Abstract
It has been demonstrated (Brady, Konkle, Gill, Oliva & Alvarez, 2013) that the limit of fidelity for feature representations is stable across memory subsystems. However, it is known that long-term memory is different from working memory when it comes to temporal processes (Cowan, 2008). Specifically, consolidation processes (Born & Wilhelm, 2012), as well as the act of retrieval (Karpicke & Roediger, 2007), influence contents of long-term memory. We examined the influence of time and retrieval on feature representations in long-term memory using the continuous report paradigm (Wilken & Ma, 2004; Zhang & Luck, 2008). The experiment consisted of three stages. During the first stage, all participants were presented with 330 images of real-world objects and required to remember each object and its color. During the second stage, immediately after the study, all the participants were presented with grayscale objects. The experimental group had to choose each object’s color on a color wheel, whereas the control group had to distinguish between old and new objects in a 2AFC task. In the third stage, which took place 24–26 hours after the second, all the participants had to choose the color of each object on a color wheel. The responses given by subjects were analyzed using mixture models of a uniform and a von Mises distribution (Zhang & Luck, 2008). The obtained results showed that the fidelity in long-term memory was not influenced by time or retrieval and support the conclusion by Brady et al. (2013) that the fidelity limit in memory subsystems is caused by a higher-order limitation. Additionally, our data show that repeated retrieval in long-term memory leads to occurrence of false memories — specifically, real memories of the presented color are replaced with the participants’ wrong answers during first retrieval.