Intuitive simplification is the dominant industry practice of writing or rewriting texts for second-language readers based on the author's judgement of what learners at a given level can handle, rather than by applying a numerical readability formula. Graded readers, ELT coursebook passages, and levelled news services almost all rely on it. The author works from a wordlist and structural syllabus, but moment-by-moment decisions about which clause to break, which referent to make explicit, and which item to gloss sit in the writer's intuition.
Crossley, Allen, and McNamara (2011) assembled a corpus of texts intuitively simplified at advanced, intermediate, and beginner levels, then ran traditional readability formulas (Flesch–Kincaid, Dale–Chall) and the Coh-Metrix L2 Reading Index, built on psycholinguistic and cognitive models of reading. Coh-Metrix, which weights word frequency, syntactic similarity, and content-word overlap, classified the levels far more accurately than the surface-feature formulas. When authors simplify by feel, they track the same lexical and cohesive variables that read-time research has shown to predict L2 processing difficulty.
The finding cuts two ways. Author intuition is more principled than skeptics assume; experienced graders are sensitive to referential cohesion, lexical co-occurrence, and predicate density, features surface formulas ignore. At the same time, readability formulas remain inadequate as L2 design tools: a text scoring at Flesch–Kincaid grade 4 may still load heavily on cohesion gaps or low-frequency academic vocabulary. Crossley, Allen, and McNamara (2012) extended the argument: rather than treat simplification as deviation from authentic text, the field should recognise it as comprehensible-input design that approximates what cognitive readability research recommends.
Treat published readability scores as a coarse filter, not a target. The features that distinguish a usable B1 text from an unusable one (pronoun chains a learner can resolve, recurring content words, propositions sequenced without ellipsis) do not appear in Flesch–Kincaid output but do show up in Coh-Metrix indices and in what a careful grader notices on a fourth read. Allen's (2009) corpus work showed intuitive simplifiers reliably reduce embedded and non-restrictive relative clauses, a pattern formulas cannot detect.