Coleman-Liau Index & Automated Readability Index: Character-Based Readability Metrics

1. Introduction

Here’s a problem that plagued readability researchers in the 1960s: Syllable counting is hard to automate.

When Rudolf Flesch created his reading ease formula in 1948, computers were room-sized machines that could barely do addition. Counting syllables in thousands of words required manual effort or complex programming.

So readability researchers did what smart researchers do: they worked around the problem.

If syllable counting is hard to automate, what’s easy? Character counting. Computers can count characters instantly — no linguistic knowledge required. Just count letters.

This led to a fascinating question: Can you predict reading difficulty using only character counts instead of syllables?

The answer, it turns out, is yes — and it’s surprisingly effective.

In this article, we’ll explore two character-based readability formulas:

Coleman-Liau Index (1975) — Uses characters per word Automated Readability Index (1967) — Uses characters per word and characters per sentence

These formulas were designed when computer processing was limited, but they’re still relevant today because:

• They’re computationally simple (no linguistic parsing needed)
• They work in any language that uses the Latin alphabet
• They’re surprisingly accurate for English
• They’re built into some automated systems and code libraries

Whether you’re a developer building readability into a tool, a researcher comparing readability metrics, or simply curious about how character counts predict reading difficulty, this guide will explain these underrated formulas.

2. Define the Core Concepts: Coleman-Liau & ARI

Coleman-Liau Index

Coleman-Liau Index is a readability metric that expresses text difficulty as a U.S. grade level (1–18+), using only:

• Average word length (characters per word)
• No syllable counting required

Formula:

Grade = 0.0588 × characters − 0.296 × sentences − 15.8
(divided by total words, adjusted for sentence count)

Output: Grade level (1–18+), like Flesch-Kincaid and Gunning Fog

Key advantage: No syllable counting. Just count letters and sentences.

Automated Readability Index (ARI)

Automated Readability Index is a readability metric that also expresses text difficulty as a grade level, using:

• Average word length (characters per word)
• Average sentence length (words per sentence)

Formula:

Grade = 4.71 × (characters ÷ words) + 0.5 × (words ÷ sentences) − 21.43

Output: Grade level (1–18+)

Key advantage: Similar to Coleman-Liau but weights sentence length more heavily.

The Core Insight: Characters vs. Syllables

The traditional approach (Flesch-Kincaid, Gunning Fog, SMOG):

• Count syllables per word to measure word complexity
• Assumption: More syllables = more difficult

The character-based approach (Coleman-Liau, ARI):

• Count characters per word to measure word complexity
• Assumption: Longer words (more characters) = more difficult

Why does character count work as a proxy for syllable count?

In English, there’s a strong correlation between word length (in characters) and word length (in syllables):

• Short words (1–4 characters) typically have 1–2 syllables: “cat,” “dog,” “play,” “jump”
• Medium words (5–8 characters) typically have 2–3 syllables: “computer,” “different,” “important”
• Long words (9+ characters) typically have 3+ syllables: “technology,” “responsib​ility,” “communication”

This correlation isn’t perfect, but it’s strong enough that character count predicts syllable count reasonably well.

Result: Character counts can approximate syllable-based metrics without the linguistic complexity.

3. The History: Why These Formulas Were Created

The Syllable-Counting Problem (1960s)

In the 1960s, readability research was expanding, and computers were just becoming available. But early computers had severe limitations:

• Limited memory (kilobytes, not megabytes)
• No built-in language processing
• Syllable-counting algorithms were complex to implement
• Manual syllable counting was labor-intensive

Researchers faced a dilemma: The best readability metrics (Flesch, Gunning Fog) required syllable counting, which was expensive and difficult to automate.

The Automated Readability Index (1967)

John Kincaid, Robert Fishburne, Richard Rogers, and Brad Chissom — the same team that adapted Flesch’s formula to grade levels — created the Automated Readability Index (ARI) in 1967.

Their goal: A readability metric that could be calculated by computer without linguistic preprocessing.

ARI uses only character and word counts, making it simple to implement in code. No syllable parsing needed.

Historical note: This was for U.S. military training materials, similar to their later Flesch-Kincaid work. The military needed readability metrics for technical manuals, and they needed to calculate them quickly and automatically.

The Coleman-Liau Index (1975)

Meri Coleman and T.L. Liau created the Coleman-Liau Index in 1975, also designed for automation but with a focus on using character counts more efficiently.

Coleman-Liau refined the character-based approach, creating a formula that was:

• Even simpler than ARI (fewer constants, cleaner math)
• More accurate for typical English texts
• Easy to implement in any programming language

Why These Formulas Became Less Popular

By the 1980s and 1990s, computer power increased exponentially. Syllable counting, once difficult, became trivial. Modern software could easily:

• Parse text linguistically
• Count syllables accurately
• Run complex algorithms in milliseconds

As a result:

• Syllable-based formulas (Flesch-Kincaid, Gunning Fog) remained the standard
• Character-based formulas (Coleman-Liau, ARI) became “legacy” metrics
• Academic research moved to more sophisticated language models

But character-based metrics never disappeared entirely. They persisted in:

• Simple readability checking tools (low computational overhead)
• Code libraries and software where simplicity is valued
• Multilingual contexts (some languages don’t have easy syllable parsing)
• Historical research (comparing old studies to new ones)

4. How the Formulas Work: The Technical Breakdown

Automated Readability Index (ARI)

ARI = 4.71 × (characters ÷ words) + 0.5 × (words ÷ sentences) − 21.43

Worked example:

Sample text: “The technology sector is evolving rapidly. Companies must adapt to survive. Innovation drives growth and profitability.”

Metrics:

• Characters: 156 (including spaces and punctuation, or just letters?)
  • Let’s count letters only: The(3) technology(10) sector(6) is(2) evolving(8) rapidly(7) Companies(9) must(4) adapt(5) to(2) survive(6) Innovation(10) drives(6) growth(6) profitability(12) = 96 letters
  • With spaces: ~110 characters
  • Let’s use letters only for this example: 96 letters
• Words: 15
• Sentences: 3

Calculate:

• Characters per word: 96 ÷ 15 = 6.4
• Words per sentence: 15 ÷ 3 = 5

ARI:

• ARI = 4.71 × 6.4 + 0.5 × 5 − 21.43
• ARI = 30.14 + 2.5 − 21.43
• ARI = 11.21 (11th-grade level)

Coleman-Liau Index

Coleman-Liau = 0.0588 × characters − 0.296 × sentences − 15.8
(computed per 100 words)

The formula is easier to express as:

Grade = (0.0588 × characters − 0.296 × sentences − 15.8) × (words ÷ 100)

Using the same sample text:

Metrics:

• Characters: 96 letters
• Sentences: 3
• Words: 15

Calculate (per 100 words):

• Scaled characters: 96 × (100 ÷ 15) = 96 × 6.67 = 640
• Scaled sentences: 3 × (100 ÷ 15) = 3 × 6.67 = 20

Coleman-Liau:

• Grade = 0.0588 × 640 − 0.296 × 20 − 15.8
• Grade = 37.63 − 5.92 − 15.8
• Grade = 15.91 (16th-grade level)

Note: Coleman-Liau scores much higher than ARI for this text. Why? Coleman-Liau weights character count more heavily.

5. Interpreting Scores: What They Mean

Both ARI and Coleman-Liau output grade levels (1–18+), same as Flesch-Kincaid and Gunning Fog:

Grade 1–3: Elementary school (very easy) Grade 4–6: Upper elementary (easy) Grade 7–9: Middle school (fairly easy) Grade 10–12: High school (standard to fairly difficult) Grade 13–15: College (difficult) Grade 16+: Graduate level (very difficult)

The main difference from other formulas is accuracy and interpretation, which we’ll cover in the limitations section.

6. Comparing Character-Based to Syllable-Based Formulas

How Character-Based Metrics Compare to Syllable-Based

Sample text: “Pneumonia treatment requires antibiotics. Symptoms include cough and fever. Consult a physician immediately.”

Character & syllable counts:

• Characters: ~80 letters
• Words: 15
• Sentences: 3
• Syllables: “Pneumonia” (3), “treatment” (2), “requires” (2), “antibiotics” (4), “Symptoms” (2), “include” (2), “cough” (1), “and” (1), “fever” (2), “Consult” (2), “physician” (3), “immediately” (4) = ~28 syllables

Metric comparison:

Pattern: Character-based metrics (ARI, Coleman-Liau) score this medical text higher than syllable-based metrics. Why?

• Words like “pneumonia,” “antibiotics,” “physician” have many characters but fewer syllables than you’d expect
• Character count penalizes these words more than syllable count
• Result: Higher (more difficult) grade level

This is a key difference: Character-based formulas sometimes overestimate difficulty for medical/technical writing with long Latinate words.

Side-by-Side Comparison: All 5 Major Formulas

For the same medical text above:

Takeaway: For this text, syllable-based formulas cluster around 8–9, while character-based formulas cluster around 10–14. Character-based metrics are more conservative (higher scores) for texts with complex Latinate vocabulary.

7. Strengths & Limitations of Character-Based Metrics

Strengths of ARI & Coleman-Liau

✅ No linguistic processing needed

• Just count characters and words
• Works in any language using Latin alphabet
• No syllable-counting libraries required

✅ Computationally simple

• Fast calculation, minimal memory
• Can be implemented in 5 lines of code
• Ideal for embedded systems or low-power devices

✅ No variability from syllable-counting disagreements

• “Poem” is 1 or 2 syllables depending on pronunciation
• But “poem” is always 4 characters (p-o-e-m)
• Character counting is unambiguous

✅ Works for non-English Latin-alphabet languages

• French, Spanish, German, Italian, etc.
• Character length correlates with word complexity across Romance languages
• No language-specific syllable rules needed

✅ Multilingual capability

• Same formula works across different languages
• Useful for organizations writing in multiple languages

Limitations of ARI & Coleman-Liau

❌ Less accurate than syllable-based formulas for English

• Character count is a proxy for syllable count, not a perfect measure
• Correlation breaks down for certain word types (monosyllabic long words, polysyllabic short words)

❌ Overestimates difficulty for medical/technical writing

• Medical terms often have many characters but manageable syllables: “thoracic” (8 chars, 3 syllables)
• Character count sees “thoracic” as very complex; syllable count is moderate
• Results in inflated grade levels

❌ Doesn’t account for word frequency or familiarity

• Character count treats “the” (3 chars) and “xerophyte” (9 chars) purely by length
• Doesn’t consider that “xerophyte” is an uncommon word that would be very difficult regardless of length
• A syllable-based formula would also have this limitation, but other formulas like Lexile account for word frequency

❌ Coleman-Liau inconsistency across sources

• Different sources implement Coleman-Liau slightly differently (some count spaces, some don’t; some use letters only, some use all characters)
• Can yield different scores depending on the tool

❌ ARI can produce unrealistic scores

• For very simple text: can score negative grade levels
• For very complex text: can exceed 18+ grade level dramatically
• Flesch-Kincaid bounds output between 1–18 more predictably

❌ Doesn’t measure sentence complexity

• Two sentences with identical words but different structures score identically
• “If you have liver disease, talk to your doctor before taking this medication” vs. “In the event that hepatic compromise exists, consultation with healthcare provider prior to pharmaceutical administration is advised” would score similarly (or close) if word/character counts are similar

8. When to Use Character-Based Formulas

Use ARI or Coleman-Liau If:

✅ You’re building a simple, lightweight readability tool

• No complex dependencies or NLP libraries
• Minimal computational overhead
• Can run on low-power devices or embedded systems

✅ You need language-agnostic readability metrics

• Writing in multiple languages
• Need same formula to work across Romance languages
• Don’t want language-specific syllable-counting rules

✅ You’re in a multilingual environment

• International organizations
• Websites serving multiple language communities
• Need consistency across languages

✅ You want to avoid syllable-counting variability

• Different tools count syllables differently
• Character counting is deterministic (same result every time)
• Want reproducibility across systems

✅ You’re comparing historical readability research

• Some older studies used ARI
• Need to match methodology for comparison

Avoid ARI & Coleman-Liau If:

❌ You need accurate readability for English text

• Use Flesch-Kincaid or Gunning Fog instead
• Character-based formulas are approximations

❌ You’re working with medical or technical writing

• Character-based metrics overestimate difficulty
• Use SMOG (medical) or Gunning Fog (technical)

❌ You want to match standard industry practice

• Flesch-Kincaid is default in Microsoft Word, Google Docs
• SMOG is standard in healthcare
• Character-based metrics are legacy

❌ You need nuanced vocabulary complexity assessment

• Character count doesn’t capture word frequency or domain-specificity
• Lexile or other semantic formulas are better

9. Practical Applications & Use Cases

For Software Developers

Scenario: Building a writing assistant or autocorrect tool

Character-based metrics are ideal because:

• No NLP libraries needed (lighter dependencies)
• Minimal computational overhead
• Can calculate readability in real-time as user types
• Works across languages from same code

Implementation: Most modern text editors (Grammarly, Hemingway Editor) use hybrid approaches: simple character-based checks for speed, plus more sophisticated analysis for accuracy.

For Multilingual Platforms

Scenario: Website serving English, French, Spanish, German users. Need readability scores for all.

Character-based metrics shine because:

• One formula works across all languages
• No need for language-specific syllable parsers
• Consistent methodology across regions

Trade-off: Slightly less accuracy than language-specific formulas, but consistency is more important.

For Historical Research

Scenario: Comparing readability of texts across decades or analyzing old research papers.

Character-based metrics useful because:

• Some historical studies used ARI
• Need to match methodology for valid comparison
• Legacy tool integration

For Embedded/Low-Power Systems

Scenario: Building readability analysis into e-ink devices, mobile apps with strict performance constraints, or IoT devices.

Character-based metrics ideal because:

• Minimal computational overhead
• Tiny code footprint
• Instant calculation

10. Common Questions (FAQ)

Q: Why would anyone use Coleman-Liau when Flesch-Kincaid exists?

A: Good question. In 2024, for English-language content, Flesch-Kincaid is generally better. Coleman-Liau was useful in 1975 when syllable counting was expensive. Today, it’s a legacy formula. Use it if you have a specific reason (multilingual environment, embedded system, historical research), otherwise use Flesch-Kincaid.

Q: Can I use ARI/Coleman-Liau for non-English languages?

A: Partially. Character-based metrics work better for Romance languages (French, Spanish, Italian) where word-length patterns are similar to English. They’re less reliable for languages with different character-length-to-complexity relationships, like German (many long compound words) or English (many short common words). Test for your language.

Q: Why do ARI and Coleman-Liau score so differently for the same text?

A: Different weighting. ARI emphasizes sentence length more; Coleman-Liau emphasizes character count more. Coleman-Liau typically scores higher (more difficult). They’re related formulas but with different calibrations.

Q: Should I use character-based metrics for medical writing?

A: No. SMOG is specifically designed for medical writing and is more accurate. Character-based metrics often overestimate difficulty for medical texts because medical words have many characters but manageable syllable counts.

Q: Is there a “best” readability formula overall?

A: No. Different formulas are better for different contexts:

• General English writing: Flesch-Kincaid
• Business/marketing: Gunning Fog
• Medical writing: SMOG
• Multilingual: Character-based (ARI/Coleman-Liau)
• Academic/sophisticated: Lexile (accounts for word frequency)
• Comparison: Use multiple formulas and look at the gap between them

Q: Why aren’t character-based metrics used more widely if they’re simpler?

A: Because they’re less accurate. Simplicity isn’t always worth a trade-off in accuracy. Modern computing makes syllable counting trivial, so there’s no reason to use approximations when exact measurements are available.

11. Further Resources & Tools

Related Articles on This Site

• What Is Readability & How to Measure It — foundational overview of all readability formulas
• Flesch Reading Ease Explained — deep dive into the most popular formula
• Flesch-Kincaid Grade Level: What It Means — grade-level version of Flesch
• Gunning Fog Index vs Flesch-Kincaid — comparison of two major syllable-based formulas
• SMOG Index: Measuring Text Complexity — healthcare-specific readability metric

External Resources

• Coleman, M., & Liau, T.L. (1975): “A Computerized Readability Formula Designed for Machine Scoring” — Original Coleman-Liau research
• Senter, R.J., & Smith, E.A. (1967): “Automated Readability Index” — Original ARI research, U.S. Air Force Technical Report
• Kincaid, J.P., et al. (1975): “Derivation of New Readability Formulas (ARI, SMOG, Flesch-Kincaid)” — Comprehensive military readability research
• Flesch, R. (1974): “The Art of Readable Writing” — Classic reference on readability (mentions character-based alternatives)

Try the Tool

Want to see how character-based metrics score differently than syllable-based ones? Use our interactive readability checker to:

• Paste any text
• See Flesch-Kincaid, Gunning Fog, SMOG (syllable-based)
• Compare against theoretical ARI/Coleman-Liau scores
• Understand where character-based and syllable-based metrics diverge
• See which formula best matches your audience

11. Conclusion: Character-Based Readability Metrics in Context

Automated Readability Index (ARI) and Coleman-Liau Index are character-based readability formulas designed when syllable counting was computationally expensive. They use character counts as a proxy for syllable counts.

The core insight: Word length (in characters) correlates with word complexity, allowing character-based formulas to approximate syllable-based formulas without linguistic preprocessing.

Key takeaways:

1. Character-based formulas are simpler but less accurate. They work reasonably well for English but aren’t the standard for a reason.
2. ARI and Coleman-Liau are useful in specific contexts: multilingual environments, embedded systems, historical research, lightweight tools.
3. For English-language content, syllable-based formulas are better: Flesch-Kincaid for general use, Gunning Fog for business, SMOG for medical.
4. Character-based metrics often overestimate difficulty for medical/technical writing where long Latinate words have fewer syllables than their character count suggests.
5. When comparing character-based to syllable-based: The gap tells you something about the text’s vocabulary (lots of long Latinate words vs. short common words).

Character-based readability formulas are historical artifacts that served an important purpose in the 1960s–1980s. Today, they’re most useful for multilingual contexts or as educational examples of how readability science evolved.

For most modern applications, use syllable-based formulas. But understanding character-based approaches gives you insight into the diversity of readability measurement and the clever engineering that made automatic readability checking possible before modern NLP.

Next Steps

Developers: If you’re building readability into a tool and need multilingual support, consider character-based metrics. But test their accuracy for your specific use case against human judgment.

Researchers: If comparing to historical studies using ARI, understand that it uses character counts, not syllables. Account for this systematic difference when comparing results.

Educators: Use character-based formulas as teaching examples of how linguistic science adapts to computational constraints. Show students how approximations work and why precision matters.

Content creators: For English-language content, ignore character-based metrics. Use Flesch-Kincaid, Gunning Fog, or SMOG depending on context.

Try our tool to check how different readability formulas compare on your text. The comparison between formulas often reveals more than any single score.