How to Argue for Memory’s Epistemic Generativity
Uku Tooming (University of Tartu) & Kengo Miyazono (Hokkaido University)
Epistemic preservationism of memory holds that memory’s epistemic force —its capacity to rationalize or justify beliefs that are formed on its basis— is derived from something else and memory itself simply transmits justification originally provided by another source, such as perception or reasoning. For instance, if Laura remembers that there was a fight at the pub yesterday, she is justified to believe it by memory. But according to epistemic preservationism, her memory just transmits the justification that comes from Laura’s actually having witnessed the fight, that is, it transmits justification by perception.
Epistemic generationism, on the other hand, asserts that memory can generate justification on its own, at the time of retrieval, rather than merely preserving justification from prior sources. In recent years, different forms of generationism have been defended by various authors, such as Jennifer Lackey, Kourken Michaelian, Jordi Fernández, Ali Boyle, and ourselves. In this post, we will introduce another way to defend epistemic generationism by arguing that in certain cases, memory can provide justification that does not depend on any earlier non-mnemonic source.
To clarify the kind of generationism at issue in this blog post, when we talk about ‘justification’ we have in mind propositional justification. An agent has propositional justification for a proposition when it is rationally permissible for her to believe it, which in turn requires her to possess adequate reasons in favor of the proposition in question. By possessing adequate reasons, i.e., by being aware of them and of how to cognitively use them, it is reasonable for her to believe the proposition in light of those reasons.
Our version of epistemic generationism claims that memory can put one in possession of adequate reasons in a way that does not depend on other, non-mnemonic sources.
While previous defenses of epistemic generationism have focused on cases in which memory provides justification for forming a belief while there was no earlier justification provided by a non-mnemonic source, our argument is slightly different.
We propose that memory can be considered a generative source of justification for forming a belief about a proposition even when the agent already had justification to believe that proposition by a different source. The crucial factor is not whether there was justification provided by some non-mnemonic source but whether mnemonic justification is dependent on that earlier source.
As an analogy, consider perception as a paradigmatic example of a generative source of justification. Perception is a generative source of justification because perceptual justification is not dependent on prior sources. For instance, a person may be informed by a reliable witness that there is a rat in the kitchen and thereby have testimonial justification for forming the respective belief. If she afterwards goes to the kitchen and sees the rat by herself, she is justified by perception to form the belief that there is a rat in the kitchen.
In such a situation, the earlier testimonial justification doesn't undermine the fact that perception is generating new justification. This is because perceptual justification is independent of testimonial justification. Similarly, memory can generate justification regardless of whether there was prior justification by another source, as long as mnemonic justification is independent of other sources of justification.
What does it take for mnemonic justification to be independent of other sources of justification? Our proposal is that it requires at least “counterfactual independence” by which we mean that even if a person hadn’t been justified by another source earlier, memory would still justify the belief. If this counterfactual doesn’t hold, then memory is simply preserving earlier justification, supporting preservationism. But if it does hold, memory serves as a generative source of justification.
Are there cases of epistemic justification by memory that satisfy the counterfactual independence requirement? We think there are. In particular, memory consolidation processes can improve an agent’s epistemic position in such a way that the justification acquired through those processes is counterfactually independent of justification provided by other sources.
Memory consolidation refers to neural processes through which recently formed neural representations are selectively strengthened and restructured. As a result, newly acquired information becomes more stable and integrated with a person’s existing knowledge. Over time, memory consolidation can result in the schematization of memories, where specific details may fade, but overarching patterns and structures become emphasized. This process enables individuals to extract general principles from experiences, which can be applied to new situations.
Take sleep-dependent memory consolidation which has been shown to improve cognitive performance in various tasks, such as visual learning and decision-making. For instance, Wagner and colleagues conducted a study where participants were tasked with transforming an 8-digit sequence according to specific rules. There was a hidden rule, which wasn’t revealed to participants.
After a period of sleep, participants were twice as likely to discover the hidden rule compared to those who didn’t sleep. Consolidation processes enabled participants to restructure their representation of the task, making it possible to grasp the hidden rule. This suggests that memory consolidation during sleep was essential at least for some participants to put them in the position to reasonably form the belief that the task was governed by that rule.
Another example comes from Pace-Schott and colleagues, who investigated how sleep affects performance in the Iowa Gambling Task. Participants had to determine which decks of cards were advantageous and which were not. Those in the Sleep group not only performed better but were also more likely to explain their reasoning for preferring the advantageous decks.
This suggests that consolidation processes were instrumental in improving their comprehension of the structure of the task and putting them in the position to articulate the reasons for their preference. The consolidation process during sleep were thus crucial for those participants to justify their beliefs about which decks were more advantageous.
These studies highlight that sleep-based memory consolidation can change how people process and understand information, especially when it comes to gaining new insights. For participants who formed new beliefs about their tasks and about the rules that governed them only after sleep, it seems that memory consolidation enabled them to justify these beliefs. In both cases, sleep was instrumental in helping participants form well-grounded beliefs about the structure of the tasks they were given.
Most significantly for present purposes, it is also plausible that the memory justification that was provided in those cases can be a generative form of justification because it is not counterfactually dependent on some earlier non-mnemonic justification. To illustrate this further, consider a hypothetical scenario involving Jane, who is participating in the Iowa Gambling Task.
Before sleep-dependent memory consolidation processes kicking in, Jane doesn’t fully understand which decks are advantageous. However, after sleeping she realizes that she was performing a task in which such-and-such decks (say, C and D) were the advantageous ones. Her justification for her belief is based on memory, making it an example of mnemonic justification.
Now, it’s possible that Jane could have had some form of non-mnemonic justification for this belief before sleep, perhaps through reasoning or inference. However, our key point is that even if Jane had some non-mnemonic justification earlier, her mnemonic justification doesn’t necessarily depend on that earlier reasoning. Jane could still be justified purely through memory, without relying on any previous justification.
Why? Remember that we understood having (propositional) justification in terms of the possession of adequate reasons. In Jane’s case, the following counterfactual holds: even if she hadn’t possessed an adequate reason to believe the rule about the advantageous decks before sleep in virtue of a non-mnemonic source, such as reasoning, she would have still possessed that reason at after sleep-dependent memory consolidation, with memory serving as the source of justification. In the Iowa Gambling Task, the reason could be expressed as follows: “Although one wins less from decks C and D, one also loses less over time.”
Since possessing a reason means being in a position to recognize or become aware of it, Jane could have come to understand this through memory after sleep, even if she hadn’t been aware of it through reasoning before sleep.
This counterfactual is plausible because of empirical reasons. As Pace-Schott and colleagues showed in their study, those participants who initially performed poorly on the Iowa Gambling Task benefitted the most from sleep-dependent consolidation. Those who struggled early on still showed significant improvement in rule understanding after sleep, suggesting that memory consolidation can enhance understanding independently of prior reasoning ability.
Thus, it is plausible that Jane could have acquired the reason to believe that she was performing the task in which decks C and D were advantageous and, therefore, her justification for forming beliefs about which decks were more advantageous is not counterfactually dependent on an earlier source of justification.
In summary, memory consolidation can provide generative justification by allowing people to form justified beliefs where justification for those beliefs is counterfactually independent of earlier non-mnemonic justification. This supports the claim that memory, particularly through processes like consolidation, can generate justification in a way that aligns with epistemic generationism. Memory, therefore, has epistemic powers of its own.
