To characterize the role of the left ventral occipito-temporal cortex (lvOT) during visual word recognition in a quantitatively explicit and testable manner, we propose the lexical categorization model (LCM) according to which lvOT categorizes perceived letter strings into words or non-words. LCM simulations successfully replicate nine benchmark results from human functional brain imaging. Empirically, using functional magnetic resonance imaging and electroencephalography, we demonstrate that quantitative LCM simulations predict lvOT activation and brain activation at an expected time window, respectively. In addition, we found that word-likeness, which is assumed as input to LCM, is represented posterior to lvOT and before the lexical categorization. In contrast, a dichotomous word/non-word contrast, which is the assumed as output of the LCM, could be localized to upstream frontal brain regions. Thus, we propose a ventral-visual-stream processing framework for visual word recognition involving word-likeness extraction followed by lexical categorization, prior to the extraction of word meaning.