MIT CogNet, The Brain Sciences ConnectionFrom the MIT Press, Link to Online Catalog
SPARC Communities
Subscriber : Stanford University Libraries » LOG IN

space

Powered By Google 
Advanced Search

 

A Single Subject Statistical Approach to Modeling Neural Interactions in Verbal Working Memory.

 C. Marshuetz, H. Ombao, T. A. Salthouse and J. Jonides
  
 

Abstract:
The n-Back task, in which verbal working memory (WM) load is varied parametrically, has been used to examine the neural substrates of verbal WM (Braver et al., 1996, Jonides, et al., 1996, Cohen et al., 1997). By examining regional activation x memory load interactions, much has been learned about probable functions of those regions. For example, monotonic increases in activation with memory load have been observed in areas hypothesized to subserve storage operations. While such empirical work suggests the function of observed verbal WM areas, the evidence for those putative functions is indirect. Structural equation modeling, a theory-driven approach to modeling interrelations of variables, has typically been used to model data at the level of groups of subjects. Here, fMRI provided multiple observations of each subject's activation during performance of the n-Back task, allowing each subject's data to be modeled individually. A theoretical model was constructed, using latent variables to represent neural processing units. For example, hypothesized language areas loaded together, forming a "rehearsal" construct. This model's plausibility, which makes explicit the role of regional activation, was examined with data from four individual subjects. This analytical procedure appears promising because the large number of observations available in fMRI provides substantial power to test theoretically interesting models of the interrelations of neural regions.

 
 


© 2010 The MIT Press
MIT Logo