Transitional probabilities count more than frequency, but might not be used for memorization.

Area of Science:

• Cognitive Science
• Psychology
• Artificial Intelligence

Background:

• Extracting recurring items from continuous sequences is crucial for learning, exemplified by word segmentation in speech.
• Transitional probabilities (TPs) are proposed statistical mechanisms for item extraction and memory storage.
• Previous research shows conflicting evidence on whether TP sensitivity indicates memorization of recurring items.

Purpose of the Study:

• To investigate the role of transitional probabilities (TPs) in the visual modality for learning recurring items.
• To determine if learners prioritize TPs over chunk frequency when discriminating visual sequences.
• To evaluate the implications of findings for distributional clustering models of sequence learning.

Main Methods:

• Participants were exposed to a continuous sequence of visual items (symbols).
• A discrimination task involved choosing between high-TP (frequent transitions) and low-TP (infrequent transitions) items, some attested and some unattested in the sequence.
• Computational modeling was used to support experimental findings.

Main Results:

• Learners showed greater sensitivity to TPs than to chunk frequency.
• Unattested high-TP items were preferred over attested low-TP items, indicating TPs influence learning.
• Results challenge explanations based solely on item similarity and support the primacy of TPs in sequence learning.

Conclusions:

• Learners weigh within-item TPs more heavily than chunk frequency, even with subtle TP differences.
• Findings challenge distributional clustering mechanisms that rely on analyzing continuous sequences.
• TPs may primarily prepare learners to memorize items, rather than directly facilitating memorization itself.