Running head: METAMEMORY: SELF-REFERENCE AND MISINFORMATION

Metamemory and the Effects of Self-Reference on the

Misinformation Effect

Jennifer A. Masterson

Abstract

The purpose of this study was to investigate the effect of self-reference on the misinformation effect. Participants initially were presented a slide show in which half were asked to relate the images to themselves while the other half were asked to relate the images to an unfamiliar person. After a short delay, participants were given a narrative that contained misled information. It was believed that participants who were told to relate the images to themselves would perform better on a subsequent recognition test than those who were told to relate the images to an unfamiliar person. It was also predicted that participants who used self-reference as a way to encode the material would be more confident in their answers than participants in the other group. There were main effects of both variables, and an interaction between the types of questions asked. However, results did not support the self-reference effect in terms of misinformation. Participants who were told to relate the images to themselves actually performed worse on questions that had been misled than did participants in the unfamiliar group. Also, there was no significant effect of the group the participant was in, unfamiliar or self, and the confidence percentages. These finding do not support those of Harvley et. al. because in this study main effects were found between the variables while his study had no significant main effects of either variable.

Metamemory and the Effects of Self-Reference on the

Misinformation Effect

Metamemory refers to people’s awareness, knowledge and control of their memory (Matlin, 2002). When people are presented with misleading information about precisely learned or witnessed material, they are likely to mix up the original information with the misleading information (Loftus, 1991; Wright & Loftus, 1998). This phenomenon is referred to as the misinformation effect. Over the past quarter of a century, hundreds of studies have been conducted that demonstrate the effects of misinformation on memory performance (Wright & Loftus, 1998).

It has been found that people are especially susceptible to being influenced when there is a delay between when initial information is encoded and when the misinformation is given (Harvley, Patel, Tucker, & Thrasher, 1996). Loftus, Miller, and Burns (1978) found that misinformation effects are largest when there is a large delay between initial presentation of information and the misleading information and when testing comes close in time after the misinformation is given. In their 1978 study, Loftus and her co-authors showed participants a series of slides. In the sequence, a car stopped at an intersection and proceeded to turn and hit a pedestrian. Half of the participants saw a slide with a yield sign while the other half saw a slide with a stop sign. Twenty minutes to a week after the slides were presented, participants answered questions regarding the details of the accident. The questions consisted of ones that were consistent with the original slides, questions that were misleading and questions that were neutral. Loftus and her co-authors found that people who saw inconsistent information were much less accurate than people in the other two conditions. They also found that the longer the period of delay between when slides had been shown and when participants were tested, the less the percentage of answers subjects got right.

In many post-event misinformation studies, participants are shown a sequence of slides depicting scenes, are misled about various aspects of the original slides, and are then given a forced choice recognition task that requires the participant to choose between the original item and the misinformed item (Loftus et al., 1978). Some studies use tests of recognition that involve yes/no questions, and others use multiple choice questions to assess scores (Loftus, 1998). However, no matter which form of test is used, all of them can be used to not only assess scores that reflect the misinformation effect, but also investigate how the information is encoded.

In a study conducted by Harvley et al. (1996), it was predicted that familiar stimuli would be a good way of encoding information thus reducing the misinformation effect. However, they did not find significant results to support this hypothesis, which could be due to familiar stimuli not being a strong enough form of encoding. Therefore, because research has shown that self-reference is a good way to encode information, it is conceivable that it would in turn decrease the misinformation effect.

The self-reference effect (SRE) seems to result because “the self is a well-developed and often-used construct that promotes elaboration and organization of encoded information” (Symons & Johnson, 1997, p. 371). It is indicated that self-referent encoding has been shown to produce better recall than mere semantic coding (Klein & Kihlstrom, 1986). Not only has it been indicated that the SRE affects encoding, but it has also been found to be an important component of deeper level processing strategies (Lord, 1980). Lord also found that when considering actual precepts, it is obvious that people are highly likely to remember information that is related to the self. There have been many studies done that suggest the self-structure as unique, relative to other concepts ( Rogers, Kuiper & Kirker, 1977; Symons & Johnson, 1977) in its motivational and affective implications as well as in its structure and content.

Some studies have also suggested that self-referencing may aid in the learning and retention of course material (Hartlep & Forsyth, 2000). For example, D. R. Forsyth and Wibberly (1993) asked students to evaluate whether each adjective presented orally in a list was self-descriptive. When later asked to recall the lists, students recalled more of the self-referent adjectives. Therefore, it is plausible that teachers should use this method for helping students to retain material for exams.

The final part of misinformation studies examine the “quality of the resulting memories by measuring some aspect of these memories” (Loftus, 1998). In order to further explore how SRE influences the misinformation effect, the present study uses a scale of confidence to assess the participants level of confidence during the acquisition period. Harvley et al. (1996) also used a scale of confidence and did so as an entirety of all twenty questions asked during the recognition tasks, which allowed for one score of confidence for each participant. The questions asked were misinformed, reinforced or controlled. The misled questions addressed material that had been changed between the initial information given and the narrative. Reinforced questions, on the other hand, related to material that was congruent in both the initial information and the narrative. The control questions referred to material presented only in the initial information, and not mentioned in the narrative. In the present study, I also chose to include a confidence measure, requiring each participant to give a percentage relating to how accurate the participant believed they were in answering the entire set of twenty questions.

As Harvley et al. (1996) hypothesized with a familiar stimulus, I also predict that the use of a self-reference will decrease the misinformation effect. Participants who are merely told to imagine someone else partaking in the situations presented will be more apt to being misled during a subsequent test when given a narrative with misleading information. However, I do not believe that either group would perform better on control questions when given a subsequent test due to the participants not seeing this information since the slides, thereby not having new material to interfere with the old. I predict that all participants will perform best on the reinforced questions due to the material being introduced to each participant twice, allowing for repetition, which helps with retention.

As for predicted accuracy, I expect that on misleading questions, people who used the self to encode information would be more accurate in their confidence predictions than those who used an unfamiliar stimulus to code the information.

Method

Participants

Participants were 56 undergraduate psychology students who volunteered to partake in the study.

Materials

For presentation of the initial material, 46 slides were used. These slides were picture images created by a 35-mm camera projected on a television screen.

The narrative component of the experiment consisted of 12 sentences containing 9 misled items and 6 reinforced items.

For the final test, a set of 20 questions was devised in the following manner. There were 9 misled questions, 6 reinforced questions and 5 control questions. Each question was asked in a different manner (see Table 1). The nine misled questions referred to information that had been misled in the narrative, while the six reinforced questions referred to information that was consistent in both the initial slides and the subsequent narrative. The control questions pertained to images that were seen solely in the original set of slides and were not found in the narrative.

Two interpolated tasks were devised to serve as delay periods between: (1) The initial 46 slides and the narrative, and (2) the narrative and the final set of questions. The first task was a word find and the second task was a crossword puzzle. For both tasks, subjects were allotted 5 min. These tasks were used due to past research showing that a delay between initial presentation of information and presentation of misled information increases the chances of the participant being misled.

Procedure

This experiment was conducted as a 3 X 2 design. The misinformation effect variable was within subjects; all of the participants received misled, reinforced and control questions. The second variable had to do with whether the participant was told to imagine himself or herself as being part of the situation portrayed in the slides or if the participant was told to imagine someone else as being part of the situation portrayed in the slides. This variable was between subjects.

Participants were exposed to seven different phases: the instructional phase, slide presentation, interpolated task, the narrative, second interpolated task, the test questions phase and confidence percentage phase as they had been in the study done by Harvley et al (1996). In the instructional phase, half of the participants were told to relate the images being presented to themselves and the other half of the participants were told to relate the images to someone else, in this situation an unfamiliar person named Jack. Participants were randomly assigned to one of the two groups.

After reading the instructions, participants viewed the set of 46 slides and were allowed 3 s to view each slide. Both groups, the groups who was told to use self-reference and the group who did not use self-reference, viewed the same set of 46 slides. After completing this task, participants were allotted 5 min to work on a word find.

The fourth phase of the experiment was the narrative story. The participants were informed that the narrative depicted a boring day in the life of a 21-year-old student named Jack. This narrative contained 12 sentences in which some of the material was manipulated to differ from the slide presentation. Participants were given as much time as needed to read the narrative in order to ensure they had read and understood the story. Upon completion of this task, participants entered the fifth phase of the experiment and were asked to work on a crossword puzzle until told to stop. This task lasted for 5 min.

The sixth phase of this experiment was the acquisition test question phase. Here, participants were instructed to answer a series of 20 questions regarding the slide show. Nine of the questions were misled, where the narrative differed from the original slides, 6 of the questions were considered reinforced because they related to information that had been viewed in the slides and reinforced in the narrative. The final five questions were control questions, which pertained to material that had been seen in the initial 46 slides, but was not mentioned in the narrative.

After completing the forced recognition task, participants entered into the seventh and final phase of the experiment. Participants were asked to record a confidence percentage that reflected how accurate they believed they were in answering the twenty questions. The questionnaires were then collected and the data was recorded.

Results

The SRE was not demonstrated in this study (see Table 2). Although the independent ANOVA for between subjects was significant, it was opposite of what was expected, F(1, 54) = 4.349, p<.05. However, the repeated measures ANOVA showed there to be a significant difference between the group that used self-reference and the group that did not in terms of types of questions asked, F(1, 108) = 18.697, p<.01. By using this information in conjunction with the mean scores for each group, it is evident that there is an interaction between the two independent variables (see Table 3). Through the subsequent use of t-tests it was also apparent that there was a significant difference between the types of questions asked (see Table 4). However, there was a significant difference between misinformed and reinforced questions and between reinforced and control questions, but there was no significant difference between misinformed and control questions. Therefore, Tukey tests were run on the two significantly different types of questions. For Misinformed and Reinforced questions there was a significant difference F(2, 53) = 31.036, p<.01. Reinforced and Control questions were also significantly different F(2, 53) = 23.659, p<.01.

An ANOVA was also used to calculate any differences between unfamiliar and self-referent groups in terms of confidence percentages. For this between subjects ANOVA I found there to be no significant difference, F(1, 54) = .745, p>.05. On the other hand, when a repeated measures ANOVA was run to look at how well the participants performed on the recognition test in each of the three types of questions, there was a significant difference, F(1, 84) = 5.88, p<.05. This showed that although participants who used self-reference did not expect to do better than those who did not use self-reference; they were accurate in their responses in terms of how well they perceived they had done on the recognition tasks.

Discussion

Although this study did not support past research regarding the self-reference effect (Forsyth et al., 1993; Loftus, 1998; Rogers et al., 1997), it did support the preponderance of my hypotheses. Participants in both groups scored higher on questions that were reinforced than on questions that were controlled. This, as argued previously, is a result of the reinforced information being viewed twice, while the control information had only been viewed once. Also, participants in the self-reference group did have the lowest percentage of correct answers when the questions were misled. However, participants in the unfamiliar stimulus group scored higher on questions that were misled than they did with the other two types of questions.

Past research shows that Self-Reference is suppose to increase one’s retention of material (Lord, 1980). Although my findings did not support this research, I do still feel that it is a strong encoding mechanism. It is possible that participants were not entirely sure on how to relate all of the images to themselves, thereby not increasing their retention of the material.

In order to further assess participants’ accuracy of confidence, it may be necessary to include a scale for each of the twenty questions. This could show a difference between the type of question asked and how confident participants were in their answers depending on whether they used self-referencing or not.

References

Forsyth, D. R., & Wibberly, K. H. (1993). The self-reference effect: Demonstrating schematic processing in the classroom. Teaching of Psychology, 20, 237-238.

Hartlep, K. L., & Forsyth, G. A (2000). The effect of self-reference on learning and retention. Teaching of Psychology, 27, 269-271.

Harvley, P. R., Patel, B. R., Tucker, S. J., & Thrasher, L. W. (1996). Meamemory and the effects of familiar stimuli on misinformation effect. Unpublished manuscript, University of Florida. Retrieved March 26, 2003, from http://www.psych.ufl.edu/~levy/96_9.htm

Klein, S. B., & Kihlstrom, J. F. (1986). Elaboration, organization, and the self-reference effect in memory. Journal of Experimental Psychology: General, 115, 26-38.

Loftus, E. F. (1991). When a lie becomes memories of truth: Memory distortion after exposure to misinformation. American Psychological Society, 3, 121-126.

Loftus, E. F., Miller, D. G., & Burns, H. J. (1978). Semantic integration of verbal information into visual memory. Journal of Experimental Psychology: General, 4, 19-31.

Lord, C. G. (1980). Schemas and images as memory aids: Two modes of processing social information. Journal of Personality and Social Psychology, 38, 257-269.

Matlin, M. W. (2002). Cognition: Fifth Edition. SUNY Geneseo, Earl McPeek.

Rogers, T. B., Kuiper, N. A., & Kirker, W. S. (1977). Self-reference and the encoding of personal information. Journal of Personality and Social Psychology, 35, 677-688.

Symons, C. S., & Johnson, B. T. (1997). The self-reference effect in memory: A meta-analysis. Psychological Bulletin, 121, 371-394.

Wright, S. B., & Loftus, E. F. (1998). How misinformation alters memories. Journal of Experimental Psychology, 71, 155-164.

Table 1

Narrative and Question Type Examples

Question Type Question

Control Steve has a ____ phone.

A.) Cordless

B.) Rotary

Reinforced Steve snacks on a?

A.) Banana

B.) Apple

Misled Steve plays with his pet ____ before bed.

A.) Cat

B.) Dog

Narrative Excerpt:

Steve watched a little TV and decides to make a phone call to one of his friends. Before bed, Steve decides to have an apple as a snack. Steve plays with his ped dog “Danny-boy” and wishes him a goodnight (Steve actually has a cat).

(This excerpt and questions came from the Harvley et al. study, 1996)

Table 2

Mean # of questions answered correctly and Standard Deviation from the mean

Group

Task Self-Reference SD Unfamiliar SD

Misled .46 .20 .68 .18

Reinforced .89 .13 .61 .16

Control .65 .22 .51 .21

Table 3

Table 3

T-scores for each of the three types of questions in terms of group (self vs. unfamiliar)

Grouping T-score probability

Misinformed by Self vs. Unfamiliar 4.310 p<.01

Reinforced by Self vs. Unfamiliar -7.016 p<.01

Control by Self vs. Unfamiliar -2.367 p<.05