What do differences between multi-voxel and univariate analysis mean? How subject-, voxel-, and trial-level variance impact fMRI analysis

Abstract

Multi-voxel pattern analysis (MVPA) has led to major changes in how fMRI data are analyzed and interpreted. Many studies now report both MVPA results and results from standard univariate voxel-wise analysis, often with the goal of drawing different conclusions from each. Because MVPA results can be sensitive to latent multidimensional representations and processes whereas univariate voxel-wise analysis cannot, one conclusion that is often drawn when MVPA and univariate results differ is that the activation patterns underlying MVPA results contain a multidimensional code. In the current study, we conducted simulations to formally test this assumption. Our findings reveal that MVPA tests are sensitive to the magnitude of voxel-level variability in the effect of a condition within subjects, even when the same linear relationship is coded in all voxels. We also find that MVPA is insensitive to subject-level variability in mean activation across an ROI, which is the primary variance component of interest in many standard univariate tests. Together, these results illustrate that differences between MVPA and univariate tests do not afford conclusions about the nature or dimensionality of the neural code. Instead, targeted tests of the informational content and/or dimensionality of activation patterns are critical for drawing strong conclusions about the representational codes that are indicated by significant MVPA results.

Briefly Cuing Memories Leads to Suppression of Their Neural Representations

“…Other possibilities are that features will be weakened at random; that weakening reflects the introduction of noise, such that memories in the short-cue condition will show a reduced signal-to-noise ratio; or that memories will become more schematic (i.e., showing a greater influence of shared vs item-specific features). Last, recent work suggests that univariate changes can contaminate similarity measures (Davis et al, 2014), so we examined whether such a confound could account for the observed pattern of results.…”

Progress and challenges in probing the human brain

“…The legitimate enthusiasm about these methods is tempered by lingering questions regarding the interpretation of multivariate analyses 31,32 . In addition, recent work combining electrophysiology and fMRI in non-human primates has demonstrated that the sensitivity of MVPA is limited by the spatial characteristics of the neuronal representations that code for particular features, such that some kinds of neuronal patterns may be more difficult to decode using MVPA than others 33 .…”

Noninvasive Functional and Anatomical Imaging of the Human Medial Temporal Lobe

“…In particular, RSA analyses of fMRI encoding data suggest that, in some contexts, representational stability may be beneficial to later remembering (Xue et al 2010;Ward et al 2013; but see Wagner et al 2000), demonstrating that greater pattern similarity of an item's neural representations across multiple encoding trials predicts better subsequent memory for the item. Although some questions remain (Xue et al 2013;Davis et al 2014), researchers are now positioned to measure, at the individual trial level and within an individual human brain, the large-scale distributed neural representations that underlie important aspects of memory behavior.…”