This 2D companion draws the same dual-route model of reading aloud (Coltheart et al., "DRC model," Psychological Review, 2001) as a flat circuit diagram instead of a 3D scene — two competing pathways from a word on the page to speech output:
- Non-lexical (sub-lexical) route — orange path, top. Converts letters to sounds piece-by-piece using grapheme-phoneme correspondence (GPC) rules. Works for any letter string, including nonwords, but always produces a regularized pronunciation.
- Lexical-semantic route — teal path, bottom. Looks the whole written word up in the orthographic lexicon, retrieves its meaning and stored pronunciation directly. Only available for words already in the mental lexicon.
Both routes accumulate activation toward a response threshold at once — a race/accumulator model of decision time, identical to the 3D version:
dA/dt = k (linear accumulation, LATER-style race)
Response fires when the first A(t) reaches θ = 1
k_nonlexical ≈ constant (GPC conversion rate)
k_lexical(f) = 1 / (1.2 − 0.9·f) f = frequency ∈ [0,1]
Naming latency = t at threshold crossing
The two routes only disagree for irregular words (e.g. "pint," "sieve"): the GPC route regularizes them ("pint" → rhymes with "mint"). Because k_lexical grows with word frequency, high-frequency irregular words are usually read correctly (lexical route wins the race), while low-frequency irregular words are read aloud more slowly and are more often regularized — the exact frequency × regularity interaction measured in naming-latency experiments and seen exaggerated in surface dyslexia, where damage to the lexical route forces reliance on GPC.