Frequency effects in human category learning.

Area of Science:

• Cognitive Psychology
• Computational Neuroscience
• Machine Learning

Background:

• Prototype and exemplar models are key theories in human category learning.
• Category frequency is a known factor influencing learning and decision-making.
• Understanding computational mechanisms of categorization is crucial.

Purpose of the Study:

• Investigate prototype vs. exemplar models of category learning.
• Examine the impact of category frequency on categorization decisions.
• Determine the role of recency in computational models of learning.

Main Methods:

• Utilized baseline and recency-weighted prototype and exemplar models.
• Employed stimuli from bivariate normal distributions across four experiments.
• Manipulated category frequency during training and tested novel stimuli classification.

Main Results:

• A significant frequency effect was observed: participants preferred categorizing items into more frequent categories.
• This preference extended to stimuli outside the trained region.
• The recency-weighted generalized context model exemplar model best fit the data and predicted classification patterns.

Conclusions:

• Relative category frequency significantly influences human categorization of novel items.
• Categorization decisions are based on recency-weighted, accumulative evidence, not averaged evidence.
• Prototype models failed to explain observed frequency effects, unlike exemplar models.