Establishing five derived mands in three adolescent boys with autism.

Murphy et al. (2010) worked with three teenage boys with autism.

They used match-to-sample lessons to link picture symbols with real items.

After the links were firm, the team tested if the boys could ask for new items they had never directly been taught to request.

All three boys asked for items using the new symbols.

One teen needed a few extra examples, then he too succeeded.

No one had to drill every single request; the relations did the teaching.

Murphy et al. (2009) showed the same trick works for “more” and “less” requests in younger kids.

Belisle et al. (2020) later repeated the idea with comparative words like “bigger” and “faster,” proving the method lasts.

Chastain et al. (2025) stretched the logic even further, using equivalence training to teach perspective-taking frames like “I” versus “you.”

Together these studies build a roadmap: train a small network, watch untrained language pop out.

You can slash direct-teach time. Pick a few core relations, then probe for the untrained mands. If the teen doesn’t generalize right away, add extra examples instead of drilling every single request.