What's statistical about learning? Insights from modelling statistical learning as a set of memory processes

Area of Science:

• Cognitive Science
• Psychology
• Computational Neuroscience

Background:

• Statistical learning is investigated across diverse tasks like word segmentation and category learning.
• Existing approaches often fail to unify learning mechanisms across different tasks.
• This highlights a need for a more comprehensive theoretical framework.

Purpose of the Study:

• To present a formal computational approach for a unifying framework of statistical learning.
• To explain learning across a wide variety of tasks using a single model.
• To challenge the notion that statistical learning solely relies on explicit statistical computations.

Main Methods:

• Proposes a computational framework rooted in inherent memory system processes.
• Integrates concepts such as activation, interference, information integration, and forgetting.
• Focuses on the extraction of input elements into memory traces and their subsequent integration.

Main Results:

• Suggests statistical learning arises from fundamental memory operations.
• Demonstrates how memory processes can account for learning across different tasks.
• Highlights the role of memory trace integration in emphasizing consistent information.

Conclusions:

• Statistical learning is not solely based on explicit statistical computation.
• A unified framework based on memory systems can explain diverse statistical learning phenomena.
• This approach offers a new perspective on the cognitive mechanisms underlying statistical learning.