Variability in behavior that cognitive models do not explain can be linked to neuroimaging data

It is known that behavior is substantially variable even across nearly identical situations. Many cognitive models are not able to explain this intraindividual variability but focus on explaining interindividual differences captured in model parameters. In sequential sampling models of decision making, for instance, one single threshold parameter value is estimated for every person to quantify how much evidence must be accumulated for committing to a choice. However, this threshold may vary across trials even within subjects and experimental conditions. Neuroimaging tools such as functional magnetic resonance imaging (fMRI) or electroencephalography (EEG) can reveal moment-to-moment fluctuations in the neural system that are likely to contribute to fluctuations in behavior. We propose that neural and behavioral variability could be linked to each other by assuming and estimating trial-by-trial variability in model parameters. To illustrate our proposal, we first highlight recent studies in model-based cognitive neuroscience that have gone beyond correlating model predictions with neuroimaging data. These studies made use of variance in behavior that remained unexplained by cognitive modeling but could be linked to specific fMRI or EEG signals. Second, we specify in a tutorial a novel and efficient approach, how to extract such variance and to apply it to neuroimaging data. Our proposal shows how the variability in behavior and the neural system can provide a fruitful source of theory development in cognitive neuroscience.