Acquisition and extension of syntactic repertoires by severely mentally retarded youth.

Meyer et al. (1987) taught three youth with severe intellectual disability to say new two- and three-word phrases. They used matrix training. The team set up a grid of words. They taught only the diagonal items. Then they checked if the kids could mix the words in new ways.

The design was a multiple baseline across behaviors. Each child started at a different time. The goal was to see if untaught word combos would pop out on their own.

All three kids produced brand-new phrases they had never been taught. The new phrases stayed strong even when the setting changed. The study calls this 'robust recombinative generalization.'

Jimenez-Gomez et al. (2019) did the same thing with toddlers with autism. The little ones also made new noun-verb combos after diagonal training. That study extends this one down to younger kids and a new diagnosis.

Marya et al. (2021) kept the matrix idea but swapped spoken words for speech-generating devices. Kids with autism still got the untaught combos. This shows the method works in high-tech and low-tech forms.

Perez et al. (2015) moved from two-word phrases to full sentences with deaf children who had cochlear implants. Matrix training again created new, untaught sentences. The core finding holds across hearing status and sentence length.

Sailor (1971) and Sanders et al. (1971) are grand-dads to this work. They used old-school drills to teach plural endings and adjective rules. Kids with ID still generalized. Meyer et al. (1987) simply packaged the same 'teach a rule, get more for free' idea into a tidy grid.

You can get more language for less teaching. Pick a few key words, line them up in a grid, teach the diagonal, then probe. If novel combos emerge, you just saved hours of drill time. If they don’t, you know extra training is needed. Use this trick with any child who has sparse multi-word speech, whether they speak, sign, or use a device.