English reading practice bands, explained

What MindCraft practices

This construct separation follows the original decoding/comprehension model and evidence-rated K–3 guidance; neither source validates a MindCraft score or one-sentence decision. [EN-CONSTRUCT-01]

The ten English practice bands

The number of bands and their boundaries are MindCraft design decisions, versioned so changes can be tested and explained. Bands 7-10 require non-intuitive spellings such as through and enough. A hard word-count ceiling prevents one strong session from suddenly producing a much longer sentence. [EN-BANDS-01]

Computed features are screening gates, not proof of item difficulty. Readability research cautions that surface formulas omit important lexical, syntactic, cohesion, and semantic information. [EN-TEXT-01]Meaning, naturalness, Minecraft accuracy, and background-knowledge demand still require human and learner evidence.

How a sentence is chosen

1. 1. Freeze the current band. Each band has a hard envelope for length, syllable load, word coverage, morphology, and clauses.
2. 2. Use minimal game context. A looked-at block or entity, biome, dimension, time, and weather can inspire the item. MindCraft does not need chat, signs, coordinates, world seed, server identity, or player names for sentence generation.
3. 3. Generate and check. English candidates are checked deterministically against the active English envelope. Unknown or out-of-band context words cannot bypass it. [EN-LEXICON-01]
4. 4. Keep uncertainty honest. A generated item is experimental until human and learner evidence supports it; a frozen fallback bank remains separate from calibrated assessment claims.

How attempts affect difficulty

Gameplay credit and evidence for increasing difficulty are separate. Every completed attempt advances play, while the durable pathway-specific progression state treats the current single-transcript result as neutral for promotion. Difficulty is frozen for a capped session, repeated difficulty signals can make a later session easier, and any automatic increase requires qualifying evidence across at least two sessions and moves only one band. Parents can apply or clear an immediate band override without rewriting the learner's mastery history.

Speech recognition can be wrong, especially with child speech, dialects, accents, noise, and second-language speech. Research has found substantial disparities in commercial ASR systems; that study used adult speech and other vendors, so it is a validation warning rather than a measurement of MindCraft. [EN-ASR-01] MindCraft is intentionally slightly soft when uncertain and does not require a learner to repeat an item.

Four pathways, separate engines

This page documents native-literacy English. Practice-language Spanish uses a separate Spanish-native tokenizer, vocabulary policy, syllable and orthographic features, item bank, hard gate, and Spanish methodology page. English→Spanish and Spanish→English use their own second-language progressions layered over those language-owned text envelopes; their evidence, scaffolding, framework attribution, and limits are documented in the English–Spanish L2 methodology.

Known limitations

• The band boundaries and feature limits are product hypotheses, not validated school-grade mappings.
• The current frozen English item set is small; specialist review and representative learner calibration are pending.
• The conservative multi-session promotion contract is not yet persisted at session boundaries; current product points still update after individual attempts.
• Production vocabulary data combines SUBTLEX-US frequency rank, CMUdict pronunciations/syllables, and Dolch flags. Exact source snapshots and license provenance are still being reconstructed, so these features are screening inputs, not curriculum claims.
• SUBTLEX-US reflects adult film subtitles; frequency there is not the same as child familiarity or printed-word exposure.
• Automatic checks cannot reliably judge meaning, naturalness, factual Minecraft fit, cultural fit, or all background knowledge.
• MindCraft has not yet published reliability, external-criterion, fairness, comprehension, or learning-outcome studies.
• MindCraft does not diagnose dyslexia, speech/language conditions, or any learning disability.

Claim and evidence registry

Version methodology-evidence-1.0.0. Each record limits the claim to the source's scope and labels product decisions separately.

Version and change log

Methodology

english-native-1.2.0

Published

August 14, 2026

Review owner

MindCraft product owner; external English literacy review pending

Changes

Expanded to ten age-independent practice bands. Bands 1-5 keep their score boundaries. Bands 7-10 require irregular spellings such as through and enough. Hard sentence envelopes, construct separation, conservative multi-session promotion, frozen golden items, and explicit validation limits remain.