The mechanisms underlying crowding are analyzed in terms of explicit neural processes mediating its perceptual characteristics as originally described by W. Korte (1923). A full understanding of crowding in letter recognition requires a detailed conceptualization of the process of recognition among large numbers of alternatives. The observed masking properties suggest the operation of recursive inhibition from V3 to V1 as a component of the crowding effect. The plausibility of six accounts of the neural basis of crowding (the template matching, feature integrator, attentional feature conjunction, propositional enumeration, attentional tracking, and relaxation network concepts) is then assessed in relation to the task of encoding the spatial structure of the letter forms. We conclude that the relaxation network approach is the most plausible hypothesis to account for the full-spectrum letter recognition performance.