Chinese Oral Reading Fluency Assessment: A Screening Guide for Immersion Programs

In many immersion programs, students do most of their early reading in Chinese. English reading has well-known screening tools and national norms. Chinese reading often gets a leveled-book check or nothing at all, so a child's growth in the language they read most is the growth nobody can see.

A Chinese oral reading fluency (ORF) screen makes that growth visible. The student reads an unfamiliar Chinese passage aloud for one minute, and you count characters correct per minute and accuracy. It is the same idea as an English ORF screen, adjusted for how Chinese is written.

This guide is for literacy leads, principals, and immersion teachers setting that screen up. It covers why Chinese is counted in characters, how to mark a miscue on a character, which benchmarks exist and which don't, and how partner-language screening fits the state reading laws districts now work under, with Minnesota's READ Act as a worked example. Minnetonka Public Schools uses ReadingFluency.app with its Chinese immersion teachers, and a Mandarin version of this guide is written for teachers to read and pass along.

Why Chinese ORF counts characters, not words

Written Chinese puts no spaces between words, and readers do not always agree where one word ends. Is 公园里 one word or two? Is 看见 one word or two? A words-per-minute count would depend on whose segmentation you used, and two teachers could score the same read differently.

Characters don't have that problem. Each character is one syllable and one unit on the page, so two teachers counting the same recording get the same number. A Chinese ORF screen therefore counts:

• each character the student reads correctly
• per minute of reading time
• and no punctuation: ，。！？ and quotation marks are not read aloud and are not counted

One consequence matters for anyone reading the data: a Chinese rate and an English WCPM are different measures on different scales. Compare Chinese to Chinese, whether that is the same student across the year or students in your program reading the same passages.

In ReadingFluency.app, a passage whose language is set to Chinese is split one character at a time everywhere: on the reading screen, in marking, and in scoring. Punctuation is never scored.

A sample passage with pinyin

Here is a short passage marked the way a teacher would mark a one-minute read.

Sample passage, marked

An original early-elementary passage written for this guide. Pinyin is here for you; the student reads characters only.

星xīng期qī天tiān，天tiān气qì很hěn好hǎo。小xiǎo明míng和hé妈mā妈ma去qù公gōng园yuán。公gōng园yuán里lǐ有yǒu很hěn多duō花huā，红hóng的de、黄huáng的de，真zhēn好hǎo看kàn。小xiǎo明míng看kàn见jiàn一yì只zhī小xiǎo鸟niǎo→ 岛在zài树shù上shàng唱chàng歌gē。他tā说shuō：“小xiǎo鸟niǎo，你nǐ好hǎo！”

SkippedMispronounced→ what the student said

Both miscues are typical of early readers. The student dropped 里 from 公园里 because the sentence still makes sense without it. They read 鸟 as 岛 because the two characters share most of their strokes; look-alike characters (形近字) are the most common misreads you will mark.

The pinyin is for you, not the student. In a screen, students read characters only, and the ReadingFluency.app reading screen shows characters without pinyin. If your program assesses pinyin, keep it as its own measure rather than printing it on the ORF passage.

How to mark a miscue on a character

Score each character as right or wrong. The rules follow the usual English ORF conventions, applied one character at a time:

Two calls are specific to Chinese. Make them as a team before the first window, write them down, and apply them the same way in fall, winter, and spring:

1. Tones. There is no single convention for whether a tone error on the right character counts as a miscue. Some teams count it once students have been taught the character; others are more lenient in the early grades. Either works if everyone scores the same way.
2. Accent and home varieties. A student who speaks Cantonese or another variety at home, or who blends sounds such as n and l or zh and z, may pronounce a character differently from standard Mandarin without misreading it. The English rule carries over: pronunciation differences that come from dialect rather than decoding are not miscues.

In ReadingFluency.app you mark by tapping a character. One tap marks it Mispronounced (yellow), a second marks it Skipped (struck through), and a third sets it back to correct. There is no separate self-correction mark, because a self-correction is scored as correct. When AI scoring hears an added character, it shows it as a small + chip and does not count it.

Characters correct per minute

The two numbers a Chinese ORF screen produces:

• Characters correct per minute = (characters attempted − miscues) ÷ minutes of reading
• Accuracy = characters correct ÷ characters attempted

Using the sample passage: say the student read all 48 characters in 40 seconds with the 2 miscues shown.

• Characters correct: 48 − 2 = 46
• Rate: 46 ÷ (40 ÷ 60) = 69 characters correct per minute
• Accuracy: 46 ÷ 48 = 95.8%

On a real one-minute screen, the passage is longer than anyone in the grade can finish. You stop at 60 seconds, "characters attempted" runs to the last character the student reached, and the rate is simply the number of characters correct.

Those numbers illustrate the arithmetic. They are not a benchmark for any grade.

In ReadingFluency.app, the assessment report shows the rate on the Reading Rate card, labeled CPM, which counts correct characters only. Accuracy is on the Fluency Score card, and each card's How it's calculated popover shows the counts behind the number. Progress charts and roster-group goals share one rate field across languages, labeled WCPM or WPM; for Chinese passages, read the number in it as characters correct per minute.

Benchmarks and norms: what exists and the local path

English ORF has widely used national reference points, such as the Hasbrouck–Tindal norms in the benchmark table in our ORF guide. Don't carry those numbers over to Chinese. The unit is different, the script is different, and an immersion student's exposure to Chinese is nothing like a monolingual English reader's exposure to English.

Published Chinese ORF norms for US immersion students are thin, and a norm only means something for the passages and students it was built on. So the practical path is local norms, built from your own program:

1. Fix a passage set for each grade and each window (fall, winter, spring), and keep it the same from year to year.
2. Screen every immersion student in the grade with the same directions, timing, and scoring rules.
3. Look at the spread within each grade and window: medians and percentiles, not a single cut score, in the first year.
4. Set provisional decision rules, such as a percentile below which a student gets a closer look, and check them against teacher knowledge and your other measures (high-frequency characters, comprehension).
5. Revisit the rules once a second year of data is in.

ReadingFluency.app has no national Chinese norm built in. The percentile bands on the Reports page are calculated from your own assessments, filtered however you choose, which is exactly the local comparison: this student against your program, on your passages.

Choosing passages for a Chinese screen

A screening passage should be:

• Unfamiliar. Not the lesson text the class practiced this week.
• Matched to the grade's characters. Mostly characters the students have been taught, with few new characters (生字).
• Long enough. Longer than the fastest reader in the grade can finish in a minute.
• Stable. The same passages every window, or forms your team has checked are equivalent.
• In the script you teach. ReadingFluency.app offers one language option, Chinese, with no separate Traditional option.

There are three ways to get passages into ReadingFluency.app:

• Import a curriculum PDF, a photo of a page, or pasted text. Set the language to Chinese first; the AI extracts each passage verbatim and you review every one before it lands in your library. The PDF import tutorial walks through it.
• AI Generate. Choose Chinese, a grade, a length from 30 seconds to 3 minutes, an optional topic, and a level: YCT 1–4, HSK 1–6, or a custom label. The level only guides generation. Read every generated passage before you screen with it, checking for characters your students haven't learned and for traditional forms if you teach simplified. See the passage generator.
• Manual entry for a benchmark passage you already have as text.

Teachers who mark on paper can print a Chinese passage with 28 characters per line and a running character count on each line.

Running the screen: one-to-one or as a class

Keep the administration identical for every student: the same directions, the same time limit, a quiet spot, and the device microphone close to the reader. A short script in Chinese helps every adult say the same thing, for example:

请你把这篇短文大声读出来，尽量读好。遇到不认识的字，先跳过，接着往下读。

If you screen one-to-one, you can instead supply a character after about three seconds and count it as a miscue. Pick one approach and use it every time.

One-to-one. Open the passage, start a time-limited read (one minute is the usual screen), and either tap characters as the student reads or record and mark afterward.

As a class. Host a live group reading session: pick the students, set the reading mode to Time-limited (30, 60, or 90 seconds, or a custom length), and share the join code. Students read aloud on their own devices and submit their recordings, with no student accounts. With AI Auto-Analysis on, each recording is analyzed as it arrives, and you review them from one queue.

Reviewing AI scoring in Chinese

When a recording is analyzed, speech recognition transcribes it in Chinese and the scorer lines the transcript up with the passage one character at a time. If the recognizer writes a different character with the same sound (same pinyin, same tone), the character counts as correct, since a homophone in the transcript is not a reading error. The result is a draft: nothing is final until you review it.

What to check before you trust the score:

• Tones. Tones are not judged directly from the audio. A tone slip is flagged only when the recognizer hears a different character. If your team counts tone errors, listen for them yourself.
• Accent and home varieties. Regional or heritage pronunciation can produce false flags. Replay before you accept one.
• The end of the read. Confirm the last character the student reached, since it sets "characters attempted".
• Anything that changes a decision. Each sentence has a play button, and a character the AI marked as mispronounced has its own short clip. Tap to override a mark, and the rate and accuracy update.

For a longer checklist, see what teachers should verify in AI oral reading fluency assessment, or the AI reading fluency tracker for how the review screen works.

State reading laws and partner-language screening

Most states now have a reading law. At least 40 have passed flagship literacy laws, according to the Albert Shanker Institute, and most of them aim their requirements at kindergarten through grade 3 (Education Week, November 2025). Many include screening requirements for young readers, often several times a year, and some, Minnesota among them, require a state-approved screener.

What those laws say about students who learn to read in another language varies from state to state, and many say little. Two questions are worth answering for your program before the next window:

• Does your state accept screening in the partner language, and who approves the tool: the state, or the district?
• Do immersion students also need an English screen, and from which grade?

Whatever the answers, a Chinese ORF screen gives the program a named partner-language measure, fall, winter, and spring data on the same schedule as English, a written decision rule, and a way to progress-monitor students who receive partner-language intervention. It also gives teachers something concrete to share with families about a child's reading in Chinese.

Example: Minnesota's READ Act

Minnesota is one state whose law is explicit about immersion programs. Minn. Stat. § 120B.12, the READ Act, says:

• Every student in kindergarten through grade 3, including students in dual language immersion programs, is screened three times a year with an MDE-approved screener: within the first six weeks of school, by February 15, and within the last six weeks.
• If the program teaches foundational reading in the partner language, those students must be screened in the partner language, at the same interval as the English screens. If it teaches foundational reading in English, they are screened in English.
• The partner-language screening tool is approved by the district for kindergarten through grade 3.
• If no screener is available in the partner language, the district must identify how it assesses reading proficiency and how it provides targeted instruction in that language.
• Local Literacy Plans are updated and submitted to MDE by June 15 each year and posted on the district website. Since December 31, 2025, a district with an immersion program reports the partner language, the grades in the program, and the language it screens foundational reading in.
• The statute also created a dual language immersion advisory committee (by January 1, 2026), whose charge includes guidance to immersion programs on screeners.

Example: a district's literacy plan

For a published example of how one district documents this, Minnetonka Public Schools' Local Literacy Plan (2025–26 data, submitted June 2026) lists Chinese Simplified Oral Reading Fluency among its partner-language screeners: in winter and spring of grade 1 and in fall, winter, and spring in grades 2 through 5, alongside pinyin and high-frequency-character measures in the early grades. That schedule and battery are Minnetonka's own decisions. This guide describes a general approach, not how Minnetonka runs its screening.

Glossary: Chinese terms teachers use

A checklist for program leads

Before the window:

• Choose a passage set for each grade and window, and have teachers read it for unfamiliar characters.
• Write down the scoring rules: tones, home varieties, supplied characters, and the time limit.
• Agree on one set of directions in Chinese.
• Decide who screens whom, and whether each grade reads one-to-one or in a group session.

After the window:

• Review AI marks before anyone uses the numbers.
• Look at each grade's spread, flag students for a closer look, and check those flags against high-frequency characters, comprehension, and what teachers know.
• If your state asks for a literacy plan (Minnesota's Local Literacy Plan, for example), record the measure, schedule, and decision rule in it.
• Share the Mandarin version of this guide with your teachers.

Passages, rosters, manual marking, and progress tracking in ReadingFluency.app work with Chinese passages the same way they do with English ones. For individual, school, and district plans, see pricing.