Wednesday 29 November 2023

Remembering requires no reliability

This post is by Changsheng Lai (Shanghai Jiao Tong University).


Changsheng Lai

You believe that you locked the door before you left your house, but do you really remember that? Your belief about the past episode might be true, but in what sense is the past episode genuinely remembered rather than being just accurately imagined or veridical confabulated? A popular view, which I refer to as ‘mnemic reliabilism’, suggests that the process of remembering is distinguished by its reliability condition. That is, successful remembering must be produced by a reliable memory process. 

Prominent champions of this view include the simulationist Kourken Michaelian (Michaelian 2016) and the causalist Markus Werning (Werning 2020). Besides, you might also find mnemic reliabilism attractive if you are sympathetic to both the orthodox view that ‘remembering entails knowing’ and the idea that ‘knowledge requires reliability’.

In my recent paper entitled ‘Remembering requires no reliability’, I argue against mnemic reliabilism. I demonstrate that there are cases where past events can be successfully remembered despite the unreliability of the corresponding memory processes. Roughly, my cases can be divided into two types:

The first type of cases focuses on patients with neurodegenerative diseases such as Alzheimer’s disease (AD). There is ample empirical evidence to show that AD severely undermines the reliability of patients’ episodic memory by causing frequent memory distortions. However, AD does not deprive patients of the capability of remembering, at least occasionally, some past events. After all, discussions about ‘true memories’ or ‘remembering’ in AD patients are ubiquitous in the psychological literature. It is also untenable to claim that AD patients cannot successfully remember any past event just because they cannot remember many past events reliably.

The second type of cases illustrates how healthy subjects can remember the past unreliably. Empirical studies have discovered many factors that can render our memory processes unreliable. For instance, divided attention (during the stage of encoding or retrieval) has proven to be able to result in more inaccurate memories than accurate ones (Perez Mata et al. 2002). Likewise, high mental stress during post-encoding consolidation also tends to impair memory performance and increase false memories (Pardilla-Delgado et al. 2016). 

Memory processes involving those reliability-affecting factors tend to end up with false memories; however, they can still sometimes end up with true memories. For example, in Pardilla-Delgado and colleagues’ (2015) DRM-style experiments, while stressed participants remembered more false than true words, there was still a 29% chance that presented words could be accurately remembered. Therefore, an overall unreliable memory process can nevertheless lead to successful remembering.

The upshot is, any satisfactory analysis of remembering should avoid including the reliability condition. Moreover, if successful remembering can still provide defeasible justification for our memory beliefs without requiring reliability, then a reliabilist account of (memory) justification is also questionable. Finally, if reliability-affecting factors are prevalent enough in daily life, then perhaps we ought to reexamine the received view seeing memory as a reliable epistemic source.