Thirty Children, Two Languages Each

What the research found

A team based in Italy set out to answer a narrow, practical question. Can you screen a bilingual child for the risk of dyslexia over the internet, testing them in both of their languages, without the examiner needing to speak the second language? To find out, they built a web platform called MuLiMi and tried it on real children.

The sample was small and the authors say so plainly. Thirty children took part, all living and schooled in Italy. Eleven were monolingual Italian speakers. Nineteen were bilingual: seven spoke Mandarin alongside Italian, and twelve spoke English alongside Italian. The reading tasks the platform delivered were then compared against established, standardised reading tests, along with questionnaires filled in by parents and teachers. The aim was to see whether the screening picked up the same signals the trusted tests do.

For the most part, it did. Performance on the computerised tasks tracked performance on the standardised tests. Some of the associations were strong. Reading speed on a self-paced sentence task, for example, closely matched sentence-reading speed on the conventional test (a correlation of 0.942, which is very high). Accuracy in identifying written word forms on screen lined up with standardised word reading (a correlation of 0.584). Crucially, this held in the children’s home languages too: for the English-Italian group, reading time on the English sentence task matched their English-language reading risk (a correlation of 0.717). The tool was measuring reading in a language the examiner did not have to know.

The children mostly liked it. In a separate usability study, the eleven monolingual children and ten examiners rated the experience, and the general verdict was positive: easy to learn, pleasant to use, not boring. The main complaint was speed. Both children and examiners found the system slow to respond, which the authors trace to network delays rather than the design itself.

Here is the honest boundary the authors themselves draw. This is a preliminary validation, not a full one. Very few children in the sample had a known reading disorder or were flagged as clearly at risk. “Risk” here was inferred from test scores, not from any clinical assessment. No child was diagnosed. The study shows the machinery works and the signals point the right way. It does not show the tool can tell a struggling reader from a dyslexic one.

What it means for you and your child

The reason a study like this matters to parents has nothing to do with the software and everything to do with a persistent, avoidable mistake. Bilingual children are routinely misjudged. A child still building a second language can look, on a test written for monolingual peers, like a child with a learning difficulty. The reverse also happens: a genuine reading difference gets waved away as “just the bilingualism.” This work is part of a broader effort to stop guessing and start measuring a child in both of their languages, because a difficulty that shows up in only one language, or in neither, tells you something a single-language test cannot.

That principle is well founded and worth holding onto. If your child is bilingual and reading is a worry, the sensible question to ask any school or clinician is a simple one. Are you assessing my child in both languages, or only in English? A picture built from one language is an incomplete picture, and decisions made on incomplete pictures are how children end up in the wrong support, or in none.

Be equally clear about what this particular study does not give you. Thirty children is a small group, and the bilingual subgroups were smaller still, seven and twelve. Findings from numbers that small are fragile. They can shift or vanish when a larger, more varied sample is tested. Every child here was schooled in both of their languages, in ordinary or bilingual schools, which is not the situation of many bilingual children. The authors are explicit that their results cannot be stretched to bilingual children in general.

Most important, a screening tool is not a diagnosis, and the authors never claim otherwise. A screen is a filter. Its job is to raise a hand and say this child may warrant a closer look, not to name a condition. The paper states directly that its results “cannot directly lead to the diagnosis of a specific learning disorder” but instead flag an increased risk and the need for deeper investigation. Anyone who tells you an online test has diagnosed your child with dyslexia has misunderstood what these tools do.

So treat the finding for what it is: encouraging early evidence that fair, two-language, remote screening is possible and that children tolerate it well. That is genuinely useful, because the alternative is often no screening at all for children whose languages the local system cannot assess. But it is a first step. If a screen of any kind raises a concern about your child, the right response is a full assessment by a qualified professional who can work across your child’s languages, not a conclusion drawn from the screen itself.