Fluent component performances may be more readily available for recombination into more complex repertoires. This experiment considered the stimulus equivalence preparation as a laboratory analog for the co-adduction said to occur in generative instruction. Seven adults received minimum training on 18 conditional discriminations, components of 9 potential stimulus equivalence classes. Training was interrupted periodically with tests to determine whether fluency of original relations predicted emergence of derived relations. Fluency predicted emergence in 2 of 17 instances of emergent derived relations for 4 subjects. One subject demonstrated fluency without derived relations. Training accuracies as low as 58% preceded emergence for 3 subjects. Fluency appears to be neither necessary nor sufficient for derived relations. Fluency's role may be in retention and complex application tasks rather than acquisition of behavioral relations.

The effects of point loss for symmetrical probe performances on other performances of an observed equivalence class, on the emergence of equivalence performances, and on performances in other contexts were examined. After training six conditional discriminations in three contexts, probes (symmetry, transitivity, symmetrical transitivity) were introduced in contexts 1 and 2. In context 3, only trained conditional discrimination trials were delivered. After demonstrations of equivalence in contexts 1 and 2, point loss was placed on symmetrical performances in one of these contexts; probe trials and point loss for symmetrical performances were simultaneously introduced in context 3. Point loss for symmetrical performances may disrupt other probe performances of an observed equivalence class in that same context; does not necessarily disrupt the emergence of equivalence performances; and may disrupt probe performances in other contexts.