One Cue's Loss Is Another Cue's Gain-Learning Morphophonology Through Unlearning

Area of Science:

• Cognitive Science
• Psycholinguistics
• Computational Linguistics

Background:

• Morphological functions in language present learning challenges due to unclear cue-outcome mappings.
• Traditional linguistic views emphasize cue-outcome co-occurrence for learning, contrasting with error-driven learning theories.
• Error-driven learning posits that cue-outcome contingency, not just contiguity, is vital, involving cue strengthening and unlearning (association reduction).

Purpose of the Study:

• To investigate the consequences of unlearning on cue-outcome associations beyond simple reduction.
• To determine if unlearning a cue's association with an absent outcome facilitates learning of the absent cue-absent outcome relationship.
• To test if negative evidence (unlearning) plays a role in learning morphological patterns.

Main Methods:

• An artificial language learning experiment was conducted with human participants.
• Participants were taught morphophonological patterns with varying cue-outcome structures across blocks.
• One block featured co-occurring cues for two outcomes (plural, diminutive); another had a single cue for one outcome.

Main Results:

• Participants demonstrated improved learning of the absent cue-absent outcome association when unlearning was possible.
• This effect was more pronounced when the unlearned cue was more salient.
• Unlearning occurred even without explicit alternative cues for the absent outcome.

Conclusions:

• Unlearning has significant consequences for cue-outcome associations, facilitating the learning of absent cues.
• Learners actively use negative evidence (unlearning) alongside positive evidence, aligning with domain-general error-driven learning principles.
• The salience of cues influences the effectiveness of unlearning in language acquisition.