A quantitative model of ensemble perception as summed activation in feature space

Area of Science:

• Cognitive Science
• Computational Neuroscience
• Psychology

Background:

• Ensemble perception, crucial for summarizing complex visual scenes, lacks robust computational models.
• Existing models do not formally link individual item memory to ensemble representations.

Purpose of the Study:

• To develop and test a computational model of ensemble perception based on the sum of activation signals.
• To formally connect models of individual item memory with ensemble representations.
• To predict interindividual and intraindividual differences in visual memory tasks.

Main Methods:

• Developed a novel computational model where ensemble representations are the global sum of individual item activation signals.
• Designed five experiments comparing the proposed ensemble model against alternatives.
• Utilized performance on individual item visual memory tasks to generate predictions for ensemble tasks.

Main Results:

• The proposed model, based on minimal assumptions, successfully predicted performance on ensemble continuous-report tasks.
• The model accurately captured both interindividual and intraindividual variations in visual memory.
• A formal link was established between models of individual item memory and ensemble perception.

Conclusions:

• The developed model provides a formal account of ensemble perception by unifying individual and ensemble memory processes.
• This approach offers a framework for building and comparing computational models of various memory systems.
• The findings highlight the utility of a top-down modeling approach in cognitive science.