The Effect of Training Structure on the Latency of Responses to a Five-Term Linear Chain

The current experiment investigated the effect of differential training histories on responses to a 5-term linear chain of nonsense syllables (described here with sequential, alphabetical characters; A [is less than] B [is less than] C [is less than] D [is less than] E) across unreinforced probe trials. Participants' responses to nonarbitrary stimulus relations of More-than and Less-than were first brought under contextual control. Participants were then exposed to 1 of 3 training structures, in which each training structure was defined by the trial types that were presented: Less-More (A [is less than] B | B [is less than] C | D>C | E>D), All-Less (A [is less than] B | B [is less than] C | C [is less than] D | D [is less than] E) and All-More (B>A | C>B | D>C | E>D. The contextual cues served as sample stimuli, and 2 nonsense words as comparison stimuli. Twenty unreinforced probe trials were subsequently administered on all possible derived relations: directly trained, mutually entailed, and 1-, 2- and 3-node combinatorially entailed relation types. Comparisons of response latencies among 1-, 2-, and 3-node combinatorially entailed relations, for accurate performances on all 20 test trials, indicated that the former produced significantly longer latencies than the latter 2 relations. Comparisons of response latencies across the 3 training structures indicated that latencies were significantly lower in the All-More condition relative to both the Less-More and the All-Less conditions. The effects of nodal distance are readily predicted by both associative accounts and Relational Frame Theory, but the effect of training structure is readily predicted only by the latter theory. The reported findings are also broadly in accordance with previous studies on relational inferences reported in the literature from mainstream cognitive psychology.