Adjustments That Work7 min read31 August 2026

The Three Conditions Your Adjustment Template Handles Worst

Different activities can need different formats and support.

Reading, moving and working with numbers each run on their own systems. When one processes differently it tends to get filed under effort, and the workplace response tends to be more training. A run of 2026 research maps what is happening underneath each of the three, and in every case it points at format rather than effort. That matters commercially, because format is the cheapest thing an employer can change.

What is going on

Dyslexia Genes Build Architecture, Not Just Reading Circuits

Researchers went looking for a reading gene and found two different jobs instead.

University of Houston researchers led by Elena Grigorenko examined the roughly 175 genes already linked to reading difficulty and found they split into two functionally distinct groups. One is active early in fetal development, building the brain’s physical wiring. The other tunes the reading circuits that come online once a child learns letters and sounds. Dyslexia is a difference in how broad neurodevelopmental networks are built rather than one faulty switch.

→Reading load is a legitimate design variable in a role. A summary at the top, a recorded briefing as an option, and documents that carry their own structure are changes to the format rather than concessions on the standard.

The Definition Changed, And Your Evidence Requirements Should Follow

The old definition treated dyslexia as fixed. The new one treats it as a continuum that moves across a lifetime.

In 2025 the International Dyslexia Association revised its definition of dyslexia for the first time since 2002, and a 2026 consensus paper built from 58 researchers, practitioners and dyslexic adults expanded it further. Dyslexia is now defined as difficulty with word reading and spelling accuracy or speed, sitting on a continuum of severity, shaped by genetic, environmental and cognitive factors together.

→If your adjustment process requires a formal diagnostic report before anything changes, it runs on a definition that has been revised and on assessment waiting times you do not control. Adjust on the difficulty described, and let the paperwork follow.

An AI Model Of Dyslexic Reading Found Fonts That Measurably Help

Researchers built a digital twin of dyslexic reading, then tested what changes on the page.

EPFL researchers used vision-language models to explore aspects of dyslexic reading and investigate typography. Results in an AI model are not evidence that a particular font improves reading for every dyslexic person. Human studies and individual preferences still matter.

→Typography in your internal comms and e-learning is an evidence based lever rather than a preference. It is also the rarest kind of adjustment: implemented once, centrally, for everybody, with no ongoing cost.

Dyspraxia Shows Up In Brain Structure And Genes Together

Clumsy was never the right word. The coordination difference has a measurable structural signature.

Developmental coordination disorder, commonly known as dyspraxia, affects planning and carrying out physical movement. A 2026 study modelling brain structure and genetics together found increased cortical volume paired with decreased subcortical volume tracking with poorer motor coordination, and that same pattern correlating with genetic markers already linked to coordination difficulty.

→Check where your processes still require handwriting, manual dexterity or speed of physical execution for no operational reason. Each is a barrier with a structural cause and a free workaround.

Dyscalculia Has Its Own Circuit, And Your Numerical Reasoning Test Sits On Top Of It

Number sense was assumed to be hardwired. New modelling suggests it may be learned, and trainable.

Children with dyscalculia, a specific difficulty processing numbers and quantities, show hippocampal hyperactivity when perceiving symbolic numbers, which suggests active compensation rather than a missing sense. Newer work using neural networks is testing whether number sense is learned through experience rather than present at birth, which would make the gap a training target rather than a fixed ceiling.

→A timed numerical test that is not genuinely related to the job is both a barrier and an exposure under the Equality Act. Untimed, calculator permitted, or replaced with a work sample costs nothing and widens the field.

What this could mean in practice

What to do about it

Match the format to the mechanism. It is almost always cheaper than the alternative.

None of these findings point at more repetition. A gene network that builds brain architecture is not addressed by more phonics, and a coordination difference rooted in brain structure is not addressed by marking down handwriting. What each one points at is a specific change of format: typography tested against a model of dyslexic reading, processes that do not require manual dexterity for no reason, and numerical screening that is either relevant to the job or removed from it. Those are process decisions rather than therapy.

"The format needs to match the mechanism, and the format is ours to change"

Offer readable formats and adjustable text where possible. Ask staff what helps and account for implementation work.

Audit where handwriting and manual dexterity are still required. Most instances turn out to be historical rather than operational.

Timed numerical screening unrelated to the role filters out capable candidates and carries Equality Act exposure. Make it relevant or remove it.

One audit is worth running: every point in your recruitment and appraisal processes that requires reading at speed, handwriting, or mental arithmetic. For each, ask whether the job genuinely requires it. The ones that do not are barriers you are choosing to keep.

Sources