Is working memory inherently more ‘precise’ than long-term memory? Extremely high fidelity visual long-term memories for frequently encountered objects

Abstract

33Long-term memory is often considered easily corruptible, imprecise and inaccurate, especially in 34 comparison to working memory. However, most research used to support these findings relies on 35 weak long-term memories: those where people have had only one brief exposure to an item. Here 36 we investigated the fidelity of visual long-term memory in more naturalistic setting, with 37 repeated exposures, and ask how it compares to visual working memory fidelity. Using 38 psychophysical methods designed to precisely measure the fidelity of visual memory, we 39 demonstrate that long-term memory for the color of frequently seen objects is as accurate as 40 working memory for the color of a single item seen 1 second ago. In particular, we show that 41 repetition greatly improves long-term memory, including the ability to discriminate an item from 42 a very similar item ('fidelity'), in both a lab setting (Exps. 1-3) and a naturalistic setting (brand 43 logos, Exp. 4). Overall our results demonstrate the impressive nature of visual long-term memory 44 fidelity, which we find is even higher fidelity than previously indicated in situations involving 45 repetitions. Furthermore, our results suggest that there is no distinction between the fidelity of 46 visual working memory and visual long-term memory, but instead both memory systems are 47 capable of storing similar incredibly high fidelity memories under the right circumstances. Our 48 results also provide further evidence that there is no fundamental distinction between the 49 'precision' of memory and the 'likelihood of retrieving a memory', instead suggesting a single 50 continuous measure of memory strength best accounts for working and long-term memory. 51 52 53 Miner, Schurgin, Brady: Repetition and long-term memory fidelity 2 Public Significance Statement 54Visual working memory appears to be based on persistence of perceptual representations in 55 visual cortex. By contrast, visual long-term memory depends critically on semantically 56 meaningful stimuli and is organized by categories and concepts. Does this mean visual long-term 57 memory is fundamentally incapable of storing as precise perceptual information as visual 58 working memory? In the current work, we show that after being shown multiple repetitions of 59 the same item, visual long-term memory can represent incredibly precise visual details. In fact, 60 after just 8 repetitions, visual long-term memory can be as precise as our very best visual 61 working memories. This provides evidence that there is not a fundamental distinction between 62 the fidelity of visual working memory and visual long-term memory. 63 64 Keywords: visual long-term memory; visual working memory; repetition; memory fidelity; 65 memory capacity 66 Miner, Schurgin, Brady: Repetition and long-term memory fidelity 3 Humans have remarkable visual long-term memory abilities, capable of storing thousands of 67 items (Standing, Conezio & Hyber, 1970) with high fidelity (Brady, Konkle, Alvarez & Oliva, 682008). Thus, while working...

Psychophysical Scaling Reveals a Unified Theory of Visual Memory Strength

“…To demonstrate TCC's applicability to multiple memory systems, not just visual working memory, we fit data from a visual long-term memory continuous report task with colors. Unlike the previous datasets, this data had been previously reported in the literature 30 . Participants performed blocks where they sequentially saw 40 real-world objects' that were randomly colored, and then after a delay, reported the color of the object using a color wheel (as in Brady et al 31 ).…”

“…Another way to test this is to fit the mixture model -which purports to measure two distinct parameters -to data from a single stimulus space (e.g., from a single color wheel) and ask whether the state-trace plot shows evidence of a single way memory changes or multiple ways 35 . Figure 8 shows this plot for all data from the current paper (e.g., the 22 conditions shown above, plus the other experiments) and from all the conditions in Miner et al 30 , which provided the long-term memory data fit above. As can be clearly seen in this plot, the two parameters always change together: while not linear in their relationship, they are nearly perfectly related -and their relationship is well predicted by the zero-free-parameter prediction of TCC (e.g., TCC's prediction across a range of d′ values).…”

“…Thus, TCC claims that the particular manipulation (encoding, set size, delay) used to change memory strength should not selectively change one mixture model parameter or another (e.g., encoding changing precision; high set sizes affecting only guess rate, etc), but that both should always change together. To visualize this, we show a state-trace plot of mixture model parameters across a wide range of manipulations of working memory (from the current paper) and long-term memory (from Miner et al 30 ), with one point per condition. We find that despite the huge number of different ways we vary memory strength, all the points lie on a single line, consistent with only a single parameter being varied; and that this line is extremely well predicted by the 0-free-parameter prediction of TCC.…”

“…We take a fundamentally different view, seeing our measure of memory strength (d') as a measure of signal-to-noise that is likely modulated by many factors, and which has a shared structure not only in working memory, where set size matters so much, but also in long-term memory, which appears to follow fundamentally the same rules of memory confusability and a similar decision process ( Fig. 8; Miner et al 30 ). Notably, we find that while set size modulates memory strength in the current work, there are many other factors that affect memory strength nearly as much.…”

“…TCC accurately captures face identity data, demonstrating its generalizability across diverse stimulus spaces. (D) To show generalization to other memory systems, we fit data on a visual long-term memory continuous report task with colors 30 . N=30 participants performed blocks of memorizing 40 items, and then after a delay, reported the colors of the items using a color wheel.…”