion topic. The group mean and standard deviation for each topic was collated, and an overall mean and standard deviation of the eight topics was calculated. This data is shown in Table 1.
Table 1
Mean Shift Score for Each Discussed and Non-Discussed Topic
Although subjects altered their opinion in six of the eight discussed topics (which can be seen from the positive shift score values), mean shift scores and standard deviations for topics one and five were zero, showing no change in opinion. At the opposite end of the spectrum, participants promoted topic six one point towards the group mean (on average). However, because the group standard deviation is 1.33, and thus higher than the mean, it can be seen that some subjects disagreed with the group judgment, and moved away from the group rating. Also note that the subjects that did not discuss topic six did not alter their initial attitudes at all (mean = 0, standard deviation = 0). This is also true for topics two and eight.
The overall standard deviation for the mean discussed shift score was 0.5. This is high when compared to the mean, which was 0.36. This suggests that there was a large shift score in both directions (positive = towards the group mean, negative = away from the group mean).
For the non-discussed topics, the overall mean shift score was zero, demonstrating that there was no general change in opinion for the topics. However, because the standard deviation was 0.5, it can be seen that not only did subjects initial opinions change, but there was also high variability. Topic five showed a mean increase toward the group mean of 0.1, but a standard deviation of 0.88 again indicating high variability.
Figure 1 Mean shift scores for discussed and non-discussed topics. The error bars represent standard deviation.
Figure 1 shows the two levels of the manipulated variable, in which the variability of the shift scores is clear. The variability of topic five is especially obvious here, where the range that the standard deviation covers is shown.
Overall, six of the eight discussed topics had moderate positive shift scores. Non-discussed topics had low-to-zero shift scores. In general, the trend of the data showed that any shift in opinion following group discussion was insubstantial.
Discussion
The results show that there was a positive shift of score in the discussed topics. Although it is not considerable, a positive shift score of 0.5 shows group members attempted to conform to group opinions. From the results it can also be seen that non-discussed topics had an overall mean shift score of zero, indicating that subjects did not change their attitudes during testing. However, both discussed and non-discussed topics had high overall standard deviations when compared to the mean (standard deviation of 0.5 for each), demonstrating high variability in the shift scores provided by group members.
The mean overall shift scores thus support the hypothesis that for discussed topics there would be a positive shift score, and non-discussed topics would have a mean shift score of zero. The large variability of results means however, that this is a tentative conclusion.
It is difficult to compare the present experiment with that of Festinger et al (1950), as that study was conducted over a year, and was observational rather than experimentally manipulated: the subjects were not put through a specific testing regime, but rather their behaviours observed (though there was an interview process to gather information on changes in beliefs) over time. However, both Festinger et al, and the present study found that individual’s opinions converged towards group consensus.
Sherif’s (1935) results showed clearly, that in group environments, group members would eventually agree on a topic or situation. The most striking similarity between Sherif’s study and the present experiment, is that subjects in the groups were allowed to come to their own conclusions (make there own interpretation of the situation). This would suggest that if the groups in the current study repeated phase II a number of times, the variation in opinion would be less.
Participants in Asch’s (1955) experiment came to an incorrect conclusion on nearly 37% of trials, as they followed group opinion. However, this leaves two-thirds of subjects frequently going against group attitudes, and thinking independently. Like participants in Asch’s (1955) experiment, it appears that participants in the current experiment were also independent thinkers, who made minimal effort to concede to the group views.
Sherif (1935) and Asch (1955) were confident in accepting their theories of social conformity. Why then, did the results from the present experiment not agree so convincingly with them? One possibility lies in the nature of the participants. University students are thought to be confident in their beliefs, and may, therefore be less prone to the larger shifts in opinion that would be expected from community subjects. This would suggest that conformity is not necessarily a rule that all of society exhibit. Rather some people may be independent enough in their beliefs to go against group consensus. Both Sherif and Asch used only college students in their respective studies. However, both the autokinetic effect, and the judgement of line lengths are based on the subject’s perception, rather than a (moral) belief or opinion, as it is in this case. This may explain the differing results between historic studies and the present experiment.
It would be interesting to extend this experiment to include a wider and larger group of participants. Using a larger cross-section of the community would allow a more descriptive set of results. It may also be seen that the current hypothesis can be accepted (or rejected) with more authority, as the evidence will be more substantial.
Another limitation of this study would be the bias that the subjects experienced. Most participants knew before the start of the experiment that their opinions were expected to change. Subjects may have therefore, consciously or not, decided to go against that hypothesis. This could be addressed by not allowing the subjects to know what the experiment was investigating.
It is also worth considering how much knowledge or prejudice towards a topic that a subject may have. It can be seen in the results that some of the discussed topics were more susceptible to attitude change than others (topics one and five had no shift, while topic six moved one point toward the group mean). This would suggest that subjects had a fairly determined opinion about the death penalty and the use of repressed memory in court cases before the experiment. However, subjects were unsure and open to influence when it came to drug testing in the work place. Having no knowledge and faced with a persuasive argument may result in high shift scores. A subject with an emotional link to a topic maybe overly biased against it. By choosing discussion questions that are less topical or less emotionally charged, will mean that preset attitudes will be less of a factor. This would be difficult as subject’s knowledge, opinions or experiences cannot be predetermined.
Future research into social influence seems to be limitless. Studies into the criminal justice system, and specifically how jury’s may be swayed in their verdict, may lead to the revamping of the court systems throughout the world.
Propaganda and politics attempt to make use of social influence when trying to sway voters or social opinion to their cause. The theory suggested by Festinger, Schachter and Back (1950) that repeated exposure to a belief will eventually lead to conformity is applied heavily in election campaigns and in war zones to garner support. Investigating the extent to which these tactics influence peoples decision making could lead to new methods of persuading voters.
Another avenue open to investigation is that of peer-pressure amongst adolescents. Teenage years seem to be the time when humans are most susceptible to social influence. This pressure from friendship groups can lead to bullying and overall, be counter productive to society. Further study on this topic could lead to substantial gains in the extent of knowledge on social influence.
The conclusion of the present experiment is that individual opinions are influenced by group discussion, though the strength of the subjects’ individualism does control the extent of which this social influence occurs.
References
Argyle, M. (1969). Social interaction. London: Methuen
Asch, S. (1955). Opinions and social pressure. Scientific American, 193 (5), 31 – 35.
Baron, R. S., Vandello, J. A., Brunsman, B. (1996). The forgotten variable in conformity research: Impact of task importance on social influence. Journal of Personality and Social Psychology. 71 (5), 915 – 927.
Carlson, N., & Buskist, W. (1997). Psychology: The science of behaviour (5th ed.) Boston: Allyn and Bacon.
Demarzo, P.M., Vayanos, D., Zwiebel, J. (2001). Persuasion Bias, Social Influence and Uni-dimensional Opinions. Graduate School of Business. Stanford University.
Festinger, L., Schachter, S., Back, K., (1950). Social pressures in informal groups: A study of human factors in housing. New York: Harper.
Laboratory Manual Psychology 111/112 (2002). Dunedin. Department of Psychology. University of Otago.
Sherif, M., (1935). A study of some social factors in perception. Archives of Psychology, 107, 60.
Sherif’s autokinetic effect is defined as the apparent movement of a stationary light in a dark room (Lab Manual, Psychology 111/112, 2002)