The current study investigated the role played by conflict monitoring in a lexical-decision task involving competing word representations, using event-related potentials. We extended the multiple read-out model (Grainger and Jacobs, 1996), a connectionist model of word recognition, to quantify conflict by means of Hopfield Energy, which is defined as the sum of the products of all orthographic word node pair activations within the artificial mental lexicon of this model. With increasing conflict levels in nonwords, a late negativity increased in amplitude (400-600 ms) accompanied by activation of the anterior cingulate cortex and the medial frontal gyrus. The simulated conflict predicted the amplitudes associated with this mediofrontal conflict-monitoring network on an item level, and is consistent with the conflict-monitoring theory.