Der Einfluss verteilten Lernens auf das kognitive und neuronale System von Feedback-Lernen
Kognitive und systemische Humanneurowissenschaften
Zusammenfassung der Projektergebnisse
The accumulation of positive and negative experiences exerts a strong influence on later decision-making. For example, as adults, when making a choice between an apple and a banana, we can rely on preferences shaped by many experiences spread across years. In most situations, we learn from related experiences that are separated by minutes, hours, days, or even years. Decades of research in animals has shown that repeated pairings of stimuli and rewards across days, or “spaced” training, leads to behaviors that are resistant to forgetting, similar to what people call habits. The neural systems underlying such learning, such as the striatum and midbrain dopamine system, continue to be illuminated using the most advanced techniques, most of which are specific to animal research. However, there is a striking difference between the timeframe of learning studies in animals and in humans. In human studies, events are very closely spaced in time (“massed”), with repeated events only separated by a few seconds. Initial research has shown that learning in these conditions is largely supported by short-term memory, which cannot support learning that lasts across days. There has been remarkably little research in humans that examines how value associations are acquired or maintained past the last learning event. To understand such learning in humans, and the underlying neural systems, we examined the effect of spaced training. These and future studies have the potential to provide insight into maladaptive learning over time, such as in addiction, and in conditions where people fail to seek positive outcomes, as in some mood disorders. In our first experiments, we used a simple reward-based learning task in combination with functional magnetic resonance imaging (fMRI). During spaced learning, participants learned associations between pictures and reward in short online sessions spread across weeks. In contrast, during massed learning, associations were learned during a single short session, mirroring conditions used in most other human studies. After learning, we found that patterns of activity in the medial temporal lobe, including the hippocampus, more strongly separated rewarding and aversive stimuli in the spaced condition. In a follow-up test three weeks later, we found that spaced learning led to striking improvements in the long-term maintenance of associations relative to massed learning. Supporting and extending these findings, we carried out a second study where we manipulated spacing between learning repetitions within a single session. We found that performance during closely-spaced “massed” learning was strongly related to a person’s short-term memory capacity. When we tested memory for the learned associations after a brief delay, we found that performance in the massed learning condition decayed, as expected, while performance in the spaced learning condition actually improved. The improvement in performance for spaced learning is an exciting effect which we speculate could be due to automatic reactivation and strengthening of associations only when they are spaced across time. Together, our studies demonstrate neural substrates of learning reward associations that can effectively guide future decision-making. Further, our findings emphasize the importance of studying learning across time in order to both connect research in humans to neuroscience research in animals, and to translate our findings to understand learning outside the lab. Finally, this work provides a platform for future research aimed at understanding learned associations that are too well-engrained, such as in addiction.
Projektbezogene Publikationen (Auswahl)
-
(2018). Reward learning over weeks versus minutes increases the neural representation of value. Journal of Neuroscience, 38(35): 7649-7666
Wimmer, G.E., Li, J.K., Gorgolewski, K.J., Poldrack, R.A.
-
(2021). Reward learning and working memory: effects of massed versus spaced training and post-learning delay period. Memory & Cognition
Wimmer, G.E., Poldrack, R.A.
