Misremembering

Abstract

The Archival and Constructive views of memory offer contrasting characterizations of remembering and its relation to memory errors. I evaluate the descriptive adequacy of each by offering a close analysis of one of the most prominent experimental techniques by which memory errors are elicited—the Deese-Roediger-McDermott (DRM) paradigm (Deese, 1959; Roediger & McDermott, 1995). Explaining the DRM effect requires appreciating it as a distinct form of memory error, which I refer to as misremembering. Misremembering is a memory error that relies on successful retention of the targeted event. It differs from both successful remembering and from confabulation errors, where the representation produced is wholly inaccurate. As I show, neither the Archival nor the Constructive View can account for the DRM effect because they are insensitive to misremembering’s unique explanatory demands. Fortunately, the explanatory limitations of the Archival and Constructive Views are complementary. This suggests a way forward. Explaining misremembering—including how it differs from both successful remembering and confabulation—requires a hybrid theory of memory, combining the Archival commitment to discrete retention with the Constructive approach to retrieval. I conclude the paper with the beginning sketches of such an account.

Keywords:

• Constructive memory
• DRM effect
• memory
• memory errors

Acknowledgements

Many thanks to Lauren Ashwell, Carl F. Craver, Corey J. Maley, Karen Neander, Katherine Rickus, Henry L. Roediger, and Shannon Spaulding for helpful comments on drafts of this article.

Notes

¹ For more details about the DRM paradigm, see Robins (2014).

² When stimuli are lists of related words, backward associative strength is a strong predictor of the magnitude of false recognition (Cahn, McRae, & Katz, 2011).

³ Recall: Brainerd, Payne, Wright, and Reyna (2003). Pictures: Koustaal (2006). Faces: Homa et al., (2001). Dot arrays: Nosofsky (1991). The magnitude of the misremembering effect differs as a function of stimuli, which may warrant positing distinct mechanisms (e.g., Koutstaal & Schacter, 1997).

⁴ Children below the age of nine or ten are rarely susceptible to the DRM effect (e.g., Sugrue, Strange, & Hayne, 2009). I do not have the space to discuss the developmental trajectory of this error here, but doing so is an important further test of any theory’s descriptive adequacy.

⁵ I do not intend to endorse any particular theory of perception, only the widely accepted characterization of perceptual errors (Fish, 2009).

⁶ I am grateful to an anonymous reviewer for this suggestion. In the future, it will be of interest to explore this issue and the errors near that fall near the alleged boundary. For now it is enough that there are clear cases on either side.

⁷ In referring to this view as archival, I am following Campbell (2006).

⁸ This view of retrieval is illustrated clearly in both Atkinson and Shiffrin’s (1968) functional depiction of short and long-term memory and Anderson’s ACT-R model of cognitive processing (Anderson, Bothell, Lebiere, & Matessa, 1998).

⁹ These four forms of error are intended to reflect the types of malfunction identified in section 3.3. This list may not be exhaustive.

¹⁰ The effect is well replicated (Gallo, 2006).

¹¹ Support for Constructivism can also be found amongst philosophers who support theories of the mind that push back against traditional, faculty-based views of cognition and computational models of cognitive architecture (e.g., Bechtel & Abrahamsen, 2002; Dreyfus, 2002).

¹² Gist-based views are also popular amongst memory scientists—see, for example, Brainerd and Reyna (2005) and Klein (2013).

¹³ Some call this construction, others reconstruction, depending on whether they care to emphasize the point that the initial encoding during perception is also constructive.

¹⁴ According to Fuzzy Trace Theory, there are two traces that are formed as the result of any given experience: a verbatim trace and a gist trace (Brainerd & Reyna, 2005). The verbatim trace—the detailed record of a past experience—decays rapidly and thus fails to play a role in most instances of remembering.

¹⁵ The Archivalist is not committed to the claim that everything is retained in memory, only that the information that is retained can be retrieved.

¹⁶ This view of retrieval is reflected in the methodological design of many memory experiments. Consider, for example, Ebbinghaus’ (1913) diligent records of how many learning trials were required to master a list of syllables, or Craik and Tulving’s (1975) exploration of which learning strategies lead to the best recall.

¹⁷ Another error—relearning—occurs when the second condition is met, but not the first. I have not discussed relearning in this paper, but it plays an important role in the Causal Theory of Memory (Bernecker, 2010; Martin & Deutscher, 1966). It is a virtue of the hybrid approach I propose that it can accommodate this error as well.

¹⁸ I am grateful to two anonymous reviewers for suggesting this opportunity for interface between these research programs.