D01 · Grade A · dyslexia specificDo not invent a strange dyslexia alphabet
Finding: A 2026 meta-analysis of 15 studies and 688 dyslexic readers found no reliable overall reading advantage for specialist dyslexia fonts as a category.
Decision: Keep familiar, socially neutral letter skeletons. Do not use heavy bottoms, arbitrary rotation, irregular baselines, or novelty merely to signal dyslexia.
Font change: Both Harmonia and High Speed use familiar, socially neutral skeletons. High Speed increases local distinction without heavy bottoms, irregular baselines or decorative disability cues.
Evidence records: E002 E019 E020
D02 · Grade A/B · dyslexia specific plus design inferencePreserve familiar orthographic forms
Finding: Orthographic knowledge is an important component of reading, and unusual specialist letterforms have not reliably improved passage reading.
Decision: Increase distinction locally instead of redesigning every glyph.
Font change: Harmonia prioritizes familiar forms. High Speed provides stronger alternative silhouettes as a separate explicit family rather than hidden contextual alternates.
Evidence records: E005 E020 E061
D03 · Grade A · general visual recognitionProtect open apertures and counters
Finding: Controlled experiments show more identification errors when apertures are closed.
Decision: Keep c, e, r, s, f and t open, including in heavier weights.
Font change: Both Reading families protect open internal white space. High Speed further emphasizes selected openings only where recognition benefits from clearer silhouette separation.
Evidence records: E078
D04 · Grade A · general visual recognitionUse moderate weight, never extremes
Finding: Very thin and extremely bold strokes reduce recognition; moderate additional boldness helps more at smaller visual angles.
Decision: Regular is the main reading weight. Medium is recommended for small screens or small visual sizes.
Font change: Light, Regular, Medium, SemiBold and Bold are included. Regular and Medium are the preferred starting points for sustained body reading; readers can compare weights in the Type Studio.
Evidence records: E079
D05 · Grade B · general typographyUse low stroke contrast
Finding: Controlled research shows performance depends on stroke contrast as well as serif presence.
Decision: Keep the approved low-contrast sans structure.
Font change: The family remains visually clean and consistent rather than introducing high-contrast display strokes.
Evidence records: E080 E050
D06 · Grade B · general visual recognitionWiden reading forms subtly, not globally
Finding: Wider letters improved parafoveal and peripheral letter recognition, while other work shows width also changes eye-movement strategy.
Decision: Create a separate Reading cut with restrained width increases.
Font change: Each family has a dedicated Reading cut with calmer text proportions and a tighter Display cut for larger typography. The cuts are explicit font files rather than synthetic browser transforms.
Evidence records: E081 E062
D07 · Grade A · dyslexia specificTreat letter spacing and word spacing as one system
Finding: Increasing letter spacing without adequate word spacing can slow reading, and spacing can explain apparent specialist-font advantages.
Decision: Avoid extreme tracking. Preserve strong word boundaries.
Font change: Built in spacing remains moderate. The website Type Studio exposes additional letter spacing for reader controlled comparison rather than forcing exaggerated tracking into every setting.
Evidence records: E016 E017 E018 E047
D08 · Grade B · dyslexia specificIncrease word boundary salience cautiously
Finding: A 2024 study reported improved comprehension and fewer migration errors with increased inter-word spacing in students with dyslexia.
Decision: Increase word spacing modestly while retaining normal word texture.
Font change: Word boundaries are kept clearly visible while letter spacing remains restrained. Word spacing is independently adjustable in the Type Studio.
Evidence records: E077 E016
D09 · Grade A · dyslexia specificCrowding matters, but more spacing is not a cure
Finding: Crowding and spacing studies are mixed; some readers benefit, while other studies show no general speed or comprehension benefit from wider spacing.
Decision: Keep spacing moderate and make future personalization possible.
Font change: Reading cut is a strong default, not an extreme dyslexia preset.
Evidence records: E013 E014 E015 E047 E003
D10 · Grade B · recognition focused design inferenceKeep I, l, 1 and i separable
Finding: Distinctive forms can improve isolated letter identification even when passage benefits are not guaranteed.
Decision: Strengthen high-risk local distinctions without making the alphabet unfamiliar.
Font change: Harmonia keeps calm familiar I, l, 1 and i forms. High Speed increases local silhouette differences while preserving the overall visual language.
Evidence records: E020 E082
D11 · Grade B/C · personalization hypothesisOffer stronger a and o separation through explicit family choice
Finding: A more differentiated a may improve local recognition for some readers, but unfamiliar orthographic forms can impose a familiarity cost.
Decision: Keep Harmonia familiar, and expose stronger a and o silhouette separation in High Speed as an explicit family choice rather than a hidden alternate.
Font change: The stronger recognition treatment is exposed as the High Speed family. It is a visible family choice, not a hidden OpenType switch.
Evidence records: E005 E020 E082
D12 · Grade B · general recognitionDifferentiate O and 0 through structure, not decoration
Finding: Alphanumeric confusability matters, but slash-zero can collide conceptually with Ø in supported languages.
Decision: Use proportion and construction to separate O and 0. Do not rely on an unusual default slash, dot or decorative mark merely to signal accessibility.
Font change: Harmonia differentiates O and 0 through proportion and construction. High Speed increases the distinction structurally while preserving a professional texture.
Evidence records: E082
D13 · Grade A/B · dyslexia theory guardrailDo not distort b d p q around reversal folklore
Finding: The evidence does not support treating letter reversal as the universal cause of dyslexia, and specialist letter-shape novelty does not reliably improve passage reading.
Decision: Keep directional forms conventional and distinct through normal typographic cues.
Font change: No heavy bottoms, rotation or dramatic asymmetry is forced on b d p q.
Evidence records: E002 E019 E020
D14 · Grade B · crowding inferenceProtect rn and vv from accidental merging
Finding: Local crowding and letter proximity can impair identification even when more global spacing is not beneficial.
Decision: Use pair-specific protection rather than global tracking.
Font change: Potential rn and m collisions are handled through width, shoulder construction and spacing, with stronger separation in High Speed.
Evidence records: E013 E078
D15 · Grade A · dyslexia specific plus vision scienceUse enough effective visual size
Finding: Dyslexic children can require larger critical print sizes, and reading speed depends strongly on visual angle near the critical threshold.
Decision: Do not compensate for undersized text with gimmicks.
Font change: Reading guidance prioritizes adequate size; Medium weight is available for smaller text.
Evidence records: E021 E022 E051
D16 · Grade B · dyslexia specific emerging evidenceProtect word-initial clarity
Finding: Parafoveal studies suggest altered early letter-position processing, and recent gaze-contingent work reports greater foveal load in dyslexic children.
Decision: Keep word-initial glyphs simple and recognizable.
Font change: No decorative entry strokes or contextual novelty at word beginnings.
Evidence records: E034 E036 E037 E043
D17 · Grade A/B · evidence synthesisUse an optical family instead of one fixed geometry
Finding: Legibility varies by reading scenario; no single typeface geometry is universally optimal.
Decision: Separate aesthetic large-text needs from continuous-reading needs.
Font change: Harmonia and High Speed each provide dedicated Reading and Display cuts across five weights with upright and italic styles.
Evidence records: E082 E056
D18 · Grade B · conservative design inferenceDo not use mandatory ligatures
Finding: The product prioritizes explicit grapheme identity and local character recognition.
Decision: Avoid automatically merging visible letters in the Reading cut.
Font change: Standard reading CSS disables optional ligatures for body text.
Evidence records: E020 E061
D19 · Grade B · general reading designPunctuation must remain visible
Finding: Reading is not only letter recognition; segmentation and navigation cues should remain visible at text sizes.
Decision: Keep punctuation optically visible and rhythmically consistent at reading sizes without novelty enlargement.
Font change: Punctuation is kept optically visible and rhythmically consistent in both Reading cuts without novelty enlargement.
Evidence records: E082
D20 · Grade A technical · multilingual engineeringPublish only audited diacritic and language coverage
Finding: A font cannot claim multilingual support if stacked marks clip or combine incorrectly.
Decision: Treat diacritics as first class reading information and limit language claims to the audited character repertoire.
Font change: The published release contains broad precomposed Latin coverage plus combining mark glyphs. Language support claims are limited to the audited repertoire and do not imply complete script shaping for unsupported writing systems.
Evidence records: E026 E027
D21 · Grade A · dyslexia specificDo not prescribe color as dyslexia treatment
Finding: Systematic-review evidence does not support colored overlays as a general dyslexia treatment.
Decision: Offer theme choice for comfort, not therapeutic claims.
Font change: Website supports light and dark themes; the font itself remains color neutral.
Evidence records: E025 E055
D22 · Grade A · direct experimental evidenceNo Bionic-style prefix bolding
Finding: Controlled eye-movement research found no reading benefit and costs from bolding the first part of every word.
Decision: Do not build prefix bolding into the font or reading system.
Font change: Bold is reserved for normal semantic hierarchy.
Evidence records: E063
D23 · Grade A · dyslexia specificTreat eye movements as outcomes, not causes
Finding: Dyslexic readers often show different reading eye movements, but basic oculomotor function can be normal and lexical difficulty strongly modulates gaze.
Decision: Use gaze as a validation signal, not as a mythology for glyph design.
Font change: Human validation protocol measures gaze optionally alongside accuracy, comprehension and effort.
Evidence records: E023 E024 E039 E047
D24 · Grade C · product hypothesisMake personalization granular instead of strange by default
Finding: Dyslexia and visual-attention profiles are heterogeneous across readers and languages.
Decision: Keep strong defaults while exposing meaningful controls that correspond to real shipping font files and normal typographic variables.
Font change: Personalization is primarily explicit: Harmonia or High Speed family, Reading or Display cut, weight, size, letter spacing, word spacing and line height. The website makes these controls directly testable.
Evidence records: E004 E035 E056 E058