Smith's 1973 model of articulatory phonological development between the ages of two and four is re-examined in an attempt to develop a model that includes the possibility of both perceptual and articulatory learning. Smith's data, regarding phonological transformations of words after rules established by his infant son's pronunciation of the words "puzzle,""puddle," and "pickle," are re-analyzed in support of two hypotheses associated with a perception model. The re-analysis shows that the child does have, in some cases, a unique, non-adult underlying representation based on a misperception of the adult form. This finding is interpreted as evidence for a model that recognizes two types of phonological rules (perceptual-encoding rules and output rules). Several properties that distinguish the two rule types are identified. (Author/JB)