State and trait epistemic curiosity can promote various aspects of cognition and behavior. However, possible age-related changes, a link to surprise, and formal education remain understudied. In three behavioral experiments, with n = 54 in experiment 1, n = 81 in experiment 2, and n = 196 in experiment 3, we could show that state but not trait epistemic curiosity drives long-term memory for relevant information in young and healthy older adults (experiment 1 and 2). This effect was further modulated by surprise in young but less in older adults, which is compatible with predictive coding theories and the notion of a quadratic effect of arousal on cognition that vanishes with age (experiment 2). Finally, in a mediation model, state epistemic curiosity could be predicted directly from trait epistemic curiosity and indirectly via formal education (experiment 3). Our findings specify how state and trait curiosity, together with surprise, promote learning and facilitate knowledge acquisition across the adult lifespan.

Curiosity is a powerful form of intrinsic motivation enabling us to seek and explore novel information, thereby shaping individual development, including educational achievement. While many studies, that will be explained below, gave valuable insights, the lifelong developmental trajectories and possible age-related changes remain unclear. Addressing this important gap may not only have practical implications with regard to healthy aging but it could also advance our conceptual understanding and theoretical models. Here, we specifically focus on epistemic curiosity (EC), i.e. the drive to acquire new knowledge in order to close an "information gap". EC can further be distinguished into state and trait EC, which relates to a rather stable and consistent desire to acquire new knowledge (i.e. trait EC), or a transient and situational intrinsic motivation to acquire new knowledge (i.e. state EC).

The ability to learn and encode novel information into long-term memory (LTM) strongly depends on different types of motivation. Studies from both psychology and cognitive neuroscience have shown that monetary rewards (i.e. extrinsic motivation) and state EC (i.e. intrinsic motivation) promote learning and memory performance via the dopaminergic mesolimbic system and interconnected brain regions. For instance, superior memory performance is being observed when images serve as cues for monetary feedbackand these behavioral effects were linked to hemodynamic activity in the substantia nigra / ventral tegmental area (SN/VTA), nucleus accumbens and hippocampus. Similarly, being in a state of high curiosity, that can be induced by unknown answers to trivia questions or magic tricks, leads to enhanced mesolimbic activity and promotes subsequent memory formation not only for the curios item but also for incidental information. From a conceptual point of view, the positive effects of curiosity (and reward) on LTM can be explained by predictive coding theories and studies demonstrating that "information prediction errors" (i.e. differences in perception and prediction) can enhance learning. Therefore, intrinsic (i.e. curiosity) and extrinsic (i.e. money) types of motivation modulate declarative learning via the mesolimbic system, which points towards common psychological and neurobiological underpinnings.

This latter notion is important since the mesolimbic system typically degenerates during healthy aging with functional consequences for reward processing and the ability to acquire novel information. For instance, age-related degeneration of the dopaminergic midbrain affects neural novelty signals and LTM performance, while iron levels and myelin content in the ventral striatum predict memory performance in the aging brain. Moreover, reward anticipation relates to neural theta oscillations (i.e. 4-8 Hz) and is predicted by dopaminergic midbrain integrity in healthy older adults. In line with this observation, healthy older adults did not show the typical reward anticipation signal in the mesolimbic system, suggesting that impairments in motivational processing can be explained by region specific degenerations. However, the positive effect of state EC on learning and memory performance in healthy older adults appears not to be impaired as compared to their younger counterparts. In fact, two initial studies showed that state EC increased subsequent LTM in both young and older humans with no significant differences between both age groups. While this suggests that learning in older adults still benefits from state EC, it remains unclear how this relates to the previously described degeneration of the mesolimbic system and typical age-related declarative memory impairments.

One possibility is that other factors - in particular surprise about the answer to the curious question - play an important role. This hypothesis is based on predictive coding theories, suggesting a close relationship of novelty encoding, surprise and prediction errors. In fact, prediction errors are typically reduced in older adults possibly due to more accurate generative models of the world (due to larger semantic networks) or the inability to properly detect deviations (due to cognitive impairments). In contrast, sensitivity to surprise appears to be enhanced in older adults as shown in studies on reinforcement learningand facial processing. Moreover, a study in young adults indicates that surprise can have a mediating effect on subsequent memory performance via the rostrolateral prefrontal cortex, which is compatible with a recent framework suggesting a close link between prediction errors, appraisal, curiosity, and exploration (PACE). Therefore, in study 1 (Fig. 1), which was based on previous work, we aimed to replicate the absence of an interaction between age and curiosity on LTM performance; and in experiment 2 (Fig. 1), we further investigated a possible interaction between age, curiosity and surprise on LTM performance.

In contrast to state EC, trait EC is considered a stable personality characteristic that can be further divided into interest (I) and deprivation (D) type EC. While trait I-EC indicates the intrinsic pleasure of learning, trait D-EC is associated with an intense need to resolve knowledge gaps. Both types have been linked to motivation and performance in academic contexts, including goal setting and achievementbut also emotional intelligence, intellectual developmentas well as empathetic and research abilities. Moreover, recent research highlights a bidirectional relationship between curiosity and knowledge, suggesting that curiosity not only fosters academic achievement but can also be shaped by prior learning experiences. This raises the question of how trait and state curiosity are related to long-term developmental outcomes - most notably, formal educational attainment, which represents a key marker of cognitive engagement. To address this issue, experiment 3 explored the relationship between trait EC, state EC, and formal education, and tested whether educational achievements mediate the link between trait and state EC.

With regard to age-related changes, trait curiosity, especially intellectual curiosity, typically declines with ageconsistent with findings from personality research on openness. Interestingly, the association between age and intellectual curiosity appears to be mediated by future time perspective, which is compatible with the socioemotional selectivity theorysuggesting that older adults often prioritize emotionally meaningful over exploratory goals. Age-dependent changes in state curiosity, on the other hand, are context-dependent and influenced by subjective aspects. For instance, despite lower levels in trait curiosity, older adults can exhibit higher state curiosity when the interaction with robots was personally relevant.

While experiment 1 and 2 each included two groups of healthy young and older adults in a laboratory setting (Fig. 1), experiment 3 was based on an online survey with subjects across the adult lifespan. We expected a positive effect of state EC on learning and memory of intentional information (i.e. answers to trivia questions) and incidental information (i.e. faces presented in between questions and answers) with no significant differences between age groups replicating previous findings, experiment 1. Note that positive effects of curiosity on incidental information are rather weak and inconsistentand, therefore, require further attention. We also hypothesized a positive effect of surprise on LTM performance and explored differences between both age groups given the typical structural and functional changes in underlying brain regions, experiment 2, Finally, we expected that state EC could be predicted based on trait EC and educational attainment, and we further investigated their relationship in a mediation analysis, experiment 3.