This work evaluated how portions of the environment uncorrelated with the consequence influenced the establishment of stimulus control. In seven experiments, graduate students had simultaneous discrimination tasks in which a small portion of the antecedent stimuli correlated with the consequence while the other portions did not. Depending on the experiment, a computer presented human-like stimuli, the same stimuli in a scrambled format, or crosses formed by colored squares. The crosses were formed by squares joined by their corners or squares separated some inches. A categorization task measured the control gained by the uncorrelated portions. In all the experiments, portions of the stimuli uncorrelated with the consequence gained control over some responding. The control gained by uncorrelated portions hindered or totally prevented the gaining of control by the correlated portions in some cases. Uncorrelated portions of stimuli in human format had a greater chance of gaining control over responding than the same stimuli in a nonhuman format. Only the uncorrelated components of the crosses formed by separated squares gained control over responding; those of crosses formed by joined squares did not. The results demonstrate that the arrangement of antecedent stimuli affect the establishment of control by correlated and uncorrelated portions.
From 1982 to 2015, many concepts have been added to Stimulus Equivalence paradigm and they do not explain how a person can acquire so many words into his /her repertory. This paper presents a short review of such concepts, from Sidman & Tailby's sets that produced unexplained untrained relations among arbitrary elements emerging new equivalent classes, to the change of the paradigm with Fields, Verhave and Fath introducing the nodal distance and naming as node a bi-linked stimulus through training, and the MTS (Matching-To-Sample) graph operator of Oliveira. A net of nodes linked by trained relations may show special properties, such as weight, distance, density, among others. The papers on Stimulus Equivalence propose mostly the study on the limits of the network growth (changing number of nodes, changing the number of elements in each set, strength of the link on trained and untrained relations, etc.) and the hybrid experiments composing more than one of those characteristics. A discussion on the limits based on the strategies of teaching such as SaN and CaN, linear chains or mixed structures is also provided. This theoretical study concludes that composing and multiplication of such nets is still lacking of study.