Russell Poldrack

Poldrack’s earliest work studied the brain systems involved in nondeclarative memory. His dissertation work examined the relation between stimulus-specific learning and general skill in a motor skill learning task.[3] His first neuroimaging paper demonstrated changes in brain activity associated with learning of mirror-reading skill, showing that it was associated with a shift from activity in parietal regions toward activation in inferior temporal regions.[4] He later showed that learning of classification learning was associated with a tradeoff between activity in the basal ganglia and medial temporal lobe, and proposed that this reflected a competition between declarative and nondeclarative memory systems in learning.[5] In 2006, his group published work showing that this tradeoff between systems is modulated by dual-task interference[6]

Poldrack’s group has also studied the brain systems involved in the inhibition of motor responses. With Adam Aron, Poldrack published two papers that established the role of a circuit involving right prefrontal cortex and the subthalamic nucleus in the inhibition of motor responses.[7][8] They subsequently showed that it was possible to predict individual differences in inhibitory behavior from functional magnetic resonance imaging (fMRI) data using high-dimensional regression machines.[9]

In 2007, Poldrack and colleagues demonstrated that brain activity during decisions under risk exhibited the pattern of gain- and loss-responsiveness predicted by prospect theory.[10] In subsequent work, they found that risky decisions in the Balloon Analog Risk Task could be predicted from fMRI data, and that these decisions were related to a balance of activity between large-scale brain systems involved in value processing and cognitive control respectively.[11]

In 2006, Poldrack published a paper in Trends in Cognitive Sciences that criticized the field for the use of “reverse inference”, in which the presence of activation in a brain region is used to infer the engagement of a specific psychological process.[12] Using a Bayesian analysis, he showed that this form of inference generally provides weak evidence in favor of specific psychological processes. His lab subsequently applied machine learning methods to fMRI data, demonstrating that it is possible to accurately infer mental states in a way that generalizes across individuals.[13]

Poldrack has also been engaged in criticizing the unwarranted use of reverse inference in the media. In 2007 he was part of a group of researchers who published a letter in the New York Times that criticized the use of reverse inference in regard to the US Presidential election.[14] In 2010 he led a group of 45 researchers who published a letter in the New York Times criticizing the use of reverse inference in an Op-Editorial on neuromarketing.[15]

Poldrack’s group has developed a formal ontology for cognitive neuroscience, known as the Cognitive Atlas.[16] In 2009, Poldrack established the OpenfMRI project[17](later rebranded as OpenNeuro) which openly shares complete raw fMRI datasets. He has collaborated with Tal Yarkoni on the development of Neurosynth,[18] an online meta-analytic tool for the neuroimaging literature. In 2014 he established the Stanford Center for Reproducible Neuroscience,[19] which develops tools for reproducible data analysis including fMRIPREP and MRIQC.

In 2012, Poldrack undertook a project to collect brain imaging, behavior, and biological data on himself for an extended period of time. Called the MyConnectome Project, this study lasted 18 months, during which Poldrack was scanned with magnetic resonance imaging more than 100 times.[20][21] Analyses of these data showed that brain connectivity changed over this long period of time within specific brain networks,[22] and also showed that the nature of variability over time within an individual is qualitatively different from variability across individuals.[23] This entire dataset was made openly available to the public for further analysis[24]

Poldrack is a part of the Science of Behavior Change (SOBC) network.[25] His research uses brain imaging to understand the brain systems supporting decision making, executive control, and behavior change. Poldrack's lab also develops informatics tools to help make sense of the growing body of neuroimaging data (including the OpenNeuro and NeuroVault data sharing projects and the Cognitive Atlas[16] ontology) as well as tools to help improve the reproducibility of neuroimaging research (including the Brain Imaging Data Structure and BIDS-Apps projects).[11][26]