Type-Token Ratio
Also known as: TTR, Type/Token Ratio, Lexical Variation Ratio
The type-token ratio (TTR) is the oldest and most widely used measure of lexical diversity: the number of distinct word types in a text divided by the total number of word tokens. A text in which few words repeat yields a TTR near 1, while a text that recycles a small vocabulary yields a TTR near 0. Despite its intuitive appeal and trivial computation, the raw ratio is severely confounded by text length, which has motivated a long line of length-correcting transformations and, ultimately, the more robust indices that have largely superseded it for serious comparison.
Key highlights
- Trivially simple to compute and to explain, requiring only a count of types and tokens with no parameters or sampling.
- Fully transparent and reproducible: anyone with the same tokenization will obtain exactly the same number.
- Provides a valid, sensitive comparison when texts are constrained to equal length, as in many fixed-length elicitation designs.
- Serves as the conceptual foundation and arithmetic building block for nearly every more advanced lexical diversity measure.
Intuition
This section is available to Pro members. Upgrade to Pro
How it works
This section is available to Pro members. Upgrade to Pro
When to use it
Use raw TTR only when all texts being compared are of identical or near-identical length, or when you need a quick, transparent first look at vocabulary variation in a single text. It is appropriate for fixed-length elicitation tasks, for teaching the concept of lexical diversity, and as a baseline against which more sophisticated indices are reported. Avoid raw TTR for comparing texts of unequal length — use MSTTR, the Maas index, or preferably MTLD or vocd-D instead — and avoid it for very short samples, where a single repeated word swings the ratio dramatically.
Strengths & limitations
- Trivially simple to compute and to explain, requiring only a count of types and tokens with no parameters or sampling.
- Fully transparent and reproducible: anyone with the same tokenization will obtain exactly the same number.
- Provides a valid, sensitive comparison when texts are constrained to equal length, as in many fixed-length elicitation designs.
- Serves as the conceptual foundation and arithmetic building block for nearly every more advanced lexical diversity measure.
- Raw TTR falls systematically with text length, so it confounds vocabulary richness with sample size and cannot fairly compare unequal texts.
- It is highly unstable for short samples, where adding or repeating a single word can shift the ratio substantially.
- It treats all repeated words alike and ignores the distribution of frequencies, so two texts with very different word-frequency profiles can share a TTR.
- The length-corrected variants (RTTR, CTTR, Maas) reduce but do not remove the length dependence, because they lack an underlying model of vocabulary growth.
Common pitfalls
This section is available to Pro members. Upgrade to Pro
Applications
This section is available to Pro members. Upgrade to Pro
Frequently asked
Why does TTR almost always decrease as a text gets longer?
Because a language has a small, fixed set of very high-frequency function words — the, of, and, a, to — that must recur in almost every sentence. As a text grows, these words are re-used rather than adding new types, so the numerator V grows far more slowly than the denominator N. The ratio therefore declines monotonically and tends toward zero for long texts, regardless of how rich the writer's vocabulary actually is. This is why raw TTR cannot be compared across texts of different lengths.
What is the difference between a 'type' and a 'token'?
A token is each running word as it appears, so the sentence 'the cat sat on the mat' contains six tokens. A type is each distinct word form, so the same sentence contains five types because 'the' occurs twice. V counts types and N counts tokens, and TTR is their ratio. Whether 'cat' and 'cats', or 'run' and 'ran', count as the same type depends on whether you lemmatize, which is a decision you must fix before comparing texts.
Should I use raw TTR or a length-corrected version?
Use raw TTR only when your texts are the same length. Otherwise prefer a length-robust measure: MSTTR or the Maas index are simple improvements, but the curve-based vocd-D and the sequential MTLD are more robust still and are the current standards for comparing texts of unequal length. The transformations RTTR and CTTR reduce the length effect but do not remove it, so they should be reported with caution and never treated as fully length-independent.
Sources
- 1.Johnson, W. (1944). Studies in language behavior: A program of research. Psychological Monographs, 56(2), 1–15.
- 2.Malvern, D., Richards, B., Chipere, N., & Durán, P. (2004). Lexical Diversity and Language Development: Quantification and Assessment. Palgrave Macmillan.ISBN 9781403902313
- 3.McCarthy, P. M., & Jarvis, S. (2010). MTLD, vocd-D, and HD-D: A validation study of sophisticated approaches to lexical diversity assessment. Behavior Research Methods, 42(2), 381–392.
You have read it. What now?
Cite this page
ScholarGate. (2026, June 22). Type-Token Ratio. ScholarGate. https://scholargate.app/linguistics/type-token-ratio