When default is not default


This project addresses a very timely and highly important question of cognitive and clinical neuroscience studies: What factors influence the reliability of neuroimaging studies? What are the sources of individual variability? Which constraining factors may help predicting the outcome of a certain therapy?

Recent studies have estimated the reproducibility of psychological studies to be 39% or less and indicated a severe limitation of neuroimaging (fMRI) study reliability. Too small sample sizes, low to moderate effect sizes, and only partly understood neurophysiological mechanisms behind the BOLD/fMRI signal make it difficult to generalize results, thereby impeding the impact of highly needed neuroscience studies on theoretical (scientific), methodological, and clinical progress.

The overall objectives of this project are to (i) improve our understanding of the neurophysiological mechanism of the BOLD signal and its sources of variability, to (ii) extend current methods on effective and functional connectivity measures (Connectoms), to (iii) find a solution to the replication crisis by developing new Bayesian, topology-based, and machine-learning based analysis methods as alternative approaches to today’s analysis strategies, and to (iv) induce a paradigm shift from the current focus on an easy to measure but susceptible BOLD signal to the underlying, but (partly) hidden neuronal states that are presumably more stable and reliable.

The project aims to generate new insights into the neurophysiological mechanisms of the BOLD signal, its variability, dependency on endogenous and exogenous parameter, and reliability, and it will advance the research field of basic and clinical neuroimaging by providing new analysis strategies.


Project PI Karsten Specht