Extracted data for each ROI was then normalized to a mean of zero and standard deviation of one. Effective connectivity of regions activated during shift and no-shift paradigms was assessed using path analysis within a structural equation modeling framework (AMOS version 19.0, SPSS, IBM). While the typical strategy for SEM is to implement

a priori hypotheses to fully constrain the SEM models as seen in the Tourville 2008 study, EGFR inhibitor this can be misleading. Instead, we chose to employ an approach with minimal a priori constraint which allowed for the production of data driven models for vocalization (Laird et al., 2008 and Hastie et al., 2009). While the results from Tourville’s stacked model are important, our goal differed from the Tourville study. Our goal was to provide a data driven model that reduced JQ1 solubility dmso bias introduced by a priori models. Bias is the result of a fully constrained model requiring assumptions to be made which can potentially limit the identification of vital connections within a system. Due to our data driven approach, we were able to examine key pathways that may not have been identified a priori. Furthermore, our model started with a full comprehensive model that included all possible paths

from our point of origin. To establish a starting connection for each structural equation model, we imposed a prior assumption identifying superior temporal gyrus as the initial region receiving auditory input. The use of STG as the initial region of input is supported by research indicating that information from an auditory stimulus reaches STG approximately 12–17 ms from the stimulus onset (Inui et al., 2006 and Steinschneider et al., 1999). Thus, it was hypothesized that STG interacts with one or more of the remaining variables/regions. Paths connecting the STG to all other

regions were established and a specification search was employed to determine the best combination of connected regions following the guidelines of Burnham and Anderson (2002). Specification search allows for multiple candidate models to be tested using optional unidirectional path loadings. The Browne–Cudeck criterion value (BCC) is an information-theoretic index that represents the predictive fit index and is used to select among PRKACG competing models fit to the same data (Schumacker & Lomax, 2010, p. 230). In this analysis, the model with the lowest BCC value was selected as the model that best represented the data (Laird et al., 2008). The next sets of candidate pathways were identified in an exploratory manner through the use of modification indices (MI). Paths with the highest MI were chosen as the next likely paths. The new paths were added to the model, and an additional specification search was conducted. This search procedure continued in an iterative manner until a root mean square error of approximation (RMSEA) value of less than .