Abstract This paper outlines a detailed theory of relations between word perception and production, and reports five experiments testing prediction, of the theory for the verbal transformation effect (VTE), the perceptual changes that occur when an acoustically presented word is repeated via computer loop for prolonged periods. Within this interactive activation theory, VTEs occur to the extent that lexical nodes determining the perception of speech are satiated due to repeated activation. However, perception shares lexical nodes with production in the theory, so that when subjects concurrently produce (shadow) a recurring word in VTE experiments, lexical nodes for production provide the basis for concurrent perception, and not the acoustic input. However, satiation of lexical nodes is offset by aspects of the production process, so that few VTEs accompany concurrent production, even silent production (mouthing and inner speech). Findings reported here supported this Node Structure theory account of VTEs, but challenged both Parallel Distributed Processing theories (that postulate inhibition between lexical nodes), and corollary discharge theories (that postulate inhibition at the phonetic level between production and perception nodes).