Draft v2.1Framework 2NAPE-alignedTeam NASS

অক্ষর থেকে অর্থবাংলা পড়ার স্কেল · লেভেল ০ to ৬, in the order of NAPE's five reading components

One shared way to describe how well anyone reads Bangla, from a learner who can only hear the sounds in words to an expert reader. Seven levels and one number per learner, with 17 blocks inside them for facilitators, in three groups. Each level is one of the five components Bangladeshi teachers are trained on1 (ধ্বনি, বর্ণ, শব্দ, সাবলীলতা, বোঝা), so a result tells a facilitator what to teach next.1,2,4

The whole scale at a glance

নতুন পাঠকস্বাধীন পাঠকদক্ষ পাঠকNot yet tested
নতুন পাঠক New reader · লেভেল ০ to ২
স্বাধীন পাঠক Independent reader · ৩ and ৪
দক্ষ পাঠক Skilled reader · ৫ and ৬

Click any block to see what it means, how the check finds it, and what to teach next.

01NAPE's five components, in order

The levels follow the order NAPE uses in primary teacher training (DPEd). The same order is the "Big 5" of reading research2 and the subtasks of EGRA4. A learner moves up one component at a time.

০ধ্বনিHearing the sounds in spoken words
১বর্ণLetters and their sounds
২ – ৩শব্দDecoding words, then conjunct words in short sentences
৪সাবলীলতাAccurate reading at a steady pace, phrase by phrase
৫ – ৬বোঝাFacts first, then implied meaning, opinion and trust

02Every level and block

The reported result is one level, such as ৪. Facilitators also see the block, such as ৩খ, which they use to form groups. Learners never see the block. Level rule: the level of the highest block passed, with every lower block passed or skipped, as in ASER's single ladder7. The steps inside ২ and ৩ (কার before যুক্তবর্ণ) follow research on reading akshara scripts such as Bangla6.

03What one result looks like

One level, with the block for the facilitator, plus flags that never change the level, as ASER and EGRA report results4,7. This is the facilitator's record; the learner hears only the last line.

Code L-0412 · age band 15 to 187 Nov 2026 · self-rating 🌳 · facilitator F-07
০১২৩৪৫৬
লেভেলReading level · block ৩খ
৩
শোনা-বোঝাListening · separate flag
৫
বাংলা সংখ্যা ✓English digits ✓English words ✗পড়া: word by wordPath: ৩খ ✓ · ৫ক ✗ · ৪ক ✗
Told to the learner: "আপনি ছোট বাক্য পড়ে বুঝতে পারেন। এখন ডাক্তার, পরীক্ষার মতো কঠিন শব্দ আর ছোট মেসেজ পড়া ধরব।" (You can read short sentences and understand them. Next, we'll read hard words like doctor and exam, and short messages.)

লেভেল ২ · শোনা-বোঝা ৫

Understands well when read to. The only barrier is the script.

লেভেল ৫ · বোঝা add-on ✗

Finds facts but misses why. Teach inference before longer texts.

লেভেল ৪ · পড়া word by word

Answers right but reads slowly. Check placement; practise phrase reading.

04Who sees the levels

Public scale, private results. Public "can do" lines follow CEFR and LASLLIAM13,14; results stay low-stakes, as in ASER and TaRL7,8.

PublicThe scale

  • What each level means
  • Sample texts and "can do" lines
  • Shared by programmes, NGOs and government

PrivateA person's result

  • Recorded under a code; no name or phone number
  • Block, path and flags seen by the facilitator and in programme reports
  • Learners see the level, one strength and one next step; no percentile

NeverGates

  • No certificate, no pass mark
  • Never shared with employers
  • Never used for entry, stipends or jobs

05How the check runs

Two ways to place a learner on the same scale: a 10-minute oral check with a facilitator, one-on-one like ASER and EGRA4,7, or the self-run, voice-guided check in the app.

Facilitator · 10 minutes

1 minকথা · TalkWhat they read in daily life. Nobody fails; they can stop anytime.
StartSentence card (৩)For anyone 12 or older. Never start an older learner on sounds or letters.
MoveUp or downPass: go up. Miss: go down. One retry on a parallel card. Stop after 2 misses at a level.
2 minData, not marksAt ৪: correct words a minute. Below ৪: a story read aloud for শোনা-বোঝা.
1 minFlags and closeDigits and English print. End on a strength.
Pass ৩ → short text (৪) → passages (৫, ৬)৪: read aloud, about 90% of words, phrase by phrase or better. ৫ and ৬: 2 passages each, 5 open questions per passage, spoken, written or pointed answers. Placed at 7 of 10.
Miss ৩ → word card (২) → letter card (১) → sound card (০)Word and letter cards of 8 items, pass at 6; the word card has 2 made-up words. Sound card: 4 spoken items, 3 right, no print.

Self-run app · voice-guided · about 6 to 11 minutes

StartWelcome and practice"এটা পরীক্ষা না।" Sound check, one practice tap, "জানি না" explained.
Self-ratingFirst block🌱 ১খ · 🌿 ২খ · 🌳 ৩খ · 🚀 ৫ক
BlocksRoutedJumps grow while blocks pass, then the gap halves. Up to 7 blocks; stops when decided; one near-miss retry.
Add-onsRecorded onlyবোঝা add-on after ৫ক. Listening below ৪. Digits and English.
EndScore cardLevel, group, a road from ০ to ৬, one strength, one next step, progress since last time.

Scoring rules

  • A pause of 3 seconds counts as an error; the facilitator says the word.4
  • Self-corrections count as correct.
  • Accent and dialect are never errors (শ, ষ, স merged; Chittagonian or Sylheti sounds).4
  • Made-up words (পাফা, নিটো) show real decoding, not guessing.4,5

Block pass marks (app)

BlocksPass
০ · ১ক3 of 4
১খ · ২ক5 of 6
২খ to ৫ক4 of 5
৫খ to ৬ঙ3 of 3 (2 = "প্রায়", not a pass)

Reading rubric for ৩ and ৪11,12

ReadsMeans
Word by wordStill at ৩
Phrase by phrase৪
Whole sentences, with expressionReady for ৫

06What makes a text belong to a level

Levels are defined by features of the text, not the kind of document, so the check measures reading rather than familiarity, as in the Global Proficiency Framework's text descriptors10. Check texts use general topics; work-specific texts are for teaching.

LevelLengthSentencesযুক্তবর্ণLanguage
০
Spoken words only
None·Everyday spoken words
১
Letters
None·Letters
২
Words, 2 to 4 letters, made-up words
None·Everyday words
৩
Conjunct words; 3 to 4 sentences
6 words or fewer●●Everyday চলিত
৪
30 to 150 words
10 words or fewer●●●●Everyday চলিত
৫
100 to 400 words
Some longer, joined●●●●●●Everyday and semi-formal
৬
Any
Complex●●●●●●●●Opinion, persuasive, formal, literary, legal, academic, সাধু ভাষা

Bars and dots show relative size only. All lengths and counts are starting points for the pilot. Today's passages are shorter than the spec (job ad 52 words, news 97); longer ones are a gap to fill before the pilot.

07Flags: recorded, never part of the level

The level measures reading Bangla script. Flags record what else a learner can do, without changing the level.

শোনা-বোঝা · Listening

For learners below ৪: a story read aloud, picture answers, in up to three stages (শ১ to শ৩), 2 of 3 per stage. Shows whether understanding is there and only the script is in the way: reading = decoding × language comprehension3. EGRA has the same subtask4.

পড়া · Reading aloud

Correct words a minute at ৪, and word by word, phrase by phrase or whole sentences. Data only; no official Bangla benchmark yet11,12. In the app, read-aloud scoring by speech-to-text waits on accuracy and data-retention checks.

সংখ্যা and English

Bangla digits (৳৪৫, বাস ৮) and English digits (Tk 45, Bus 8), credited separately, plus common English print such as Send Money and Cash Out.

বোঝা add-on

After a ৫ক pass: two "why" questions and a scam SMS. Separates finding facts from understanding why, like MICS's literal and inferential questions9.

08From results to groups in one room

Levels and blocks let a facilitator form three or four groups with neutral names, never numbers. Grouping by level rather than grade is the core of Teaching at the Right Level8. An example room of 18 learners:

পদ্মা

০ and ১ · blocks ০ to ১খ
Sounds, letters from own name, shop signs

মেঘনা

২ · blocks ২ক to ২গ
কার, ফলা, simple words

যমুনা

৩ · blocks ৩ক to ৩গ
Conjunct words, short messages

তিস্তা

৪ and above · ৪ক up
Fluency, notices, finding information

09Re-checks over time

Each block has a retry set, and the bank grows towards about 12 texts per level, so forms are assembled fresh and texts spread less in a community. The app shows progress since the last check on the same device.

Day 1Full check
place and group
Days 1 to 10Teaching cycle 1
level-group practice
Day 11Progress card
3 minutes, one block at level and one above
Days 11 to 20Teaching cycle 2
Day 21Full re-check
new form, regroup

10Running it safely

From the team's Reading Steps v0.1, plus what the voice and the app add, in line with UNICEF's guidance on research involving children15.

Setting and consent

  • Side by side, private from peers but in sight of another adult.
  • Never behind a closed door with a minor; offer girls a female facilitator.
  • Not at the workplace or in front of supervisors.
  • Assent from every learner; guardian consent for under-18s, arranged before the day.
  • Read-aloud recording: honest consent wording, since the voice service keeps data unless zero-retention is confirmed. Under-18s skip it in the self-run app.

Distress and data

  • If a learner is upset: an easy item, end on a success, offer to stop.
  • "জানি না" on every scored item; the same neutral star after every answer.
  • A code, not a name. Never record employer, address, photos or the word নিরক্ষর.
  • The tool applies fixed rules; no AI in placement.
  • Results stay on the device.

11Writing items for the check

Text

  • Bangladeshi vocabulary (পানি, not জল) and Bangla Academy spelling.
  • Bangla digits in Bangla texts; English only in flag items.
  • General, familiar topics: market, journey, family, health, phone.
  • No conjuncts in print before block ৩ক.
  • Matched length and conjunct count across forms; no brand names or logos.
  • The right answer is never the longest option.

Voice and screen

  • The voice says only the instruction or question, never the text being tested.
  • Hear-and-tap captions always read "শুনুন, তারপর ক্লিক করুন", never the spoken target.
  • Picture choices carry no text labels, except at ০ and in listening.
  • আপনি in every learner-facing line; a chat message may keep তুমি or তুই.
  • Every conjunct print-tested in a full-coverage Bangla font.

12Internal reference points

Internal only, never on learner, employer or public materials

  • Rough school match: ০ and ১ before Grade 1 · ২ Grade 1 · ৩ Grades 2 to 3 · ৪ Grades 4 to 5 · ৫ Grades 6 to 8 · ৬ Grades 9 to 12 and university.
  • Only about half of Grade 3 and 5 pupils in Bangladesh read at grade level, so a grade is not a fixed standard.
  • MICS foundational reading (a short Grade 2-level story) sits roughly at the ৩ to ৪ boundary.9
  • Any of these needs a linking study before it is claimed.

13References: where the structure comes from

These show the structure is not invented. They do not validate the cut-offs; only pilot data does that. Numbers in the text, such as 4, point here.

  1. NAPE (National Academy for Primary Education). Diploma in Primary Education (DPEd) teacher-training materials on teaching Bangla reading, Mymensingh, Bangladesh.
    The five components and their order, in the language Bangladeshi teachers already use. Page reference still to add before citing.
    Used in: Levels ০–৬, section 01
  2. National Reading Panel (2000). Teaching Children to Read: An Evidence-Based Assessment of the Scientific Research Literature on Reading. NICHD, NIH Pub. No. 00-4769. www.nichd.nih.gov/publications/pubs/nrp/smallbook
    The five components of reading ("Big 5"): phonemic awareness, phonics, fluency, vocabulary, comprehension.
    Used in: Section 01, level ০
  3. Gough, P. B. & Tunmer, W. E. (1986). Decoding, reading, and reading disability. Remedial and Special Education 7(1), 6–10. doi.org/10.1177/074193258600700104
    The Simple View of Reading: reading comprehension needs both decoding and language comprehension.
    Used in: শোনা-বোঝা flag
  4. RTI International (2016). Early Grade Reading Assessment (EGRA) Toolkit, 2nd edition. USAID. shared.rti.org/content/early-grade-reading-assessment-egra-toolkit-second-edition
    Each component as a subtask (sounds, letters, made-up words, oral reading fluency, comprehension, listening) and the scoring rules: 3-second pause, self-corrections, dialect. Used in Bangladesh.
    Used in: Levels ০–৪, scoring rules, flags
  5. Gove, A. & Wetterberg, A. (eds.) (2011). The Early Grade Reading Assessment: Applications and Interventions to Improve Basic Literacy. RTI Press. www.rti.org/rti-press-publication/early-grade-reading-assessment
    How EGRA results are used to plan teaching, including made-up word decoding.
    Used in: Level ২
  6. Nag, S. (2017). Learning to read alphasyllabaries. In L. Verhoeven & C. Perfetti (eds.), Learning to Read across Languages and Writing Systems. Cambridge University Press. doi.org/10.1017/9781316155752
    Akshara scripts such as Bangla are learned in steps: simple akshara, vowel signs, then conjuncts. Backs blocks ২ক–৩গ.
    Used in: Levels ২ and ৩
  7. ASER Centre. Annual Status of Education Report and the ASER reading tool (letter, word, paragraph, story). asercentre.org
    One ordered level per learner from a short one-on-one oral check; the highest level passed is the result. Uwezo uses the same design in East Africa.
    Used in: Level rule, check, result
  8. Banerjee, A., Banerji, R., Berry, J., Duflo, E., Kannan, H., Mukerji, S., Shotland, M. & Walton, M. (2017). From proof of concept to scalable policies: challenges and solutions, with an application. Journal of Economic Perspectives 31(4), 73–102. www.aeaweb.org/articles?id=10.1257/jep.31.4.73
    Randomised evidence that grouping learners by level (Teaching at the Right Level) raises reading. See also teachingattherightlevel.org.
    Used in: Groups
  9. BBS & UNICEF (2019). Progotir Pathey Bangladesh: Multiple Indicator Cluster Survey 2019. www.unicef.org/bangladesh/en/reports/progotir-pathey-bangladesh
    The Foundational Learning Skills module: a short Grade 2-level story with literal and inferential questions. The local anchor at the ৩ to ৪ boundary; its two question types mirror ৫ and ৬.
    Used in: Levels ৫ and ৬, বোঝা add-on, internal points
  10. UNESCO Institute for Statistics, USAID, World Bank and partners (2020). Global Proficiency Framework: Reading. tcg.uis.unesco.org/wp-content/uploads/sites/2/2021/03/Global-Proficiency-Framework-Reading.pdf
    Global minimum proficiency levels for SDG 4.1.1, with text-complexity and comprehension descriptors (retrieve, interpret, evaluate).
    Used in: Text features, levels ৫–৬
  11. NCERT (2022). Foundational Learning Study 2022, under NIPUN Bharat, Ministry of Education, India. nipunbharat.education.gov.in/fls/fls.aspx
    Oral reading fluency and comprehension benchmarks for Grade 3 in 20 Indian languages, set by policy linking: a model for a Bangla words-a-minute cut-off.
    Used in: Level ৪, পড়া flag
  12. Hasbrouck, J. & Tindal, G. (2006). Oral reading fluency norms: a valuable assessment tool for reading teachers. The Reading Teacher 59(7), 636–644. doi.org/10.1598/RT.59.7.3
    Words correct per minute as a fluency measure, read against norms rather than as a pass mark.
    Used in: Level ৪, rubric
  13. Council of Europe (2020). Common European Framework of Reference for Languages: Companion Volume. www.coe.int/en/web/common-european-framework-reference-languages/cefr-companion-volume
    A public level scale with "can do" descriptors, the model behind IELTS-style reporting.
    Used in: Who sees the levels
  14. Council of Europe (2022). Literacy and Second Language Learning for the Linguistic Integration of Adult Migrants (LASLLIAM). www.coe.int/en/web/lang-migrants/laslliam
    Literacy levels for adults and teens who are new readers, below CEFR A1.
    Used in: Who sees the levels
  15. Graham, A., Powell, M., Taylor, N., Anderson, D. & Fitzgerald, R. (2013). Ethical Research Involving Children. UNICEF Office of Research – Innocenti. www.unicef-irc.org/publications/706-ethical-research-involving-children.html
    Assent, guardian consent, privacy and avoiding distress when assessing under-18s.
    Used in: Running it safely

Links checked 7 Nov 2026. Still to add: the NAPE / DPEd manual page.

14Validating it

Expert reviewNAPE / DPEd trainers
and a Bangla teacher read every block
PilotReal learners
blocks are ordered: passing ৪ means passing ৩
AgreementTwo facilitators
same learner, same level
CompareEGRA or MICS items4,9
results line up
BenchmarkWords a minute
a Bangla cut-off for ৪

15What changed

Wasv2.1Why
v2.0: one number, no steps inside a level17 blocks inside the levels (০, ১ক … ৬গ, ৬ঘ, ৬ঙ)Facilitators group by block; the learner still gets one level
v2.0: ৬+ parked, no itemsOne level ৬, with সাহিত্য ও সাধু (৬ঘ) and আইন ও গবেষণা (৬ঙ) as its top two blocksStrong readers get a ceiling inside the same level; still not yet tested
v2.0: interest pick and self-ratingSelf-rating picks the first block; routing in growing jumps, then halving; up to 7 blocksShorter check, same placement
Score card with a percentileLevel, group, road, one strength, one next step, progressThe percentile used invented data
তুমি in learner linesআপনি everywhereRespect for teens and adults
v1.4: 7 levels, 16 sublevels, starts at printLevels ০ to ৬ from v2.0, with ধ্বনি at ০ and made-up words at ২NAPE's component order

Mapping v1.4 results: ১.১–১.২ → ১ · ২.১–২.৩ → ২ · ৩.১–৩.৩ → ৩ · ৪.১ → ৪ · ৪.২–৫.১ → ৫ · ৫.২–৭.২ → ৬.

16Still open

Pilot

All cut-offs, text lengths and block boundaries need testing with real learners. ৬ passages, including ৬ঘ and ৬ঙ, need trying with good readers.

Read-aloud and age

Speech-to-text scoring and the under-18 tap wait on accuracy, cost and zero-retention checks. Consent wording to confirm.

Content gaps

Own-name card, retry sets for ২ক and ৪ক, longer ৫ passages, an academic text, and a teacher's review of every line.