This study investigated the relationship between parenting stress and relational conditioning. Fourteen students who were not mothers, 14 mothers who reported high parenting stress and 14 mothers with low parenting stress completed two matching-to-sample (MTS) computer tasks, each requiring formation of three 3-member classes. The first MTS task examined class acquisition, with neutral symbols and stimuli theorized to be emotionally relevant to parents (i.e., negative child behavior words) or not (i.e., fruit or medical words). The second task required participants to form classes that included symbols, negative child behaviors and positive parenting words. Statistical analyses were conducted on number of trials needed to meet criterion, percent correct in training and testing, and latency to respond. Overall, distressed mothers were slower to respond than other participants. However, in MTS 1, they responded more quickly and made fewer errors in response to stimuli in the child behavior class. Further, they obtained a higher percent correct for these items than the other participants. In the second MTS, distressed mothers required more trials to reach criterion and made more errors than the other participants. Findings indicate that emotionally relevant stimuli are easily related to other things. However, once formed, emotionally relevant classes are difficult to relate in new ways. The findings may have important clinical and treatment implications for parents, particularly if computer responding relates to "real-world" inflexible parenting. (Contains 4 figures.)