The verbal working memory (vWM) has been widely studied to understand individual’s cognitive development and learning. Role of subcortical structure in vWM features binding is also well established in literature. However, all such studies make use of intracranial electroencephalogram (iEEG) or fMRI because of their anatomical specificity. Additionally, simultaneous analysis of all subcortical structures, like hippocampus, amygdala, and thalamus, is sparsely present. In this work, subcortical source localization-based dipole activation pattern difference is utilized for detecting vWM stages using surface EEG. The EEG potential is converted into neuroimaging modality using sLORETA method. The EEG data was collected from fifteen healthy subjects for the proposed vWM framework. The localized source current information of targeted subcortical structures is utilized for quantitative comparisons and lateralization analysis to identify the significant subcortical structure during the vWM stages. Source domain performance analysis is compared with sensor domain counterparts in classifying encoding, recall, retrieval, and rest. A consistent improvement is seen in the source domain across all the subjects. In particular, the hippocampus is found to be an active and suitable node for the entire vWM process. Subject variability for memory stages is additionally captured. The findings are relevant in early detection of Alzheimer's disease (AD).