Imaging Research Center, The University of Texas at Austin, TX 78712, USA.
Neuropsychologia. 2012 Mar;50(4):544-52. doi: 10.1016/j.neuropsychologia.2011.11.007. Epub 2011 Nov 11.
Multivariate pattern analysis (MVPA) has recently received increasing attention in functional neuroimaging due to its ability to decode mental states from fMRI signals. However, questions remain regarding both the empirical and conceptual relationships between results from MVPA and standard univariate analyses. In the current study, whole-brain univariate and searchlight MVPAs of parametric manipulations of monetary gain and loss in a decision making task (Tom et al., 2007) were compared to identify the differences in the results across these methods and the implications for understanding the underlying mental processes. The MVPA and univariate results did identify some overlapping regions in whole brain analyses. However, an analysis of consistency revealed that in many regions the effect size estimates obtained from MVPA and univariate analysis were uncorrelated. Moreover, comparison of sensitivity showed a general trend towards greater sensitivity to task manipulations by MVPA compared to univariate analysis. These results demonstrate that MVPA methods may provide a different view of the functional organization of mental processing compared to univariate analysis, wherein MVPA is more sensitive to distributed coding of information whereas univariate analysis is more sensitive to global engagement in ongoing tasks. The results also highlight the need for better ways to integrate these methods.
多元模式分析（MVPA）最近在功能神经影像学中受到越来越多的关注，因为它能够从 fMRI 信号中解码心理状态。然而，关于 MVPA 和标准单变量分析的结果之间的经验和概念关系仍然存在问题。在本研究中，对决策任务中货币收益和损失的参数操作的全脑单变量和搜索灯 MVPA（Tom 等人，2007）进行了比较，以确定这些方法的结果差异及其对理解潜在心理过程的影响。MVPA 和单变量结果确实在全脑分析中确定了一些重叠区域。然而，一致性分析表明，在许多区域中，MVPA 和单变量分析获得的效应大小估计值是不相关的。此外，敏感性比较表明，与单变量分析相比，MVPA 对任务操作的敏感性通常更高。这些结果表明，MVPA 方法可能提供了一种与单变量分析不同的心理加工功能组织视图，其中 MVPA 对信息的分布式编码更敏感，而单变量分析对正在进行的任务的整体参与更敏感。结果还强调需要更好的方法来整合这些方法。
