The present study investigated differential effects of training design on probability of an equivalence outcome. Forty normal adults were assigned to four different groups. Subjects in the first three groups were exposed to a C-A equivalence test directly following linear series (AB and BC), many-to-one (AB and CB), or one-to-many (BA and BC) training, respectively, while the subjects in the fourth group, following linear series training, were exposed to a symmetry test before the C-A test. Three comparison stimuli were used throughout the experiment to minimize effects of control by negative comparisons. Number of subjects showing equivalence were highest following one-tomany training and lowest following linear series training. Several previously suggested explanations of the differential effects of training design are discussed and shown to be unsatisfactory. The minimal conditional discriminations that are prerequisites for the testing of emergent equivalence relations are established by any of the following match-to-sample procedures: (a) linear series (LS), A-B and B-C, in which the comparison stimuli from the first conditional discrimination training serve as samples in the second, (b) many-to-one (MTO; comparison as node), A-B and C-B, in which the same stimuli are comparisons for different sets of samples in the first and in the second task, or (c) one-to-many (OTM; sample as node), 8-A and 8-C, in which the same stimuli serve as samples in two tasks with different sets of comparison stimuli (see Fields & Verhave, 1987; Saunders, Saunders, Williams, & Spradlin, 1993). Differences in outcome as a function of training design, that is, OTM versus MTO training, have been studied with mentally retarded subjects

A lack of joint attention skills may constitute a core impairment in autism. In the present study, a training protocol was developed, based on the literature on joint attention and on behavioral interventions. The training was organized into a sequence of three main parts respectively aimed at establishing each of the following skills: (1) responding to joint attention bids, (2) engaging in turn-taking activities based on joint attention skills, and (3) initiating joint attention. Two novel components were incorporated in the training: (a) a discrimination training procedure aimed at establishing the adult's nods as conditioned reinforcers and (b) tasks based on turn-taking, where joint attention skills were targeted and reinforced. The study was conducted according to a single-subject experimental design, in which joint attention skills were measured before and after intervention, using the ''behavioral assessment of joint attention.'' Four 3.5-5.5 year-old children diagnosed with autism participated in the study. All four children completed the training successfully and made significant progress in engaging in joint attention and in initiating joint attention skills. Following the completion of training and at 1 month follow-up, parents reported that their children used their skills in different settings. Moreover, at follow-up, all four children were reported to engage in joint attention behaviors and to enjoy doing so.

Stimulus equivalence, defined as C-A matching, was tested in 80 adults following training in a matching-to-sample task involving arbitrary A-B and B-C matching. In Experiment 1, 50 subjects, successively assigned to one of five groups, were exposed to specific stimulus material. The stimuli for subjects in the first group were Greek letters only. In the remaining groups, pictures were incorporated as A-, B-, and/or C-stimuli. The probability of equivalence was low when the stimulus material consisted only of Greek letters. For the remaining groups, the probability of equivalence varied considerably depending upon whether the A-, B-, and/or C-stimuli were pictures. The results indicate that seemingly minimal procedural variations can yield markedly differential outcomes not predicted by any existing model. The results also showed that responding gradually may become consistent with a pattern other than the predicted equivalence pattern. Experiment 2 replicated the results of Experiment 1 showing (a) differential probabilities of equivalence in individual subjects, depending upon the configuration of Greek letters and pictures during training/testing, (b) consistent patterns of responding even when the responding was not in accord with equivalence, and (c) higher reaction times to comparison stimuli initially during testing. Probabilities of equivalence increased in a second exposure to the tasks involving only Greek letters, whether or not the subjects were exposed to a task with pictures prior to the second Greek-letter task. Higher reaction times initially during testing may indicate precurrent problem solving behavior prior to the selection of a comparison stimulus. The finding of delayed emergence of consistent responding suggests that even the slightest tendency toward responding that partitions the stimuli into the experimenter-planned equivalence classes may evolve into consistent responding in accord with those classes.The authors gratefully acknowledge Charles Catania for his comments on an earlier version of this manuscript.Order of authorship was determined by the flip of a coin . Correspondence and reprint requests may be sent to Per Holth,

No abstract

Previous studies comparing groups of subjects have indicated differential probabilities of stimulus equivalence oute,ome as a function of training structure. Both one-to-many (OTM) and manyto-one (MTO) training structures seem to produce stimulus equivalence more often than a linear series training structure. The purpose of the present study was to explore whether or not corresponding differential probabilities of equivalence outcome as a function of training structure can be demonstrated in the performances ' of single subjects. In Experiment 1, equivalence outcome was tested successively following training according to each of the three training structures. All subjects responded in accord with equivalence following the OTM training structure independent of the training order, except for 1· subject who did not respond in accord with equivalence following neither of the three training structures. Furthermore, 2 subjects demonstrated individual success following both one-to-many and many-to-one training even when they did not demonstrate success following linear series training, while the reverse never happened. In Experiment 2, equivalence outcome was tested successively following training according to a many-to-one and a one-to-many training structure with both 2 and 3 classes of stimuli. The results showed that all subjects responded in accord with equivalence following the OTM training structure, while 2 subjects did not respond in accord with equivalence following the MTO training structure. In Experiment 3, equivalence outcome was tested successively following training according to a many-to-one training structure and one-to-many training structure with both 3 and 4 members in each class. In accord with the results of Arntzen and Holth (1997) the present results indicated a superiority of the oneto-many over alternative training structures as regards the . probability of an equivalence outcome. There was no difference in the probability of an equivalence outcome following one-to-many