Christoffer Eilifsen

The Effects of Different Training Structures in the Establishment of Conditional Discriminations and Subsequent Performance on Tests for Stimulus Equivalence

Previous studies comparing groups of subjects have indicated differential probabilities of stimulus equivalence outcome as a function of training structures. One-to-Many (OTM) and Many-to-One (MTO) training structures seem to produce positive outcomes on tests for stimulus equivalence more often than a Linear Series (LS) training structure does. One of the predictions from the discrimination analysis of R. R. Saunders and Green (1999) is that the differences in outcome between training structures should increase with number of class members. The purpose of the present experiment was to replicate and expand earlier findings on the effect of training structures and the stimulus equivalence outcome in a single-subject design. We wanted to compare the stimulus equivalence outcome in three 3-member classes to the outcome in three 4-member classes. In addition, we included all trial types in the tests and also changed the density of feedback before testing. The results from the current study replicated some earlier findings and showed that OTM gave a slightly better outcome on the stimulus equivalence test than MTO, and that both gave better outcome than LS. Thus, we did not find that MTO was superior to OTM with increasing number of members in each class. Reaction time data also replicated earlier findings that showed an increase from baseline to testing, and a more pronounced increase in reaction time on equivalence than symmetry trials. Differential procedural issues and some contingencies that could be important in understanding the results are discussed.

On the role of trial types in tests for stimulus equivalence

Some studies which have shown that differences in outcome on tests for stimulus equivalence dependent on different training structures, have run the tests as separate blocks without baseline trials interspersed in between test trials. Saunders and co-workers have argued that the differences in test outcome could be related to differences in the retention of trained discriminations during testing (R. R. Saunders, Drake, & Spradlin, 1999; R. R. Saunders & Green, 1999). In the current experiment, 20 adult participants were taught conditional relations by employing a linear series training structure. Following this training, non-reinforced trials of the directly trained relations were randomly interspersed in a mixed test for symmetry, transitivity, and global equivalence. After being exposed to the training procedure once, 17 of the participants did not perform in accord with stimulus equivalence, but 9 of these participants still responded in accord with the directly trained relations. After being exposed to the training procedure again, 10 participants still did not respond in accord with stimulus equivalence, while 7 out of these did respond consistent with the directly trained relations. This indicates that a “destroyed” baseline could not be responsible for these participants’ failure to respond in accord with stimulus equivalence. In addition the reaction time between the appearance of comparison stimuli and subsequent responding during the test were recorded. Data show that the average reaction time varies as a function of which type of relation that is presented. There were also distinct differences in reaction time patterns for those participants who responded in accord with stimulus equivalence compared to those not responding in such a manner.

Response to sample requirements in conditional discrimination procedures

The present experiment examined number of training trials as a function of a required response to sample in establishing conditional discriminations and testing for responding accordance with stimulus equivalence. Reaction times were also examined. Twenty participants in 4 different experimental groups were compared in a between-groups design. The participants were trained to form 4 classes of 3 stimuli each with a linear series (LS) training structure. Unreinforced trials for directly trained relations were included in the test. Results from the present study show that a required response to sample increased chances to learn the discriminations more quickly. This may imply that the required response can increase the observing behavior. Required response to sample did not affect responding according to emergent relations in the test in any substantial way. Reaction time data show that there are systematic differences in reaction latency depending on the trial type in the test, and that incorrect responses may not be randomly selected.

Single-Subject Withdrawal Designs in Delayed Matching-to-Sample Procedures

In most studies of delayed matching-to-sample (DMTS) and stimulus equivalence, the delay has remained fixed throughout a single experimental condition. We wanted to expand on the DMTS and stimulus equivalence literature by examining the effects of using titrating delays with different starting points during the establishment of conditional discriminations prerequisite for stimulus equivalence. In Experiment 1, a variation of a single-subject withdrawal design was used. Ten adults were exposed to one condition where the delay titrated between 0 and 3000 ms and another condition where the delay varied from 5000 to 8000 ms. Subsequently, participants were re-exposed to the condition they had first experienced. Results show that starting the titration of the delay at 5000 ms may have had a facilitatory effect on stimulus equivalence responding for some participants. For several participants, however, performance was stable throughout the experiment, apparently not affected by either the variations of the starting point of the titrating delay or by previous exposure to stimulus equivalence training and test procedures. In Experiment 2, we examined the effects of being exposed to the same stimulus equivalence procedure three times, again with adults as participants. Results show that such extended exposure had very limited effects on stimulus equivalence responding. This indicates that single-subject withdrawal designs may be an appropriate approach for studying stimulus equivalence.