The Made-Up Words That Told Them Apart

What the research found

Bilingual children are hard to assess. A child who is still building English or German or Italian may sound, on a standard test, much like a child who has a genuine language disorder. The scores dip for both. Distinguishing the two matters, because one child needs time and exposure while the other needs support. Getting it wrong in either direction is costly.

This study, published in Frontiers in Psychology in 2022, looked at one tool that may help: a nonword repetition task. The idea is simple. The child hears an invented word, something like a real word but with no meaning, and repeats it back. Because the word is made up, no child can have learned it in advance. What the task measures is the underlying machinery of language, the ability to hold a string of sounds in mind and reproduce it, rather than vocabulary a bilingual child may not yet have acquired.

The researchers, Maren Rebecca Eikerling, Theresa Sophie Bloder and Maria Luisa Lorusso, studied 37 Italian-German bilingual children aged three years ten months to six years one month. The children fell into three groups: 13 who were developing typically, 17 judged to be at risk for language difficulties, and 7 with a diagnosed developmental language disorder. The task itself contained 21 nonwords, each around three syllables long. Six were built from sounds specific to Italian, six from sounds specific to German, and nine used only sounds common to both languages.

The task sorted the children well. Using the total score, it separated children with the disorder from typically developing children with an area under the curve of 0.898, a measure where 1.0 is perfect and 0.5 is a coin toss. At a cut-off score of 12.5, it correctly flagged 86 per cent of the children with a disorder and correctly cleared 77 per cent of those without one. For identifying phonological risk more broadly, the figures were similar, with an area under the curve of 0.917.

The more striking result concerned which nonwords worked best. The words built from language-specific sounds, the ones carrying phonemes and consonant clusters unique to Italian or to German, discriminated more sharply than the words using only shared sounds. The language-specific items reached an area under the curve of 0.931 for the disorder, against 0.826 for the shared-sound items. In other words, the words that leaned on each language’s particular texture were the more revealing test.

What it means for you and your child

If your child is growing up with two languages and someone has raised a concern about their speech or language, this study offers a modest, useful point. A nonword repetition task is one of the fairer ways to look. Because it does not depend on knowing words, it does not penalise a child simply for being partway through learning two systems at once. It probes something closer to the underlying capacity.

The finding about language-specific nonwords carries a practical lesson. A test that draws on the specific sounds of your child’s languages is likely to tell you more than a generic one built from sounds every language shares. This is worth keeping in mind if an assessment is offered. A tool designed for the actual language pair a child speaks is better placed to separate a real difficulty from the ordinary business of bilingual development.

It is important to be honest about what this single study can and cannot show. The sample was small: 37 children, and only 7 with a confirmed disorder. Numbers that small can shift with a few children either way, and the impressive-looking figures should be read as promising rather than settled. The authors say as much. They also note that the groups were uneven, that the range of bilingual backgrounds was narrow, and that all the children with a diagnosed disorder happened to be stronger in Italian, which could have coloured the results.

There is a further caution the authors raise, and it is a subtle one. To decide which children had a disorder in the first place, they leaned partly on standard tests that are not well suited to bilingual children. That is the very problem the nonword task is meant to solve. So the study cannot fully escape a degree of circularity, checking a new tool against yardsticks that are themselves imperfect for this population. This does not undo the result, but it does mean the task’s accuracy needs confirming against cleaner benchmarks and in larger, more varied groups.

For a parent, the sensible reading is this. Repeating nonsense words is not a gimmick; it is a recognised and reasonably fair window into a bilingual child’s language system, and the evidence here supports using it. But no single number from one small study should decide anything on its own. A good assessment gathers several kinds of evidence, over time, from someone who understands bilingual development. Treat a strong or weak score here as one informative signal among several, not a verdict. If a concern persists, ask specifically whether the assessment accounts for your child being bilingual, and whether the tools used were built with that in mind.