In Experiment 1, 12 senior citizens from the community were trained with 18 sets of conditional discriminations. Training included 2-, 3-, and 4-choice matching-to-sample (MTS) configurations in linear series (LS), many-to-one (MTO), and one-to-many (OTM) training structures. Training structure order was counterbalanced across participants. The design permitted tests for class establishment ranging from 2 classes of 3 stimuli each to 4 classes of 4 stimuli each in the LS, MTO, and OTM structures. The experiment tested the hypothesis that 3- and 4-choice MTS would increase the probability of class establishment, relative to 2-choice MTS, by reducing the potential for sample/S- control to arise during training. Results showed, however, that training with 3- and 4-choice MTS did not significantly increase equivalence class establishment and unequivocal evidence of sample/S- control was found in only 1 instance of a 2-choice training and testing structure. Experiment 2 systematically replicated Experiment 1 with 6 additional senior citizens in a 0-s delayed MTS paradigm. As in Experiment 1, equivalence class establishment was not related to number of choice stimuli. The delayed MTS paradigm, however, required fewer training trials to establish the conditional relations and led to more class establishment overall. The results are compared to data from previous studies with younger and older participants.