A Language Test’s Norms, and the Children It Was Built For

What the research found

One hundred and ninety-four schoolchildren in Manizales, in central Colombia, sat two assessment appointments of roughly forty-five minutes each. Ninety-seven had a diagnosis of ADHD, made by a team of licensed psychologists and psychiatrists against DSM-5 criteria; ninety-seven were controls matched to them on sex, age, school year, socioeconomic stratum and type of school. Each group contained 24 girls and 73 boys, aged 5 to 14, with a mean age of 9.4 years.

The results were published in 2024 in the journal Children, by Diana Montoya-Londoño, Daniel Landínez-Martínez and colleagues at the Universidad de Caldas and the Universidad de Manizales, with a co-author at Georgia Southern University.

The instrument was the Children’s Neuropsychological Assessment, a battery constructed for Spanish-speaking children aged 5 to 16 and standardised on Colombian samples. Four of its tasks measure metalinguistic skill, which means the capacity to treat language as an object of attention rather than as a means of saying something. A child using metalinguistic skill is listening to a word instead of listening through it.

Those four tasks are small and specific. In phonemic synthesis, the examiner speaks a word as separate sounds and the child says which word it was. In spelling, the child spells eight words aloud. In sound counting, the child reports how many sounds make up each of eight words. In word counting, the child hears a sentence and says how many words it contained. Each task is scored out of eight.

Children with ADHD averaged 2.6 on synthesis against 3.3 for controls, 4.4 on sound counting against 5.4, and 3.7 on word counting against 4.8. By the Mann-Whitney U test, which the researchers used because none of the scores were normally distributed, those three differences exceeded chance at p values of 0.018, 0.011 and 0.009. Spelling produced a smaller gap, 4.1 against 4.8, which did not reach statistical significance at p equals 0.086.

The reading tasks divided the groups in the same direction. Accuracy on syllables, words, non-words and sentences, comprehension after reading aloud, and both oral and silent reading speed all favoured the control group. Two comprehension measures did not separate the groups: following a written instruction, and answering questions on a 92-word text read silently, where the p values were 0.077 and 0.093. The widest gap of all, at p below 0.0001, was in the number of words a child misread while reading aloud.

What this means for you and your child

Every child in this study had a full-scale IQ of at least 85, and a psychiatric interview had been used to rule out specific learning disorders before testing began. The difficulty on these tasks therefore appeared in children who did not have a diagnosed reading disorder to explain it.

Counting the sounds inside a word is an operation a child performs almost nowhere outside an assessment room, and a low score records difficulty in holding a word still long enough to inspect its parts. The study cannot show that ADHD produced that difficulty. The authors describe their design as experimental, but no one assigned ADHD to anyone; this is a matched comparison between two groups of children, and it establishes that they differed, not why.

The language the children were tested in is part of what the score means. These tasks were written for Spanish and normed on Colombian children, and the researchers argue in their introduction that Spanish exposes certain difficulties early, through its verb conjugations, its subjunctive and its rules of accentuation. They also press for international agreement on which instruments to use, on the grounds that prevalence figures from different countries are not currently comparable.

Whether these same four tasks would divide English-reading children at all is a question this research does not address, and no answer should be inferred from it. The study also reports no effect sizes, so how far apart the groups were is unrecorded; the p values establish that the gaps were unlikely to be chance and say nothing about their size. Families here volunteered, which the authors acknowledge may have produced a sample more engaged with their children’s schooling than the wider population of children with ADHD in the region.

What the assessment documents with confidence is narrow and real. Two of its measures of silent comprehension did not distinguish the groups, which leaves open whether some other task would. On the three tasks requiring a child to inspect the parts of spoken words, and in the count of words misread aloud, ninety-seven Colombian children with ADHD reading Spanish scored below ninety-seven peers matched to them by sex, age, school year and stratum.

A test report rarely names the children an instrument was standardised on, and on the report describing your child, that omission is worth a question.