Time For A Fresh Approach To Learning Difficulties? The Cognitive Profile Of Kids Struggling At School Bore No Relation To Their Official Diagnoses

Screenshot 2019-01-09 at 16.23.06.png
The study used “machine learning” to organise children into clusters based on their cognitive profiles. (Figure 4 reproduced from Astle et al, 2018. See their open-access paper for description.)

By Emma Young

Around 30 per cent of British children fail to meet expected targets in reading or maths at age 11. These children face a future of continuing difficulties in education, as well as poorer mental health and employment success. Understanding why some kids struggle – and providing them with tailored support as early as possible – is clearly vital. Some will be diagnosed with a specific disorder, such as Attention Deficient Hyperactivity Disorder or dyslexia, and get targeted help. But many will not. And even many conventional diagnostic labels may be misleading, and fail to capture the true picture of a child’s problems, according to new work by a team at the MRC Cognition and Brain Sciences Unit at the University of Cambridge, which has come up with a radical, alternative approach. 

Duncan Astle and his colleagues studied 520 children aged between 5 and 18 years (average age 9) who had been referred to a research clinic at the unit for problems with attention, memory, language or poor school progress in reading or maths. Setting aside these diagnoses, they gave all the children a battery of assessments of their cognitive and learning performance, which included measures of working memory, phonological processing, spelling, reading, and maths. Their communication skills were measured using a separate checklist. A parent or carer also completed behavioural assessments of each child – reporting on their impulse control and emotional regulation, for example. The team then fed the cognitive and learning results for each child into an artificial neural network, which looked for any patterns in the data, grouping kids with distinct similarities into clusters.

One hundred and eighty-four of these children also had MRI brain scans – as did 36 typically performing children, for comparison – allowing the researchers to look for any group differences in patterns of communication between brain regions.  

As reported in their open-access paper in Developmental Science, according to the artificial neural network analysis, the children fell into four distinct groups. But these bore no relationship to their previous diagnoses. As the researchers write, “children referred primarily for problems with attention, poor learning or memory were equally likely to be assigned to each group”. This finding is worth stressing. The conventional diagnostic labels already given to the children did not reflect their cognitive profiles identified by the thorough testing in this study: “The four groups cut across any traditional diagnostic groupings that existed within the data,” the researchers wrote.  

So what did characterise these groups? 

More than half of the children in the sample fell into two extremes. Those in the “age appropriate” group in fact scored typically for their age on the cognitive tests; they did not have learning difficulties. But they did have an elevated level of behavioural difficulties, which presumably accounted for their problems in school. Meanwhile, members of the “broad cognitive deficits” group had widespread and severe cognitive problems, scoring in the bottom 5 per cent of the population on measures of spelling, reading and maths, and also experiencing difficulties with communication. 

“Generalised cognitive deficits therefore appear to constrain multiple aspects of learning,” the authors note. Relative to the other groups, kids in this group also showed reduced connectivity between some specific areas of the brain that have previously been identified as playing a role in multiple higher-order cognitive skills (such as problem-solving). “These general struggling learners are rarely studied, but our data suggest that they are common amongst those coming to the attention of children’s specialist services,” the researchers write. 

Children in the other two groups scored in between the two extremes, overall – but they also showed some specific deficits. 

One group was particularly poor at phonological processing – processing sounds in words – and verbal short-term and working memory. As might be expected, these children had trouble with speech, syntax and general coherence. And the brain imaging data showed reduced connectivity between regions implicated in language processing, supporting the AI analysis. However, these children were also poor at maths, which was unexpected. Though impaired phonological processing is typically associated with problems with reading, this data suggests that it’s likely to indicate more general learning deficits.

Members of the fourth group had distinct difficulties in working memory, scoring significantly below average on spatial, short-term memory and on verbal and spatial working memory. (There was no obvious differences in the brain connectivity patterns of these children.) 

One of the striking findings is that though the two intermediate groups had different specific deficits, their performance at maths and reading were almost identical. This contrasts with earlier research linking phonological problems to reading difficulties (dyslexia) and spatial short-term and working memory problems with trouble with maths (dyscalculia). This may be because previous studies have tended to recruit kids with these specific deficits – i.e. reading but not numeracy problems, for example. According to the new results, such specific learning problems may in fact be relatively rare in the general population of struggling learners (which chimes with a recent study of dyscalculia prevalence that found most children with dyscalculia also had language problems).

The researchers do acknowledge a few imitations with the study. For example, as a total of only 220 children were included in the neuro-imaging comparison portion of the research, only the largest and most consistent group differences are likely to have been revealed. (Which could explain why the team did not observe brain communication differences in a few of the groups.) 

But there are clearly potentially important insights here. If, as this work suggests, children can struggle with maths and reading for very different reasons, which do not necessarily align with diagnoses made in the conventional way, then surely the diagnosis and referral process has to be looked at – and the interventions that might help these children will surely have to be differently tailored, too. 

Remapping the cognitive and neural profiles of children who struggle at school

Emma Young (@EmmaELYoung) is Staff Writer at BPS Research Digest

4 thoughts on “Time For A Fresh Approach To Learning Difficulties? The Cognitive Profile Of Kids Struggling At School Bore No Relation To Their Official Diagnoses”

  1. Pingback: CAMHS Update
  2. Pingback: LD & Autism Update

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s