The relational evaluation procedure (REP) is seen by proponents of relational frame theory (RFT) as a useful means by which to assess and train relational responding. Most REP work so far has been conducted with typically developing adults; however, given the importance of relational responding as a repertoire and of the potential utility of the REP for assessing and training this repertoire, researchers need to investigate its use with other populations including children with autism and other developmental delay. The current study presents relevant data. Nine children were initially assessed using a simple REP-based multi-level protocol (the NSD-REP) and on a measure of linguistic ability (Pre-school Language Scale, 4th Edition; PLS4). Three children found to pass Level 1 (same and different relations) but not Level 2 (affirmation and disconfirmation of same and different relations) were subsequently given a training intervention in the latter using a multiple baseline design. These findings supplement previous data suggesting the utility of the REP for training relational responding in children with autism.