The contextually controlled, feature-mediated classification of symbols.

DeRosse et al. (2010) asked four adults to sort made-up symbols into new groups. The rules changed with the room cue: red screen meant one set of matches, green screen meant a different set. The team also trained tiny picture parts (a dot or line) so the same symbol could live in two classes.

The design flipped back and forth (ABAB) to prove the new sorting was truly under contextual control.

Only two adults fully mastered the switch. One learner showed clear red-vs-green control. One followed a private rule the trainers did not plan. One never reached full mastery.

The study shows adults can relearn equivalence classes, but success is not guaranteed even with clear cues.

Duker et al. (1991) got perfect results: six adults formed context-driven classes and even used the rule with brand-new pictures. DeRosse et al. (2010) used a tougher task—reclassifying the same stimuli—so mixed outcomes make sense.

Perez et al. (2021) went further. After classes formed, a simple background color told the learner whether to approach, escape, or ignore the picture. Their clean data extend Pamela’s core idea: context can steer what an equivalence stimulus means.

Foti et al. (2015) showed contextual control itself can travel through equivalence networks. Once one cue is trained, untrained stimuli fall under the same rule. Together these papers build a timeline: first prove context works, then show it can move, then show it can command three different functions.

If you teach conditional discrimination, plan for partial learners. Build extra exemplars and probe with context switches. Use clear sensory cues (color, border, location) and teach both the cue and the feature, not just the cue alone. When a learner stalls, try simpler stimuli or fewer class members before adding context layers.