Using matrix training to promote recombinative generalization by children on the autism spectrum in China
Matrix training gives you more spatial words for less teaching time with autistic kids.
01Research in Context
What this study did
Four autistic children in China joined matrix training for spatial words.
The team taught only diagonal pairs in a grid, then checked if kids said and pointed to new combos.
They tracked learning with a multiple-baseline design across children.
What they found
Every child learned the taught words and also used untaught ones.
Kids even followed new directions without extra teaching.
Some words were still shaky weeks later.
How this fits with other research
Frampton et al. (2016) did the same thing with noun-verb pairs years earlier.
Both studies show you can teach less and get more with autistic kids.
Carr et al. (1985) taught prepositions too, but used play-based incidental teaching.
That older paper says loose, natural moments beat drills.
Lee et al. now show a tight matrix works just as well, so you can pick either style.
Why it matters
You can save hours by setting up a simple grid instead of drilling every word.
Try it next time you target on, under, next to.
Teach the diagonal pairs, then probe the rest.
If the child stalls, flip to incidental teaching during play.
Want CEUs on This Topic?
The ABA Clubhouse has 60+ free CEUs — live every Wednesday. Ethics, supervision & clinical topics.
Join Free →Draw a 3×3 grid of objects and places, teach the diagonal pairs, then test the rest.
02At a glance
03Original abstract
We implemented tact matrix training to teach tacts of spatial locations to four children (male, 4-7 years of age) on the autism spectrum in China. The experimental design involved a multiple-probe design across participants with pre- and postinstruction probes on untaught tacts and listener responses. Learning outcomes included taught tacts of object-preposition combinations, generalization of untaught tacts, and derived listener responses to all combinations in the matrix. All four participants acquired taught tacts after matrix training. Untaught tacts and listener responses were demonstrated with direct teaching, indicating the occurrence of recombinative generalization. Two participants maintained these skills with high accuracy for 4 or 8 weeks. The remaining two participants demonstrated high accuracy in untaught tacts and listener responses immediately after instruction; however, accuracy in taught and untaught tacts declined during the 4- or 8-week maintenance probes, whereas listener responses remained stable.
Journal of Applied Behavior Analysis, 2025 · doi:10.1002/jaba.70025