Statistical learning of two artificial languages presented successively: how conscious?

Area of Science:

• Cognitive Psychology
• Neuroscience
• Artificial Intelligence

Background:

• Statistical learning is often considered an automatic, implicit cognitive process.
• The extent to which learned representations are consciously accessible remains largely unexplored.

Purpose of the Study:

• To investigate if knowledge acquired through statistical learning is available for conscious control.
• To determine if participants can consciously manipulate representations learned from artificial languages.

Main Methods:

• Participants were exposed to two distinct artificial languages differing in transitional probabilities.
• Learning and differentiation of languages were assessed (Experiments 1 & 2).
• The Process-Dissociation Procedure was adapted to test conscious control over acquired knowledge (Experiment 3).

Main Results:

• Both artificial languages were equally learnable and distinguishable by participants.
• Statistical information was effectively used to parse and differentiate between the two languages.
• The acquired knowledge representations were found to be available for conscious manipulation.

Conclusions:

• Statistical learning yields representations that can be consciously controlled.
• This finding challenges the notion that statistical learning is exclusively implicit.
• Conscious awareness plays a role in utilizing learned statistical regularities.