Comparing visual memories to similar visual inputs risks lasting memory distortion

Abstract

Maintaining perceptual experiences in visual working memory (VWM) allows us to flexibly accomplish various tasks, but some tasks come at a price. For example, comparing VWM representations to novel perceptual inputs can induce inadvertent memory distortions. If these distortions persist, they may explain why everyday memories often become unreliable after people perform perceptual comparisons (e.g., eyewitness testimony). Here, we conducted two experiments to assess the consequences of perceptual comparisons using real-world objects that were temporarily maintained in VWM (n = 32) or recalled from long-term memory back into VWM (n = 30). In each experiment, young adults reported systematic memory distortions following perceptual comparisons. These distortions increased in magnitude with the delay between encoding and comparisons and were preserved when memories were retrieved again a day later. These findings suggest that perceptual comparisons play a mechanistic role in everyday memory distortions, including situations where memory accuracy is vital.

Working Memory Content Is Distorted by Its Use in Perceptual Comparisons

“…Although we did not use any secondary task to distract attention away from the original WM representation prior to the onset of the similar probe, it is possible that the similarity-induced memory bias happens only when a WM is compared with a similar input too soon after encoding. We tested this possibility in a follow-up study (Saito et al, 2021) by directly manipulating the interval between memory encoding and similarity judgments. Here, we replicated the similarity-induced memory bias even when participants performed a similarity judgment between a perceptual input and a memory encoded 24 hr before by retrieving it into WM (Atkinson & Shiffrin, 1968; Cowan, 1999; Fukuda & Woodman, 2017).…”

Judgments during perceptual comparisons predict distinct forms of memory updating

“…In the present experiments, we tested the prediction that using memories in perceptual comparisons can trigger distinct forms of memory updating. Specifically, we predicted that perceived similarity between mnemonic and perceptual representations would result in representational integration, as demonstrated previously (Fukuda et al, 2022;Saito et al, 2021), and that perceived sameness would result in probabilistic replacement by the perceptual representation. We based this prediction on the fact that perceived sameness between sequential stimuli indicates a sense of shared identity that is fundamentally untrue for perceived similarity and allows individuals to rely on the more recent stimulus to facilitate their behavioral goals.…”

“…An immediate next step will be to confirm that memory replacement following 'same' judgments reflects a bona fide change in the memory and not a temporary phenomenon brought about by decisional processes at the time of the report. Previous work has shown that biases following 'similar' judgments are recalled again 24-hours later, suggesting that memory errors following perceptual comparisons reflect veridical changes in memory that may be permanent (Saito et al, 2021). As such, it is not unreasonable to predict that memory errors following 'same' judgments reflect an analogous process.…”

“…Lastly, it will be useful for future work to build on these findings by testing the representational boundaries that perceptual comparisons operate within. While the distorting effect of perceptual comparisons has been illustrated in simple visual features (Fukuda et al, 2022;, real-world objects (Saito et al, 2021), and human faces (Teoh et al, 2021; see also Plummer et al, 2021), no study has tested the effect of performing perceptual comparisons between event representations that are perceived in different sensory modalities. An overwhelming majority of studies on the misinformation effect have implemented paradigms in which false details about a prior visual experience are provided verbally to participants (see Loftus, 2005 for review).…”

“…These similarity-induced memory biases were explained computationally by representational integration between the memory and probe items that produced a reliable bias across trials. Later work then expanded upon these findings to show that memory biases induced during perceptual comparisons persist into LTM (Saito et al, 2021) and cannot be readily explained by trial-wise variability in VWM precision nor the physical similarity between the encoded target and perceptual probe in the underlying feature space . Together, this evidence has been used to posit that perceptual comparisons act as a cognitive mechanism that updates memory representations and may explain systematic memory errors that are commonly observed JUDGMENTS DURING COMPARISONS PREDICT DISTINCT UPDATING 5 in real-world scenarios, especially those where perceptual comparisons are performed explicitly as part of an ongoing task (e.g., eyewitness lineups; Steblay & Dysart, 2016;Wixted et al, 2016).…”