Experimental and meta-analytic evidence that source variability of misinformation does not increase eyewitness suggestibility independently of repetition of misinformation

Abstract

Considerable evidence has shown that repeating the same misinformation increases its influence (i.e., repetition effects). However, very little research has examined whether having multiple witnesses present misinformation relative to one witness (i.e., source variability) increases the influence of misinformation. In two experiments, we orthogonally manipulated repetition and source variability. Experiment 1 used written interview transcripts to deliver misinformation and showed that repetition increased eyewitness suggestibility, but source variability did not. In Experiment 2, we increased source saliency by delivering the misinformation to participants via videos instead of written interviews, such that each witness was visibly and audibly distinct. Despite this stronger manipulation, there was no effect of source variability in Experiment 2. In addition, we reported a meta-analysis (k = 19) for the repeated misinformation effect and a small-scale meta-analysis (k = 8) for the source variability effect. Results from these meta-analyses were consistent with the results of our individual experiments. Altogether, our results suggest that participants respond based on retrieval fluency rather than source-specifying information.

Keywords: eyewitness memory; eyewitness suggestibility; misinformation; misinformation effect; repetition; source variability.

Figure 1

Illustration of critical items contained in each interview type and condition. Control interviews, with no misleading items, are indicated in gray. Each symbol represents a different piece of misinformation. In the 3X condition, each piece of misinformation appeared in all three interviews. In the 1X condition, all misinformation appeared only once and in a single interview. In the 1X-Distributed condition, all misinformation appeared only once, but the misinformation was distributed across three interviews.

Figure 2

Illustration of procedure for experiments 1 and 2. The figure shows the estimated completion time for Experiment 1. Experiment 2’s estimated completion time was ~90 min due to the length of the interview videos.

Figure 3

Illustration of interview information and format for Experiments 1 and 2.

Figure 4

Proportion correct by item type in Experiment 1 as a function of repetition and sources. Each dot represents the data of an individual participant. Jitter was introduced to disperse the data points horizontally for visualization purposes. The violin element displays data density.

Figure 5

Confidence accuracy calibration as a function of item type in Experiment 1. Data points represent observed probabilities, and bands represent 0.95 CI for the fitted multilevel regression lines.

Figure 6

Depiction of interviewer and interviewees for 1W and 3W conditions by interview type. The interviewer remained the same across interviews. In the 1W condition, participants viewed the same witness wearing different clothes for each interview. In the 3W condition, participants viewed three different witnesses. The order of witnesses was randomized and counterbalanced.

Figure 7

Proportion correct by item type in Experiment 2 as a function of repetition and sources. Each dot represents the data of an individual participant. Jitter was introduced to disperse the data points horizontally for visualization purposes. The violin element displays data density.

Figure 8

Confidence accuracy calibration as a function of item type in Experiment 2. Data points represent observed probabilities, and bands represent 0.95 CI for the fitted multilevel regression lines.

Figure 9

A forest plot of repetition effect sizes for misinformation effect studies.

Figure 10

A forest plot of the source variability effect in the current study and Foster et al. (2012) by repetition condition.