Observational Word Learning: Beyond Propose-But-Verify and Associative Bean Counting

Area of Science:

• Cognitive Science
• Developmental Psychology
• Linguistics

Background:

• Learning new words presents ambiguity, with multiple interpretations possible for novel words.
• Cross-situational statistics are proposed as a mechanism to overcome this ambiguity in word learning.

Purpose of the Study:

• To investigate the underlying mechanisms of observational word learning using cross-situational statistics.
• To address unresolved questions about how learners track word-object mappings and resolve ambiguity.

Main Methods:

• Two experiments were conducted with participants learning eight word-object mappings.
• Eye-movements were tracked during the learning process to analyze real-time cognitive processes.
• Trial-by-trial accuracy analyses were performed to assess statistical learning.

Main Results:

• Eye-movement data indicated learners maintain and utilize multiple word hypotheses during naming.
• Continuous statistical accumulation about word/object mappings was observed, beyond initial hypotheses.
• Learning can be hindered by consistent, probabilistic contexts, leading to false associations.
• Factors like foil knowledge, object spatial consistency, and word presentation frequency influence learning.

Conclusions:

• Observational word learning likely involves a blend of gradual statistical learning and rapid cognitive processes.
• Mechanisms such as hypothesis testing, competition, and mutual exclusivity contribute to resolving word learning ambiguity.