Visual artificial grammar learning in dyslexia: A meta-analysis.

Merel and colleagues pooled every visual artificial-grammar study done with dyslexic readers. They ran a meta-analysis to see if the group truly learns picture patterns worse than typical readers.

The papers used simple shapes that follow hidden rules. After viewing strings, participants pick new ones that fit the same rule. Fewer correct picks signal weaker implicit learning.

Across the pooled studies, dyslexic readers scored lower on average. Yet the authors spotted strong publication bias: small studies with big effects got printed, while small or null studies stayed in file drawers.

After correcting for this bias, the effect shrank and was no longer reliable. In plain words, we cannot be sure dyslexia really dampens visual pattern learning.

Cicchetti et al. (2014) is inside the meta. That single study did show a clear child deficit, so the meta both includes and tempers its strong claim.

Ortiz et al. (2014) found visual temporal-order problems in preschoolers at risk for dyslexia. Their early timing fits with the idea that low-level visual issues exist, but the meta says the later AGL task may not capture them well.

Calet et al. (2019) reported medium rhythm and prosody deficits in Spanish-speaking children. Both papers flag medium negative effects, yet the meta warns the AGL number could be inflated, so the true overlap is uncertain.

If you test rule learning to screen for dyslexia, treat low scores as a maybe, not a diagnosis. Add extra reading-based probes before making placement decisions. When you design interventions, keep using explicit phonics; do not assume fixing visual pattern games will transfer to decoding.