摘要：In this study, 12 5-year-old normal functioning children wereexposed to a respondent-type training procedure and tested foremergent conditional discriminations. During respondent-typetraining, arbitrary stimuli were presented in pairs, one at a time, usinga table-top procedure. On a given trial, for example, the arbitrarystimulus A 1 was presented on an observation card for 1 s, followed bythe arbitrary stimulus B1 presented on another observation card for 1s (represented as A1~B1). Emergent conditional discriminationswere tested using a standard matching-to-sample procedure. On onetest trial, for example, B1 was presented as a sample, with A 1 and A2as comparisons. Choosing A 1 (rather than A2) was defined as thecorrect choice (represented as B1-A1), based on the previousrespondent-type training (A1~B1).ln Experiment 1 (linear condition),subjects were trained and tested in the following sequence: trainA1~B1 , A2~B2, test B1-A1 , B2-A2; train B1~1, B2~C2, test C1-B1, C2-B2, C1-A 1, C2-A2; train C1 ~D1 , C2~D2, test D1-B1 , D2-B2,D1-A1, D2-A2. In Experiment 2 (one-to-many condition), subjectswere trained and tested in the following sequence: train B1~A1,B2~A2 , test A1-B1 , A2-B2; train B1~C1, B2~C2, test C1-B1, C2-B2, C1-A1 , C2-A2; train B1~D1, B2~D2, test D1-C1, D2-C2, D1-A1,D2-A2. In Experiment 3 (many-to-one condition), subjects weretrained and tested in the following sequence: train A 1 ~B1, A2~B2,test B1-A 1, B2-A2; train C1 ~B1, C2~B2, test B1-C1, B2-C2, C1-A 1,C2-A2; train 01 ~B1, D2~B2, test D1-C1 , D2-C2, D1-A 1, D2-A2. Thestudy demonstrated that · respondent-type training is an effectivemeans of generating equivalence classes with young children.Results also showed that it is possible to extend an equivalence classusing the respondent procedure without testing for the "mediating"symmetry relations. The training protocols (linear, one-to-many, andmany-to-one) were found to be equally effective, although a possibleceiling effect needs to be taken into account.
