Memory systems do not divide on consciousness: Reinterpreting memory in terms of activation and binding

Abstract

There is a popular hypothesis that performance on implicit and explicit memory tasks reflects 2 distinct memory systems. Explicit memory is said to store those experiences that can be consciously recollected, and implicit memory is said to store experiences and affect subsequent behavior but to be unavailable to conscious awareness. Although this division based on awareness is a useful taxonomy for memory tasks, the authors review the evidence that the unconscious character of implicit memory does not necessitate that it be treated as a separate system of human memory. They also argue that some implicit and explicit memory tasks share the same memory representations and that the important distinction is whether the task (implicit or explicit) requires the formation of a new association. The authors review and critique dissociations from the behavioral, amnesia, and neuroimaging literatures that have been advanced in support of separate explicit and implicit memory systems by highlighting contradictory evidence and by illustrating how the data can be accounted for using a simple computational memory model that assumes the same memory representation for those disparate tasks.

Figure 1

A schematic illustration of how the source of activation confusion model represents studying a word in an experiment. The figure also illustrates the aspect of the representation that is shared by implicit and explicit memory tasks.

Figure 2

A: An illustration of recognition memory performance (measured in d′ units) as a function of whether the encoding font was reinstated (original vs. swapped) and as a function of whether the font was encoded with only one word (low fan) or studied with a dozen words (high fan). The data in this figure are from “Perceptual Match Effects in Direct Tests of Memory: The Role of Contextual Fan,” by L. M. Reder, D. K. Donavos, & M. A. Erickson, 2002, Memory & Cognition, 30, p. 321. Copyright 2002 by the Psychonomic Society. Adapted with permission. B: An illustration of the proportion of remember hits as a function of whether the test font matched the encoding font (original), mismatched but was used with a different word (swapped), or was a novel font not used during encoding (novel). The model predictions are represented as open circles on each bar. The data in this figure are from “Perceptual Match Effects in Direct Tests of Memory: The Role of Contextual Fan,” by L. M. Reder, D. K. Donavos, & M. A. Erickson, 2002, Memory & Cognition, 30, p. 321. Copyright 2002 by the Psychonomic Society. Adapted with permission. Model fit from “Modeling the Role of Perceptual Features in Word Recognition and Fragment Completion,” by M. Cary & L. M. Reder, 2000, paper presented at the 41st annual meeting of the Psychonomic Society, New Orleans, LA.

Figure 3

A: An illustration of a schematic illustration of source of activation confusion’s memory representation of words studied in various unusual fonts. The different number of links fanning out of the font nodes represent the different number of words (and hence encoding contexts) associated with the different fonts. B: A schematic illustration of how the amount of activation sent to the relevant episode and word node is modulated by font fan, when the same font is reinstated at test.

Figure 4

Empirical data (vertical bars) and source of activation confusion model predictions (circles) from Cary and Reder (2000). A: The priming effect when the studied font is reinstated at test (difference in proportion of fragments completed for studied vs. unstudied words) as a function of whether the test font was high or low fan (studied with many words or only one). B: Plot of the proportion of fragments completed for old and new words as a function of whether the font at test matched the font during encoding (for old words) and whether the font was high or low fan or a novel font. The model fit was first reported in “Modeling the Role of Perceptual Features in Word Recognition and Fragment Completion,” by M. Cary & L. M. Reder, 2000, paper presented at the 41st annual meeting of the Psychonomic Society, New Orleans, LA.

Figure 5

Source of activation confusion model fit (open circles) to the young and older adult recognition memory data of (A) Bowles and Poon (1982) and (B) Light, Patterson, Chung, and Healy (2004). Two parameters were estimated for each data set, and R² = .98 for each fit. From “Modeling Age-Related Memory Deficits: A Two-Parameter Solution,” by N. E. G. Buchler & L. M. Reder, 2007, Psychology & Aging, 22, pp. 108 & 114. Copyright 2007 by the American Psychological Association. Reprinted with permission. In Figure 5A, HO = high-frequency old word; HN = high-frequency new word; LO = low-frequency old word; LN = low-frequency new word. In Figure 5B, A–B and A–D are study word pairs, and X, Y, and Z are new words. The error bars represent standard errors of the mean. Conj. = conjunction.

Figure 6

Contextual cuing effect (in ms) in a visual search task (repeated novel displays) early and late in the experiment. Top panel: Normal participants’ performance under saline and midazolam, a drug that simulates amnesia (Park et al., 2004). The data in this figure are from “The Effect of Midazolam on Visual Search: Implications for Understanding Amnesia,” by H. Park, J. J. Quinlan, E. Thornton, & L. M. Reder, 2004, Proceedings of the National Academy of Sciences, USA, 101, p. 17882. Copyright 2004 by the National Academy of Sciences. Adapted with permission. Bottom panel: Hippocampal patients compared with normal control participants (Chun & Phelps, 1999). The data in this figure are from “Memory Deficits for Implicit Contextual Information in Amnesic Participants with Hippocampal Damage,” by M. M. Chun & E. A. Phelps, 1999, Nature Neuroscience, 2, p. 845. Copyright 1999 by Nature America.

Figure 7

Proportion of hits and false alarms for each type of stimulus as a function of drug condition. From “Drug Induced Amnesia Hurts Recognition, But Only for Memories That Can Be Unitized,” by L. M. Reder et al., 2006, Psychological Science, 17, p. 565. Copyright 2006 by Blackwell. Reprinted with permission. Error bars represent standard errors of the mean.

Figure 8

Proportion correct on final test as a function of test list, pair condition, and drug condition. The empirical data are presented in bar graphs and source of activation confusion model predictions are superimposed dots on top of the respective conditions. From “Retrograde Facilitation Under Midazolam: The Role of General and Specific Interference,” by L. M. Reder, J. M. Oates, et al., 2007, Psychonomic Bulletin & Review, 14, p. 264. Copyright 2007 by the Psychonomic Society. Reprinted with permission. Error bars represent standard errors of the mean. Midaz = midazolam; Int = interference.