Retrieval can increase or decrease suggestibility depending on how memory is tested: The importance of source complexity

Abstract

Taking an intervening test between learning episodes can enhance later source recollection. Paradoxically, testing can also increase people's susceptibility to the misinformation effect -a finding termed retrievalenhanced suggestibility (RES, Chan, Thomas, & Bulevich, 2009). We conducted three experiments to examine this apparent contradiction. Experiment 1 extended the RES effect to a new set of materials. Experiments 2 and 3 showed that testing can produce opposite effects on memory suggestibility depending on the complexity of the source test. Specifically, retrieval facilitated source discriminations when the test contained only items with unique source origins. But when the source test included some items that had appeared in multiple sources, the intervening test actually increased source confusions. These results have implications for a wide variety of learning situations. We focused our discussion on eyewitness memory, source complexity, and reconsolidation. Taking an intervening test between learning episodes can enhance later source recollection.

Impairing existing declarative memory in humans by disrupting reconsolidation

“…From a procedural standpoint, this experiment is highly comparable to experiment 1, and the effect size of the reconsolidation-associated amnesia was similar in these experiments (d = 0.57 for experiment 1 and d = 0.50 for experiment 4), which suggests that the use of source-free recognition did not diminish the magnitude of reconsolidation-associated amnesia appreciably. Nonetheless, the effect size was numerically smaller in experiment 4; thus, consistent with prior research (11,31), reactivation-induced source confusions might have played a partial role in the memory impairment observed in the previous experiments.…”

“…We devised a special true/false recognition test to estimate the accessibility of the original memory (57,58). Recall tests are not suitable for our purpose because participants can withhold responses based on various metacognitive control processes (31,59), making assessment of the true strength of a memory difficult. In the recognition test, participants encountered one statement during each trial and indicated whether the statement was true (e.g., the terrorist used a hypodermic syringe on the flight attendant) or false (e.g., the terrorist used a chloroform rag on the flight attendant).…”

“…Having established the timedependent nature of this effect, we sought to rule out source confusion as the basis of this effect. Two recent papers have reported that, when people are reminded of the original learning episode immediately before new learning, they often misattribute details encountered during the new learning phase to the original learning phase (11,31). Therefore, it is possible that our manipulation did not produce an amnesic effect on the original memory per se; rather, recognition performance was reduced as a result of source confusions.…”

Testing Increases Suggestibility for Narrative‐based Misinformation but Reduces Suggestibility for Question‐based Misinformation

“…Similarly, the impact of misinformation is weaker when the witnessed event is still "fresh in mind" than if forgetting has set in (e.g., Chan & Langley, 2011;Loftus, Miller, & Burns, 1978). However, contrary to the prediction that testing would reduce suggestibility, Chan and colleagues have repeatedly found that testing increased the misinformation effect (e.g., Chan & Langley, 2011;Chan et al, 2009;Chan, Wilford, & Hughes, 2012). One explanation for this is that completing the initial test paradoxically enhanced learning of the subsequent misinformation, thus increasing its likelihood of being recalled (e.g., .…”

“…These include the length of the event video, initial testing manipulated within-vs. between-subjects, a 48 hr vs. 30 min retention interval, and misinformation introduced via questions vs. narrative. Based on current knowledge, the first three differences are likely irrelevant, because the RES effect has been demonstrated across different witnessed event lengths Chan et al, 2012), retention intervals (Chan & Langley, 2011), and in both within-and between-subjects designs (Chan & LaPaglia, 2011). We are thus left with the fourth methodological differencethat Pansky and Tenenboim presented their misinformation via written questions whereas studies showing RES presented misinformation in a narrative.…”