As per Bayesian framework, intelligent agents work to establish causalities forming and testing predictions through observing and interacting with their environment. Empirical evidence suggests that memory serves the purpose of supporting these processes. In contexts where the environmental feedback supports the learned causal structure, the previous knowledge is strengthened (Kim et al., 2020). However, highly predictable environments rarely result in detailed episodic memories (Haeuser & Kray, 2021). However, in cases of prediction violations, encoding of detailed episodes is useful for further inspection for establishing causality. Indeed, there is some behavioral evidence suggesting that environments that lack predictive value are remembered in more detail (Sherman et al., 2020; Shohamy et al., 2022).
In my study, a traditional visual statistical learning paradigm (Fiser & Aslin, 2001) paired with a simple object recognition test is used to further investigate the role of predictability in memory encoding and recall. Receiving operating characteristic analysis will be performed to investigate the contributions of familiarity (a continuous sense of having seen the object before), and recall (a binary aspect of recognition associated with remembered contextual information) to the participants’ memory of the viewed objects. (Eichenbaum et al., 2007). Previous studies have demonstrated that participants can distinguish between previously viewed and new objects in VSL paradigms demonstrating high familiarity following statistical learning in a predictable environment (Fiser & Aslin, 2001; Fiser & Aslin, 2005). In the present study, a condition with a decreasing level of predictability is introduced. Given the interplay between Bayesian principles and memory function, it is hypothesized that in this condition better recall performance will be detected, as prediction violations should result in more contextual detail being encoded.