Introduction
This was a real experiment done with students who were provided Albino rats who were to run towards a darker platform for being rewarded. Every investigator faces the problem of influencing his subjects in experiments and he takes caution not to be so influenced. However, there is an unconscious influence that one cannot control.
Unexpected findings in experiments are probably due to this uncontrollable influence. The hypothesis that an experimenter is able to get the results or data that he desires has not been tested before. This study has done just that.
Summary
Hypothesis tested
The E effect was tested in this experiment. The E effect is understood to be the outcome in an experiment which is the one expected by the experimenter. The hypothesis that Es are able to obtain from animal Ss the data they wanted or expected to obtain is studied here.
Participants
12 students who were enrolled in the senior division course were the participant Es in this experiment. The 13th student participant was an undergraduate student who had one years’ experience in the program as an experimenter. The sample was adequate.
Subjects
The subjects Ss were Sprague-Dawley albino rats of age 64-105 days. Each group was composed of 2 males and 3 females between 83 to 91 days of age.
Setting
An elevated T maze was built to the specifications of Ehrenfreund (1952). This had 2 interchangeable arms which were painted differently, one white and the other dark gray. The position of the arms would be interchanged daily.
The instrument used
A questionnaire that had 20-point rating scales and was similar to previously used ones was constructed for this experiment. The Es were required to rate various emotions during the experiment. They were to rate their satisfaction with their participation in the experiment, their feelings about the animal’s Ss, and the perception of their own behavior during the experiment. The scales were marked as -10 for extreme dissatisfaction to +10 for extreme satisfaction and the intermediate points were to be used as required.
Preparation for the experiment
The Es were given 5 rats each and was told that the experiment was a repetition of what they had been doing earlier. Some would be getting maze-bright rats and some maze-dull. The Es were told that the maze-bright rats were expected to show some evidence of learning on the first day itself and thereafter would show daily improvement and that the maze-dull rats would be expected to show little evidence of learning. The experiment would involve a discrimination-learning problem. Only those rats which go to the darker of two platforms would be rewarded.
Just before the experiment
One in each pair received the maze-bright and the other maze-dull rats.
The rats were labeled bright and dull randomly for the Es sake when actually the rats were all alike. Before the actual running of the rats, the Es were asked to judge whether they expected their rats to run well, on the 20 point scale: +10 for extremely well to -10 for extremely poorly.
The Ss each had one hour’s experience in the maze before the actual experiment. They obtained food from both arms in the maze experience.
Actual experiment
Each E ran each of the 5 rats Ss 10 times each day for 5 days. The Es recorded the performance of the Ss. They watched to see whether the response of selecting the darker arm which had food in it was achieved by each rat and noted the time required to reach the correct arm with food (gray arm). The arms were interchanged randomly but the same T was used by all the rats on each day.
None of the first 12 Es knew anything about the aim of the experiment or the role of the 13th participant. Only the 13th participant knew what the experiment was really for.
She was fully conscious about her motivation to get good performance for her rats without appearing to cheat. Her additional role as an informant on how the others performed without arousing suspicion was another fact that was hidden from the other 12.
Variables
The dependent variable was the experimenter’s bias of conscious and unconscious influence on the outcome of the experiment. The 12 students were of the latter group and the assistant was of the former group. The independent variable was the time taken by the rats to find the food in the maze correctly.
The tabulating of the results
A number of correct responses per day were recorded first. The mean time taken to get correct responses were then noted and tabulated. The ratings of self and the Ss on the 20 point instrument used revealed the behaviors of the Es before, during, and after the experiment and gave an idea of their biases working for and against the outcomes. These had been grouped into 3 clusters which were signed by an earlier experiment.
Results
The number of correct responses per day per Ss was recorded for the 6Es who believed that they were running maze-bright rats, the 6 Es who believed that they had maze-dull rats, and for the Assistant who knew that all the rats were the same and still expected her rats to perform well. Performance of the maze bright rats appeared to be better on days 1, 4, and 5 but not on the other days. However, when all the days were compared, t was significant with a one–tailed p-0.01. The maze-bright rats were indicated as showing a monotonical increase in function as if learning was occurring alongside. The maze-dull rats did not perform well and the performance was seen increasing till the 3rd day and then decreasing on the 4th and not changing on the 5th day. It was assumed that learning was not likely to have happened in the maze-dull rats.
The research assistant tried to and succeeded in getting performances better than the maze-bright rats. The t for all 5 days was 2.38 where 4 df was significant at the level of 0.05, one-tailed test. But the performance of her rats who had fewer runs was not increasing monotonically. The E who was biased enough to elicit good performances obtained good performances, better than those who were biased but not explicitly instructed. On average there were 60 occasions on each day when the rats did not run.
These could be because of the difficulty to discriminate between the platforms, limiting the pre-training to one hour, or because of the inexperience of the Es in running the rats.
Of the 60 occasions each day, the maze-dright rats did not run on 17 occasions and the maze-dull on 43. Since the dull rats made fewer responses, and the more responses made led to more correct responses, there was confounding of the results. The mean time is taken to make correct responses each day was tabulated. The maze-bright Ss made their correct decisions faster. They actually improved their timings each day. The maze-dull rats made no such improvement. The research assistant’s rats actually had the best timings. Her Ss ran the best. Instances of cheating were observed when Es prodded the rats to run. The data Es expected and the data they obtained were used as an index for the degree of experimenter bias. The Spearman rank correlation between expected and obtained performance was 0.43 for the Es who had maze-bright rats and 0.41 for the other group. Statistical significance was not obtained as only 6 Es were in each group.
However, when the results were combined, the one-tailed preached the 0.09 level.
Strengths
The Es were careful not to influence their Ss. The study design was good and could be replicated. The study had been done in an exemplary manner and the results proved the hypothesis that was mentioned earlier. The methodology was well conducted and the instrument used was appropriate.
Weakness
Not informing the 12 Es appeared unethical as they were ignorant of the real aim of the experiment. It seemed like encroaching or spying on a person’s emotions or behavior without informing them. However, that was the only way it could have been done to avoid bias.
Suggestion for improvement
There was another way this experiment could have been done. The first part of the experiment could have been done where all the Es were ignorant as in this experiment. Then the whole matter could have been revealed to the participants and the experiment repeated. The new results would have shown up the bias definitely as the same Es would have been running the same Ss. Comparison of the first and second groups of results could have given a better conclusion that bias did affect the results.
Conclusion
An E who was explicitly biased to get good performance from the Ss obtained good performance than those less explicitly biased. The instances of cheating could not be attributed to any real cheating as both groups did the prodding. Correlations could not reach a statistical significance.
In 21 of the 23 scale instruments, Es who had maze-bright rats believed in more favorable responses. These Es considered their rats were brighter, cleaner, tamer, and more pleasant. They also described themselves as more relaxed and casual and altogether pleasant. This conformed to the Perry Como Cluster. Both groups wished to have good performances from their rats. Future research needs to concentrate on bias in different kinds of Es and the method of mediation of this bias through auditory, visual, and tactile cues. This experiment has an implication in behavioral science in that a theory that bias can change the outcome of a behavior response has evolved.
The study cannot be deemed right or wrong. That experiment bias is truth has been confirmed in this experiment. The outcomes of all experiments are bound to be the result of experimenter bias. Future research needs to discover a method to account for this and correct results to eliminate bias.