Unsupervised learning of complex associations in an animal model

Area of Science:

• Cognitive Science
• Animal Behavior
• Comparative Psychology

Background:

• Supervised learning relies on explicit feedback, while unsupervised learning stems from statistical co-occurrence.
• Understanding non-human animal learning mechanisms provides insights into cognitive evolution.

Purpose of the Study:

• To investigate if pigeons can acquire associations between visual images through unsupervised learning via mere pairing.
• To compare pigeon learning mechanisms with statistical learning in human infants.

Main Methods:

• Pigeons were initially trained on an unsupervised task involving paired visual images.
• Learning was assessed using testing trials where pigeons chose between paired and unpaired image tokens.
• A subsequent supervised training phase examined learning facilitation based on prior pairing.

Main Results:

• Pigeons showed preferential choice for previously unpaired tokens, indicating learning.
• Prior pairing of object and token facilitated learning in the subsequent supervised phase.

Conclusions:

• Pigeons exhibit unsupervised learning, evidenced by association formation through statistical co-occurrence.
• The findings suggest a significant parallel between cognitive processes in pigeons and statistical learning in human infants.
• This research highlights potential shared evolutionary roots for learning mechanisms across species.