The Gow Lab is currently using support vector machine (SVM) classifiers to probe the nature of abstract phonological representation in brain regions associated with lexical representation. We deliberately use SVM rather than stronger classifiers (e.g. tensor flow, multilayer classifiers) in an attempt to identify what contrasts are represented in activation patterns within a specific region of interest, rather than what contrasts could ultimately be derived from those patterns given additional analysis. We apply these analyses to patterns of activation within regions of interest acquired through sourcespace reconstructions of MEG and EEG data. We are currently planning to extend this approach to direct cortical recordings (ECoG) collected in surgical patients undergoing treatment for intractable epilepsy.