Dickins, D. W. (2011). Transitive Inference in Stimulus Equivalence and Serial Learning. European Journal of Behavior Analysis, 12(2), 523–555.
[Abstract]The logical and behavioural properties of stimulus equivalence (SE) sets and serial learning (SL) sets are different, yet either can be derived from a randomly presented number of overlapping premise pairs, and both show transitive inference (TI). A within-participant experiment is reported which attempted to base both types of set on the same stimuli. To provide an ‘ecologically valid’ context the stimuli were photographs of 2 imaginary groups of 7 students ordered within each group by ‘exam grades’. Participants were given respondent-type training in ‘study phases’ in which the 12 premise pairs of photos were randomly presented without a response being required, alternating with ‘response phases’ in which the 10 participants in the ‘SE first’ group received matching-to-sample trials and the 10 in the ‘SL first’ group received trials with the study pairs of stimuli, in which they had to indicate whether these were in the same order as in the study phase or had been switched around. TI testing was then first conducted using the same requirement as in training, followed by similar tests using the other kind of response requirement. In a parallel sorting test participants were shown the 14 photos in random array on a screen and were asked to arrange them into 2 ordered groups. is sorting test was given 3 times, (1) after initial training on either SE or SL; (2) after TI testing with the same paradigm; (3) after TI testing with the opposite paradigm. Though the yield of accurate responding on the TI tests was poor, performance on initial TI testing was both more accurate and showed greater positive transfer to the other kind of TI test when SL preceded SE than vice versa. Results on the sorting task gave stronger indications of set formation than the TI tests, particularly in the SL first group. There were signs of the predicted increase in accuracy and decrease in RT as a function of increasing numbers of nodes in SL in the SL-first group, and some sign of the predicted inverse relation between accuracy and nodal number in SE for the SE-first group. When the groups switched to the opposite types of test to that on which they had been trained both showed an overall reduction in RTs and both showed decreasing RTs with increasing numbers of nodes. Unsurprisingly the experiment raised more questions than it could answer but suggested ways in which the similarities and differences between SL and SE, and how they interact, may be further explored.
[Citing Place (1995/6)]  
Citing Place (1995/6) in context (citations start with an asterisk *):
Section General Discussion
* [...] the possibility that SE is a precursor of language. SE classes are very like natural concepts (Fields, Reeve, Adams & Verhave, 1991), and the inclusion into them of symbolic stimuli such as spoken and written words, either by design in teaching reading (Sidman, 1971), or spontaneously in the participant’s deployment of linking names (Horne & Lowe, 1996) suggests that a positive feedback loop may have taken place in the joint evolution of SE and of language (Dickins and Dickins, 2001; Place, 1995).