Learning concepts when instances never repeat.

Area of Science:

• Cognitive Psychology
• Computational Neuroscience
• Machine Learning

Background:

• Understanding category learning is crucial for cognitive science.
• Investigating how memory representations influence generalization is key.
• Existing models offer different explanations for how humans learn categories.

Purpose of the Study:

• To investigate the impact of instance repetition on category learning and transfer.
• To compare learning and memory performance between repeating and non-repeating training conditions.
• To evaluate the consistency of findings with prototype versus exemplar models of category learning.

Main Methods:

• Three experiments were conducted with human participants.
• Training involved categories with either repeating or non-repeating instances across blocks.
• Transfer was assessed using classification and recognition tests.

Main Results:

• Substantial category learning was observed in all experiments.
• Learning speed was not affected by the non-repeating condition.
• Transfer performance was marginally better in the non-repeating condition.
• Recognition memory for specific instances was high in the repeat condition but absent in the non-repeat condition.

Conclusions:

• Category learning can be effective without instance repetition.
• Findings support prototype models of category learning, especially under non-repetition.
• Exemplar models face challenges explaining transfer without explicit instance memory.