Increase the spacing between letters in a word and second graders read better. Their comprehension improves measurably. The mechanism is visual crowding: letters positioned too close together interfere with each other's recognition. Wider spacing reduces the interference, giving the young reader's bottom-up decoding system cleaner input.
Apply the same modification to third graders and the benefit disappears. In some measures, comprehension trends worse.
Nothing changed except the reader. The crowding effect that constrains second graders' letter recognition has diminished by age nine. Their visual processing has automated the bottom-up work. They no longer decode letter by letter — they recognize word shapes, chunks, familiar patterns. Wider spacing disrupts those patterns by stretching the visual unit the reader has learned to process as a whole. The fix for yesterday's bottleneck is today's interference.
This is not the same problem as OpenDyslexic, where the intervention addressed a bottleneck that didn't exist (visual rotation instead of phonological processing). Here, the bottleneck was real. The spacing genuinely helped when crowding was the constraint. The intervention was correctly targeted at the right stage. It just didn't know when to leave.
The developmental transition from decoding to recognition is not a switch. There is no moment where spacing goes from helpful to harmful. There is a gradient — a period where the reader is simultaneously constrained by residual crowding and beginning to rely on word-shape recognition. In that window, the same spacing simultaneously solves one problem and creates another. The scaffold is both supporting and obstructing the structure it was built to protect.
Every developmental intervention has an expiration date. The ones that cause damage are the ones that don't know it.