# Bias in Structural MRI Correlates of Delay Discounting due to Head Motion

## TL;DR

This study shows that head movement during MRI scans can bias results linking impulsive behavior to brain structure, suggesting future research should account for motion.

## Contribution

The study identifies head motion as a confounding factor in MRI-based associations between delay discounting and cortical thickness.

## Key findings

- Delay discounting correlates with in-scanner head motion, which introduces artifacts in MRI data.
- Head motion biases grey matter thickness estimates in regions previously linked to delay discounting.
- Motion effects are significantly correlated with cortical thinning effects observed in delay discounting studies.

## Abstract

A growing number of studies report that delay discounting, a measure used to evaluate impulsive behaviour, is associated with the regional volume or thickness of cortical grey matter. Using 1096 participants' MRI data from the open Human Connectome Project Young Adults dataset, we show that delay discounting correlates significantly with in‐scanner head motion, a prevalent cause of artifacts in magnetic resonance imaging. Furthermore, head motion was found to bias estimates of grey matter thickness downwards across much of the neocortex, including many of the regions where delay discounting has been associated with reduced grey matter, suggesting these effects may be confounded by motion artifacts. The effects associated with delay discounting were also significantly correlated with the cortical thinning effects associated with motion under a spin permutation null model. In conclusion, we suggest that any future studies investigating structural correlates of delay discounting should correct for motion effects—either during data acquisition or by including a movement covariate when fitting models.

Associations of delay discounting and head motion with cortical thickness. (A) Summary of previous literature findings as having altered cortical thickness with delay discounting behaviour. (B) Results for reductions in cortical thickness correlating with delay discounting and head motion, using parcellated regions. (C) Results for reductions in cortical thickness, analysed vertex‐wise analyses.