How publishers build graded readers: the craft decisions behind vocabulary control, grammar grading, readability measurement, and the treatment of unknown words.
These three concepts are frequently conflated but operate differently in reader design.
Token: any running word in the text. "run", "runs", "running", "ran" = 4 tokens.
Lemma: the base form + its inflections only. "run" covers runs/running/ran = 1 lemma, 4 tokens.
Word family: lemma + all derived forms. "run" covers run/runs/ran/running/runner/runners/runnable = 1 family, many forms. Nation (2016) defines six levels of family breadth, from base-only (Level 1) to all infrequent derivations (Level 6). Research and the AWL use Level 6.
Headword: publisher terminology for a counted unit. Oxford Bookworms and most British publishers use headwords defined approximately as word families (covering inflections + common derivations). So "run" at 400 headwords means a learner needs to recognise runner, running, and ran, but publishers vary in how strictly they apply this.
Key implication for graded reader authors: a 400-headword level does not mean 400 tokens. Each headword represents a family cluster. The author controls the number of headword families deployed, not the raw token count.
| Publisher | Series | Levels | Headword Range | CEFR Approx |
|---|---|---|---|---|
| Oxford | Oxford Bookworms Library | Starter–Stage 6 | 250–2,500 | A1–B2 |
| Penguin/Pearson | Penguin Readers | Starter–Level 7 | 200–3,000 | Pre-A1–C1 |
| Cambridge | Cambridge English Readers | Starter–Level 6 | 250–3,800 | A1–C1 |
| Macmillan | Macmillan Readers | Starter–Upper | 300–2,200 | A1–B2 |
| National Geographic | Footprint Reading Library | 8 levels | 800–3,000 | A2–B2 |
Note: Claridge (2012) found significant inter-series inconsistency. A Bookworms text at 1,400 word families is labelled B1/B2; a Macmillan text at the same count is labelled A2/B1. These mismatches matter when building multi-series libraries.
The New General Service List, Graded Reader edition (NGSL-GR 1.0), was designed specifically for graded reader production. It has 11 bands in 400-word increments (bands 1–8) then 600-word increments (bands 9–11), providing finer-grained control than traditional publisher levels. A text at NGSL-GR Band 4 has 1,600 high-frequency words available. This list is corpus-derived from a 273-million-word corpus and is increasingly used by researchers for profiling reader vocabulary.
Nation (2001) and Hu & Nation (2000) established the foundational finding: a reader needs to know approximately 98% of running words for independent, pleasurable comprehension. At 95% (1 unknown in 20), comprehension becomes effortful and acquisition drops sharply.
Implications:
Coverage thresholds for vocabulary size at 98%:
| Text type | Vocabulary needed (word families) |
|---|---|
| Graded readers (controlled) | 300–2,500 depending on level |
| General fiction | ~8,000–9,000 (Nation, 2006) |
| Newspapers | ~9,000–10,000 |
| Academic text | ~8,000 + AWL |
Introduction: New headwords in well-designed graded readers appear first in a high-support context. The meaning is recoverable from surrounding text (semantic transparency), illustration, or a glossary entry. Cold introduction (using a word without contextual scaffolding) is an authoring error.
Recycling: Nation's research suggests a word needs approximately 10–15 meaningful encounters for productive acquisition. For graded readers, this means a new headword introduced at page 1 should reappear naturally 8–12 more times by the end. Waring & Takaki (2003) confirmed that single-encounter learning from graded readers is modest; repeated encounters drive retention.
Glossing: Three mechanisms:
The Claridge (2012) study noted that Cambridge English Readers deliberately exclude glossaries and support materials, positioning their texts as adult leisure reading. Oxford and Penguin include notes and glossaries. Neither approach is demonstrably superior for acquisition; the choice reflects design philosophy.
Proper nouns present a structural problem: character names, place names, brand names, and cultural references fall outside any headword list but are unavoidable in narrative. Publisher practice varies:
Grammar is graded in parallel with vocabulary. Each publisher uses a grammar syllabus that defines which structures are permissible at each level. The Oxford Bookworms graded grammar syllabus (publicly available) is the most widely cited model:
| Stage | CEFR | Headwords | Key Grammar |
|---|---|---|---|
| Starter | A1 | 250 | Present simple, past simple, basic modals (can/can't), imperatives, simple coordination |
| Stage 1 | A1–A2 | 400 | Past simple, coordination (and/but/or), subordination (before, after, when, because, so) |
| Stage 2 | A2–B1 | 700 | Present perfect, will (future), have to/must/could, comparative adjectives, simple if-clauses, past continuous, tag questions, ask/tell + infinitive |
| Stage 3 | B1 | 1,000 | Should/may, present perfect continuous, used to, past perfect, causative, relative clauses |
| Stage 4 | B1–B2 | 1,400 | Future perfect/continuous, past modals (might have, should have), more complex passives, reported speech |
| Stage 5 | B2 | 1,800 | Full range of conditionals, complex noun phrases, embedded clauses |
| Stage 6 | B2–C1 | 2,500 | Near-native grammar range; archaic forms acceptable in classics |
The key design principle is no structures above the level. An author writing a Stage 2 text cannot use the past perfect even once. This discipline is harder than vocabulary control because English grammar is recursive; sophisticated meaning often requires complex syntax.
Empirical studies of graded readers (Claridge, 2005; Grabowski, 2015) yield approximate sentence length norms:
| Level | CEFR | Mean sentence length (words) | Max clause depth |
|---|---|---|---|
| Starter | A1 | 7–10 | 1 (simple/coordinate) |
| Stage 1–2 | A1–A2 | 10–13 | 2 (one subordinate) |
| Stage 3–4 | B1 | 13–17 | 2–3 |
| Stage 5–6 | B2 | 16–22 | 3–4 |
| Authentic adult fiction | N/A | 18–25 | Unrestricted |
Syntactic depth (number of clause embeddings per sentence) is as important as raw sentence length. A 20-word sentence with flat coordination ("She ran and fell and cried and looked up") is more accessible than a 12-word sentence with heavy embedding ("The man she'd once trusted had gone").
Flesch Reading Ease (Flesch, 1948): Based on average sentence length (ASL) and average syllables per word (ASW). Score 0–100; higher = easier. Designed for native English readers. Not calibrated to CEFR. A Flesch score of 60–70 corresponds roughly to plain English for adult native readers; EFL texts typically target higher scores (70–80+) even at B2.
Flesch-Kincaid Grade Level: Converts the above to US school grade. Grade 5 ≈ 10-year-old. Again, calibrated for L1 readers; should be used cautiously for L2 material. A graded reader at 400 headwords will often score Grade 3–5 FK, but this does not mean it is appropriate for 8-year-olds.
Lexile Framework: Used primarily in US education. A Lexile score is derived from word frequency and sentence length. Lexile 500–700 = roughly A2–B1; 1000+ = C1. Lexile is increasingly used by ELT publishers but was designed for L1 comprehension, so it systematically underestimates difficulty for L2 readers who lack background vocabulary even at "easy" Lexile levels.
CEFR-J Readability Index (CEFR-J Rater): Specifically designed for EFL contexts. Developed at Tokyo University of Foreign Studies for Japanese learners but applicable cross-linguistically. It integrates lexical frequency bands, syntactic complexity measures, and text length. More aligned with L2 reading behaviour than Flesch or Lexile.
Limitations of all formula measures: Readability formulas are proxy measures. They do not directly assess:
For graded reader QA, formula measures are used as a screen, not a verdict. A text that passes a formula check still needs expert editorial review.
Krashen's Input Hypothesis (1982) states acquisition occurs when learners encounter input at i+1, one level above current competence. Graded readers operationalise this by:
The practical implication: a good graded reader at Bookworms Stage 2 should be genuinely comfortable for someone who just completed Stage 1, not someone who aspires to Stage 2. The unknown words should feel like discoveries, not obstacles.