The efficiency and efficacy of equivalence-based learning: A randomized controlled trial.

Austin et al. (2015) ran a small test with two children who had autism or developmental delay. The team taught A-B and B-C relations across sound, touch, and sight. They never directly trained the A-C link. The question: would the kids still match a sound to a picture without being told?

The whole study sat inside a single-case design. Sessions moved fast once the children learned the first two links.

Both children passed the untrained A-C tests. They heard a word and picked the correct picture, even though that pair was never taught. The same thing happened for the reverse C-A test. Cross-modal emergence showed up right after the basic training blocks.

The result fits the stimulus equivalence pattern: reflexivity, symmetry, and transference all appeared without extra trials.

Early et al. (2012) got the same positive outcome using listener-plus-speaker drills with young autistic learners. Their 2012 work is an earlier cousin; E et al. added the cross-modal twist.

Maddox et al. (2015) looks like a contradiction at first glance. That team also taught auditory-visual relations to kids with autism, but only children at ABLA-R Level 6 formed classes. The difference is in learner readiness. E et al. did not report ABLA-R scores, so their pair may have already shown the higher discrimination skill that B et al. says is needed.

Arntzen et al. (2018) later showed that a six-second delayed matching warm-up triples success with adults. The adult study extends the idea: a tiny pre-training step can unlock the same emergent leap E et al. saw in children.

You can build a bigger set of skills from fewer taught relations. After your learner masters two links, probe the untrained ones right away. If the child passes, move on; if not, check ABLA-R level or add a brief delay-matching warm-up like Arntzen et al. (2018). This saves teaching time and keeps sessions engaging by letting the learner 'discover' new relations without extra drills.