Prior experience modifies acquisition trajectories via response-strategy sampling.

Area of Science:

• Behavioral neuroscience
• Cognitive psychology
• Machine learning

Background:

• Signal detection parameters are crucial for understanding decision-making under uncertainty.
• The evolution of these parameters during learning remains underexplored.
• Prior experience can significantly influence learning trajectories and strategy adoption.

Purpose of the Study:

• To investigate how signal detection parameters change during the acquisition of an auditory task.
• To examine the impact of differential prior experience on learning dynamics.
• To propose and validate a computational model of learning in this task.

Main Methods:

• Training three groups of mice (naïve, CCRT, CDT) in an auditory signal detection task.
• Analyzing acquisition data using a two-period model (pre-solution and solution).
• Developing and simulating a computational learning model.

Main Results:

• The pre-solution period involved biased response strategy sampling, while the solution period involved conditional responding.
• Prior experience (CCRT and CDT) shortened the pre-solution period compared to naïve mice.
• CDT and naïve mice showed faster acquisition of discriminability and developed criterion biases compared to CCRT mice.

Conclusions:

• Prior experience modulates the length and strategy composition of the learning acquisition phase.
• A proposed learning model, incorporating strategy sampling and conditional responding, can explain the observed behavioral data.
• The model highlights the role of inference and generalization in learning complex tasks.