Reproducible Cognitive Neuroscience

Current projects

Multiverse and Computational Model Comparison in Developmental Reward Learning How do reward magnitude and outcome variance shape trial-and-error learning from childhood into adulthood, and how much do our answers depend on the analytic choices we make along the way? This project revisits an fMRI dataset of 73 participants aged 8 to 30 who completed a multi-armed bandit task crossing expected value with outcome variance. Rather than adding one more study to a literature with conflicting developmental trajectories, we hold the task fixed and ask where conclusions actually come from. At one level, competing reinforcement learning models attribute the same behavioral age differences to different cognitive mechanisms, such as distorted value representation versus altered learning rates, and behavior alone cannot adjudicate between them. At another, hundreds of defensible neuroimaging pipelines yield varying pictures of reward-related activity in ventral striatum, vmPFC, and anterior insula. Mapping variability across both levels turns a robustness audit into a scientific question: which developmental effects are stable, which are fragile, and what properties of the data predict the difference.

Previous projects

Experiment Factory
Experiment Factory is an “an open source framework for the development and deployment of web-based experiments.” We developed this platform to run our large scale multiwave online experiments. It consists of a Python application and a repository of behavioral experiments and surveys coded in javascript using jsPsych. As we hoped for, it has been used by other researchers for their experiments as well.