During language acquisition children often produce forms that are unattested in standard adult language. Among these forms, we find commission errors, understood as referring to cases where children overtly pronounce material that is (usually) not realized in the standard adult language. Drawing on corpus data from causative and comparative marking in child French and English, we show that one common class of errors involves redundant, multiple exponence. Investigating the question of what exactly children are getting wrong when producing these errors, we present and compare analyses in Distributed Morphology (DM) and Nanosyntax. We argue that within DM children erroneously neglect specificity differences of exponents upon Vocabulary Insertion, while they mistakenly apply spanning and regular constituent lexicalization simultaneously in Nanosyntax. However, an asymmetry in the data naturally follows from the DM implementation, whereas the Nanosyntax account requires additional assumptions.