The current study constituted the first attempt to generate repertoires of relational responding in accordance with opposite, as generalized operant behavior, when they are found to be absent in young children. Three children, aged between 4 and 6 years, participated in the study. A basic problem-solving task was adopted from previous research to test and train patterns of relational responding in accordance with opposite. This task involved presenting a child with various numbers of identically-sized paper coins and providing the following instructions, for example: “This coin buys many (or few) sweets, and is opposite to this coin, which would you take to buy as many sweets as possible”? All three participants failed to pass baseline tests for specific patterns of relational responding in accordance with opposite. Various interventions, including training and testing across different stimulus sets and across different numbers of sets, were then successfully used with all participants to establish these relational responses as well as increasingly complex patterns of opposite responding. Generalization tests also demonstrated that the relational responding generalized to novel stimuli and experimenters. In addition, the use of a non-contingent reinforcement condition for one participant, during which no improvement was made, together with contingency reversals for all three participants, indicated that the trained and tested opposite responding may be considered a form of generalized operant behavior. These findings support previous research and lend positive support to Relational Frame Theory’s approach to derived relational responding, and to the functional analysis of human language and cognition. Alternative interpretations of the data are also considered.