How reliable is fMRI, and does it actually matter? by Dr Matt Wall

About the speaker

Dr Matt Wall completed his PhD in Cambridge, then did post-doctoral positions at Royal Holloway and UCL before selling out and moving into the private sector. He currently works for Invicro, a global imaging company that provides research services for the pharmaceutical and biotechnology industries. His research interests have meandered around somewhat, taking in cognitive psychology, visual psychophysics, pain, psychopharmacology, psychedelics, addiction, neuroendocrinology, and methods development, but all focussed around the use of fMRI. His current collaborative projects involve giving teenagers cannabis and showing women pornography, as well as a number of commercial studies, and ongoing methodological work.

About the talk

fMRI is an established workhorse method in modern cognitive neuroscience but relatively little attention has been paid to characterising its basic features, in particular, the closely related concepts of reliability, replicability, and reproducibility. I’ll discuss some recent work we’ve done examining test-retest reliability metrics on a range of different cognitive and sensory task paradigms. We’ve also examined the effects of different de-noising procedures on these reliability metrics in the same set of data, as well as in resting-state fMRI data. There has been a great deal of discussion in recent years on improving power, reliability, and replicability in neuroscience, but which of these is the most important, and which are we actually able to change? Does reliability actually matter all that much, and if it does, how do we get people to care about it?