LevelReads · content specification

How a passage
gets its Level

What a passage must look like at each of twelve levels — and the method that produces one.

Grounded in Specifying a Passage, the evidence brief.

scroll ↓

why the spec has this shape

One equation,
two unknowns

FK = 0.39 × words/sentence
    + 11.8 × syllables/word − 15.59

Two inputs, one output. Any FK score is satisfiable an infinite number of ways, by trading sentence length against word length.

Level 6
20.6-word sentences
of nursery vocabulary
both score
6.0
Level 6
8.5-word sentences
of polysyllabic vocabulary

So the specification fixes both inputs separately and treats the FK score as a check, never as the target. Lexile and ATOS both work this way.

the boundaries of the instrument

Where FK
stops working

Levels 1–2 · no formula is valid

Common Core's alignment table has no K–1 row; those texts were placed by teachers. Applied to Little Bear, formulas span grade 1.5 to 7.5. Below grade 1 the specification uses word‑list coverage instead.

Poetry · outside every formula

MetaMetrics does not measure poetry, drama, song lyrics or lists; Coh‑Metrix excluded them from its corpus. Poetry therefore carries no FK gate at any level — it is specified by line, syllable and word‑list constraints only.

Genre · a 2.3-grade split

Measured by ETS, FK over‑rates informational text by ~1.28 grades and under‑rates literary text by ~1.05. The specification therefore carries two sentence‑length bands per level, narrative and informational, rather than one.

Levels 10–12 · every metric flattens

Nelson et al. found metrics discriminate worse at grades 9–11 than at 3–8. At this end the specification adds Academic Word List density and a written qualitative rubric, because no formula separates these three shelves.

the alternative we cannot use

Why not Lexile

1 · It is licensed

Lexile is a proprietary framework owned by MetaMetrics. The scale, the analyser and the word‑frequency corpus are commercial products. Printing a Lexile number on a passage, a report or a parent dashboard requires a licence agreement.

Flesch‑Kincaid was developed under U.S. Navy contract, became a Military Standard, and is a government work in the public domain. We can compute it, name it and show it to a parent with no permission from anyone.

2 · It would not close the gap anyway

The licence settles it; the evidence would not have rescued us. The gap FK leaves is a gap Lexile leaves too — which is why the specification adds word lists rather than swapping formulas.

short reference

The instruments

What sits alongside FK, and why each earns a place.

Dolch 1936 · public domain implemented

220 service words + 95 picture nouns a child is taught to recognise on sight. Used as a coverage percentage.

No readability formula is validated below grade 1; word‑list coverage is what Fountas & Pinnell and Reading Recovery rest on.

Fry 1980 · public domain need list

First 100 / 300 / 1000 words by frequency. The first 100 account for about half of all written English.

Gives Levels 2–4 a coverage target that keeps rising once Dolch is exhausted.

Spache 1953 need list

A readability formula built for grades 1–3: sentence length plus the share of words outside a 1,041‑word familiar list.

The published guidance has always been Spache below grade 4, Dale‑Chall at grade 4 and above — a boundary drawn decades ago that lands exactly where FK starts working.

Dale‑Chall 1948, rev. 1995 need list

Sentence length plus the share of words outside a 3,000‑word list known to 80% of fourth graders.

Measures familiarity, which FK cannot. FK counts syllables, so it reads “tectonic” as hard whether or not the child has met it.

CWF — Critical Word Factor derive

A count per 100 words of words outside the reference list — not a percentage threshold.

The familiar “95% / 98% known words” rule is adult second‑language research. Its originating study never tested 98%; a 661‑subject replication found no threshold; the two large native‑speaker studies land at 56–59%. A count is defensible; a threshold is not.

AWL — Academic Word List Coxhead 2000 need list

570 word families common across academic writing, measured as density.

The one signal still moving at Levels 10–12. Academic prose runs ~10% AWL; fiction ~1.4%.

the level system

Twelve levels,
one public measure

Level N means the passage measures FK N. That number is what a parent sees and what the app sorts on, at every level. What changes by band is what decides whether a passage belongs there.

LevelGradeBand
1Kage 5emergent
21age 6emergent
32age 7transitional
43age 8transitional
54age 9core
65age 10core
76age 11core
87age 12core
98age 13core
109age 14upper
1110age 15upper
1211age 16upper

The scale ends at 12. The audience is 7–14, and the published alignment places FK 13 inside the grade 11‑and‑beyond band rather than above it.

Levels 1–2 keep the FK number for continuity in the interface, but FK does not gate them.

what we measure a reader on

Three axes,
not one

A Level is one number, and one number cannot place a reader. A child sits somewhere in a space with three dimensions, and a decision about what to give them next needs all three.

X Text difficulty

How hard the prose is. Sentence length, word familiarity, word‑list coverage — the whole of the specification in this document. This is what the Level number carries, and it is the only axis a formula can reach.

Y Comprehension

Not whether they finished it — whether they understood it, and at what depth. Split in two: literal (Remember, Understand) asks whether the text went in; inferential (Apply, Analyze, Evaluate, Create) asks whether they can do anything with it. A reader can clear one and fail the other, and the two mean different things.

Z Breadth

Which kinds of writing they have actually met at this level. A child who has read six fables has not shown the same thing as a child who has read a fable, a poem, an explanation and an argument. Performance on a story tells you something different from performance on an expository text — measured at grade 4, comprehension of narrative runs 23 points above expository for the same readers.

Why all three

Difficulty alone says a passage was the right size. Difficulty and comprehension say the reader handled one text. Only breadth says they can handle the level — because the categories are not interchangeable, and a reader who has only ever met one of them has been tested on one of them.

the specification · 1

What decides
each band

Levels 1–2 emergent

decided by
Word‑list coverage · no FK gate
word list
Dolch pre‑primer & primer at 75% coverage or above; Fry first 100–300
new words
per 100
1–3
also fixed
Unique words per 100 held at 20–25 · only phonics patterns taught to date · a repeated sentence pattern carried through the passage

Level 3 transitional

decided by
Spache decides · FK cross‑checks
word list
Spache 1041‑word list; Fry first 300
new words
per 100
3–5
also fixed
Mean sentence length inside band · multisyllabic share near 48%

Level 4 transitional

decided by
Spache and FK together
word list
Spache; Fry 1000
new words
per 100
5
also fixed
Multisyllabic share near 65% · morpheme count per word

Levels 5–9 core

decided by
FK decides, banded separately per genre
word list
Dale‑Chall 3000; frequency bands 1–3
new words
per 100
5
also fixed
Narrative band runs ~2.3 grades below informational · mean and range of sentence length · age‑of‑acquisition and concreteness spread

Levels 10–12 upper

decided by
FK plus AWL density plus a written rubric
word list
Academic Word List (10% academic‑prose, 1.4% fiction)
new words
per 100
5
also fixed
Deep cohesion · abstractness · knowledge demand recorded per passage

Established: the instrument assignment, the word lists, the multisyllabic and AWL reference figures. Inference: the new‑word ceilings — a synthesis of source ranges, not a published specification.

the specification · 2

Length,
sentences, time

Attention span by age is not evidenced — the circulating figures trace to unsourced blogs. Reading time is computable, so time sets the length, and every level lands in a single sitting on a phone.

Lwordssent
nar
sent
info
min
160–904–64–62.5
2120–1606–86–82.3
3160–2008–109–111.8
4200–2409–1110–122.0
5250–29010–1211–132.0
6290–33011–1312–142.1
7330–38012–1413–152.4
8400–45013–1514–162.8
9430–48014–1615–173.0
10500–56015–1716–182.9
11550–61016–1817–193.1
12600–66017–1918–203.1

Sentence bands for Levels 3–6 are Hiebert (TextProject RRR #11.03, Table 2), the only per‑grade table found, genre‑split and flagged preliminary by its author. Levels 1–2 are our measured floor; Levels 7–12 continue the Hiebert trend and are inference. Rates for Levels 2–9 are Hasbrouck & Tindal oral‑fluency norms, 50th percentile, spring; Level 1 and Levels 10–12 are silent‑reading estimates and the softest numbers here. These are means over the passage. No maximum sentence length is specified — none is evidenced at any grade.

the specification · 3

Sentence
shape

A mean alone produces flat prose — every sentence the same length, which reads as machine‑made. The specification asks for a spread: most sentences near the mean, several short, one or two long enough to carry a whole idea.

Lshortmeanlong
1259
24713
34916
451018
561120
661222
761323
871425
981527
1081629
1181731
1291832

How to read it. Aim for roughly one sentence in five short and no more than one in five long. The mean is what the band gates on; short and long are the shape it should reach that mean by.

The long sentence is not a defect

The long end is a target, not a cap — no maximum sentence length is evidenced at any grade, and the specification carries none. A long sentence is where a cause and its effect sit together, joined by because or so. Splitting it into two short ones removes the join and, in the one study that tested it directly, made the text harder rather than easier.

the specification · 4

Voice, and
the join

Everything specified so far is countable — length, sentence shape, word lists, question mix. None of it produces good writing. A passage can satisfy every band and still read as though it were assembled rather than written.

The join — required at every level

Each sentence must pick up something from the one before: a pronoun, a repeated noun, a connective, a consequence. A reader should never feel a gap between two sentences.

This is a floor, not a dial. Measured across thousands of texts, cohesion does not rise or fall with grade — it correlates with grade at about .05, against −.67 for sentence simplicity. So it is not something to increase as levels climb. It is something every level owes the reader.

It matters most where vocabulary is tightest. A passage built from 188 sight words is where prose most easily falls into disconnected statements, because the writer has the least room to manoeuvre.

The text stands alone — there are no illustrations

These passages ship as text. Nothing may rest on a picture, and a reader must never have to supply a fact the passage did not give them.

Every step of the causal chain is on the page. If one thing follows from another, the first was stated. A mystery must be fair: the situation is established before it is called strange, the clue reaches the reader when it reaches the character, and the solution follows from what was shown rather than from a fact produced at the end.

No sentence is padded to reach a band. If a sentence should be longer, it is joined to an idea it belongs with. A reader can tell the difference between a sentence that grew and one that was stretched.

What each band should sound like

Levels 1–2 · song

Rhythm a child can feel. Rhyme where it lands naturally — internal rhyme, a refrain that returns, a line that comes back changed. Repetition should read as a chorus, not as a shortage of words. The test is reading it aloud.

Levels 3–4 · told, not reported

A story with a teller. Events connect by cause and consequence rather than sitting in sequence. Something is at stake, even if it is only a missing library book.

Levels 5–9 · explaining

The voice of someone who knows a thing and wants you to know it. New words arrive where they are needed and are explained in the sentence that carries them, not in a glossary.

Levels 10–12 · argued

A position taken and defended, with the strongest objection to it stated fairly before it is answered.

This is the part a person has to judge

Voice and cohesion are the only elements of this specification that cannot be gated by measurement. Every professional levelling system reaches the same conclusion — the ones that skip the human step do so by refusing to measure the text types where judgement is unavoidable.

the specification · 5

Inside
one passage

What a single passage contains, derived from the length and sentence bands above.

Lparasentnew
words
Qliteral /
inferential
141525100 / 0
262045100 / 0
35199675 / 25
462111675 / 25
572314860 / 40
672516860 / 40
772618850 / 50
882921850 / 50
982923840 / 60
1093226840 / 60
1193329835 / 65
12103432835 / 65

New words is the CWF ceiling applied to the passage length — the count of words outside that level's reference list. These are the vocabulary words the passage teaches, and they carry the tap‑for‑definition entries. Literal is Remember and Understand; inferential is Apply, Analyze, Evaluate and Create.

Shelf structure

Each level holds a fixed set of categories with three passages in each, so a child can read one of every kind before moving up. Categories widen with the level: four at the emergent end, rising to seven or eight from Level 5.

Poetry is specified by line count, syllables per line and word‑list coverage. It carries no FK number and no sentence band.

the method · gates

What each
band checks

Levels 1–2

Levels 3–4

Levels 5–9

Levels 10–12

Poetry, any level

example · emergent band

Level 2

realistic fiction · grade 1 · draft 2 of 3 · cleared 7/7 gates

Goodbye, Tooth!

I sit at my school table. My tooth is loose today. I wiggle it, and it wiggles back. Wiggle, wiggle, little tooth!

It wiggles, and out it comes! I look at the gap. My hand finds the gap too. I like my little gap.

So I run to the children. I say, “Look, my tooth came out!” They say, “I want to see!” We laugh, and we play.

Then I put the tooth in my box. The box is safe and small. Wiggle, wiggle, goodbye tooth—now you ride home with me.

At home, I show my mother. She says, “You are big now!” I put the box on my bed. Wiggle, wiggle, out it came. My gap, my box, my home.

words122
mean sentence7.2 (6-8)
short present12/17
long tail2
questions5
literal share100%
Dolch77%

Shape as written: 17 sentences, shortest 4 words, longest 18, mean 7.2. Score: FK 0.6 raw, 1.6 once genre is accounted for. Questions: 5, 100% literal — remember, understand. New words: —.

example · transitional band

Level 4

mystery · grade 3 · draft 6 of 3 · cleared 6/6 gates

The Missing Book

On Tuesday afternoon, Maya carried The Secret of Star Island to the front desk. She wanted to check it in before class. Mister Chen scanned the barcode, then set the book beside his computer. Twiggy the parrot rattled his cage latch, the way he always did near something shiny. A loud crash rang out near the art table. Mister Chen hurried off to help. Maya turned to look, too. When he came back, the book was gone.

Then Maya noticed something peculiar. A blue feather lay on the counter, matching the parrot on her book’s cover. She hesitated; the feather seemed too strange to ignore. So she decided to investigate. A trail of feathers led past the shelves and behind a curtain.

The trail stopped at a cozy nook by the window, where Twiggy’s favorite perch sat empty. A corner of blue cover peeked out from beneath the cushion there. Maya lifted the cushion and gasped. Her missing book lay underneath, safe but a little chewed at one corner.

A mischievous squawk rang out behind her. Twiggy had slipped his cage and perched on a shelf, clutching a scrap of shiny paper in his beak. He must have dragged the shiny cover to his nook, the way he always dragged shiny things. Mister Chen laughed and caught the bird, while Maya held her book, triumphant. Twiggy squawked once more, as though he still owned the whole mystery.

words237
mean sentence10.8 (9-11)
short present10/22
long tail6
questions6
literal share83%

Shape as written: 22 sentences, shortest 5 words, longest 19, mean 10.8. Score: FK 4.4 raw, 5.5 once genre is accounted for. Questions: 6, 83% literal — analyze, remember, understand. New words: peculiar, hesitated, investigate, nook, mischievous, perched, triumphant, squawked, rattled, latch.

example · core band

Level 7

expository · grade 6 · draft 2 of 3 · cleared 7/7 gates

How Agave Survives the Dry Season

Rain may vanish for months. Yet agave plants survive this harsh pause because their leaves, roots, and rhythms conserve water. An agave is a succulent, a plant that stores water in thick leaves. Its leaves form a rosette around a short stem, and this shape directs rare drops toward the center. The plant looks still. Its structure protects.

Each leaf has an outer skin called the epidermis, which limits damage and slows water escaping into dry air. A pale, waxy coating reflects intense sunlight, while spines discourage animals from piercing the leaves. Inside, water-holding tissue contains mucilage, a slippery gel that keeps moisture from moving too quickly. These layers work together because desert air constantly draws water from living things.

Agave roots spread through shallow soil, so they can absorb brief rain before heat returns. When the soil dries, root growth pauses until moisture becomes available again. The roots wait. Some species grow a taproot, one thick root reaching downward, whereas others make networks of fine, fibrous roots. Either design gives the plant a better chance of finding water after a rare storm.

Agave controls when it exchanges gases with the air. Tiny openings, called stomata, admit carbon dioxide for photosynthesis, but they can also release water vapor. The openings close during hot daylight hours, when evaporation would be fastest. At night, however, cooler temperatures allow the plant to take in carbon dioxide with less water loss overall. This unusual schedule is called CAM photosynthesis, a process that separates gas intake from food-making. Sunlight then supplies the energy for food-making during the day.

After rain, agave refills its leaves and resumes growth, although it does not waste the sudden supply carefully. During drought, it becomes nearly dormant, meaning that growth slows and stored resources are carefully guarded. Its saved water keeps cells firm and allows essential chemical reactions to continue. The plant waits. Eventually, flowers rise above the spiny leaves, and bats or insects carry pollen between blossoms so seeds can develop. This timing matters because reproduction must follow survival, not compete with it during the driest, longest months. Agave survives drought not by finding constant rain, but by collecting water, limiting loss, and slowing its needs. Those adaptations help it endure conditions that would soon damage a broad-leaved tree.

words380
mean sentence13.1 (13-15)
short present13/29
long tail7
questions8
literal share50%
FK adjusted6.8

Shape as written: 29 sentences, shortest 3 words, longest 19, mean 13.1. Score: FK 8.1 raw, 6.8 once genre is accounted for. Questions: 8, 50% literal — analyze, evaluate, remember, understand. New words: epidermis, mucilage, stomata, photosynthesis, CAM photosynthesis, dormant, taproot, fibrous.

example · upper band

Level 11

persuasive · grade 10 · draft 6 of 3 · cleared 7/7 gates

Every Empty Lot Is a Choice

Every empty parking space in a growing city is a decision, because scarce land accommodates vehicles instead of residents. That choice rarely gets debated, yet it determines who can live near a workplace, a school, or a bus stop. A lot can sit half empty for hours while the housing shortage next door never pauses. The imbalance keeps expanding. Cities should therefore replace some parking with housing, especially near transit, because unused asphalt should not outrank the people who need a home.

Scarcity is the argument that matters most. Homes are limited in most cities as population rises, while parking lots remain vacant. When housing demand outruns supply, rents escalate and propel families away from the schools and employment that anchored them there. Constructing on existing parking parcels adds homes without encroaching on farmland at the city's edge, so growth need not signify sprawl. Additional housing will not resolve every problem, but careful construction can alleviate the pressure.

Parking also represents what economists call an opportunity cost, the value lost when land performs one function instead of a superior alternative. A surface lot might generate a modest levy, yet an apartment building there could generate homes, customers, and tax revenue for decades. Housing reinforces nearby businesses because residents purchase groceries, meals, and services all week.

Critics raise a legitimate concern, and it deserves a legitimate answer. Drivers still need somewhere to leave their vehicles, and deliveries, workers, visitors, and people with disabilities depend on parking close to the door. If a city eliminates parking too quickly, congestion spikes and residents pay for a change they never asked for; the problem is the speed of the cuts, not the goal of replacement.

The answer, though, is not to perpetuate every space. Cities can convert parking gradually while studying how much each neighborhood truly requires, because removing more than a block can spare is what triggers backlash. They can prioritize new housing beside transit stops, where fewer residents will own a car, and designate accessible loading zones so deliveries and visitors with disabilities can reach the door. None of that means abandoning drivers; it means admitting that not every stretch of asphalt is essential.

Affordability has to sit at the center of this policy, or the whole project unravels. New apartments built on former lots can still price out the families it claims to assist, and rising rents can trigger displacement. Cities should mandate that some new homes remain below market rate and protect tenants from sudden rent increases. Parking revenue can help underwrite those safeguards, but money alone will not guarantee an equitable outcome.

This transition can also improve residents' health and air quality. When homes sit close to transit, residents can walk, cycle, or ride instead of driving, and shorter trips accumulate across a neighborhood. More housing near employment can shorten commutes if transit remains frequent. A fair plan must support residents whose mobility needs make a vehicle necessary, because accessibility is a public responsibility, not a private convenience.

Opponents are right that parking serves legitimate functions, but protecting all of it forever carries a cost that rarely gets acknowledged. Land devoted only to parked vehicles cannot shelter a family or sustain a struggling business. Cities should prioritize housing where transportation choices are strong, proceed in careful phases, and correct course whenever evidence demands it. Reclaiming some parking is not an attack on drivers; it is a decision to ask scarce urban land to serve more people than idle vehicles ever could.

words582
mean sentence18.2 (17-19)
short present14/32
long tail11
questions8
literal share38%
FK adjusted10.3

Shape as written: 32 sentences, shortest 4 words, longest 32, mean 18.2. Score: FK 11.5 raw, 10.3 once genre is accounted for. Questions: 8, 38% literal — analyze, evaluate, understand. New words: scarcity, accommodates, imbalance, escalate, propel, encroaching, sprawl, signify, alleviate, reinforces.