Lexical Diversity in English Vocabulary: Why Better Writing Is Not Just “Using More Different Words”

A student writes: > The problem is serious because the problem affects many people. The problem also creates other problems. A teacher may say: > “Use more varied vocabulary.” That advice is understandable. But another student reacts by replacing every repeated word: > The problem is serious because the issue affects many people. The difficulty also creates other complications. Now the paragraph sounds less repetitive. Has the thinking improved? Not necessarily. The writer may have gained surface variety while weakening reference. Which “problem” is: > issue? Which is: > difficulty? Which new “complications” are still the original problem, and which are consequences? This is why **lexical diversity** needs careful teaching. At its simplest, lexical diversity concerns how varied the vocabulary of a text is. But good writing is not a competition to avoid repeating any word twice. A coherent text must repeat some important words. A precise technical text may need the same noun again and again. A narrative may deliberately return to one concrete object. The real question is: > Does the writer have enough lexical choice to express distinctions when distinctions are needed? That is a richer job than: > How many different words did you use? ## Quick answer: what is lexical diversity? Lexical diversity refers to the range or variety of words used in a language sample. A simple measure is the **type-token ratio**, often shortened to **TTR**. Suppose a text contains 100 word tokens. If 60 different word forms occur, the TTR is: > 60 / 100 = 0.60. Here: – **tokens** = every word occurrence; – **types** = different word forms. A text with more repeated words generally has a lower TTR. A text with more unique forms generally has a higher TTR. That sounds straightforward. It is also easy to misuse. ## Why text length matters Imagine two texts. Text A: > 50 words. Text B: > 1,000 words. Even a skilled writer must repeat basic words such as: > the > and > is > of > to many times in the longer text. As a text grows, repeated vocabulary becomes increasingly inevitable. So raw TTR tends to decline with text length. That means: > a 100-word composition and a 1,000-word article should not be compared casually using raw TTR. This is one of the oldest practical problems in lexical-diversity measurement. ## A simple example of the length problem Short text: > The dog chased the ball. A child laughed. Suppose almost every content word is unique. Now extend the story: > The dog chased the ball. The child threw the ball again. The dog chased the ball across the field. The child laughed as the dog returned the ball. The longer text is richer as a scene. Yet repeated words such as: > dog > child > ball reduce the type-token ratio. The lower TTR does not prove that the longer passage is worse. It proves that diversity statistics depend on sample design. ## This is why researchers use alternatives to raw TTR Researchers have developed measures intended to reduce sensitivity to text length. Examples include: – **MATTR** — Moving-Average Type-Token Ratio; – **MTLD** — Measure of Textual Lexical Diversity; – **vocd-D**; – **HD-D** and related measures. The mathematical details differ. The educational principle is more important: > One number cannot be interpreted without knowing how it was calculated, how long the text was, and what kind of text it was. A vocabulary score is not self-explanatory. ## MATTR: compare fixed windows Moving-Average Type-Token Ratio takes small windows of equal length across a text and calculates diversity repeatedly. Instead of comparing one short text against one long text, it asks, roughly: > How varied is the vocabulary inside comparable local windows? This reduces one major problem of raw TTR: > overall text length. It does not solve every problem. But it illustrates good measurement thinking: > control the comparison before interpreting the result. ## MTLD: how long before repetition lowers diversity? MTLD takes a different approach. It analyses how long a sequence can continue before the type-token ratio falls below a threshold, then combines those segments into a measure. Students do not need to calculate MTLD by hand. What matters is the idea: > lexical diversity can be measured in ways that are more robust than simply counting unique words once. This is useful when AI tools, writing platforms or research papers display vocabulary statistics. Ask: > Which metric? not merely: > What score? ## Lexical diversity is not lexical sophistication A student can write many different common words. That may increase diversity. But it does not automatically mean the vocabulary is sophisticated. Compare: > The boy walked to the shop and saw a dog beside a car near a tree. Many distinct words. Now: > The committee postponed the vote because the evidence remained inconclusive. Fewer lexical items perhaps, but more specialised academic vocabulary. These are different dimensions. **Lexical diversity** asks: > How varied is the vocabulary? **Lexical sophistication** asks questions such as: > How advanced, specific, infrequent, abstract or phraseologically developed are the lexical choices? Research on L2 writing increasingly treats lexical sophistication as multidimensional rather than as a single “rare word” score. ## Lexical diversity is not lexical density Another nearby idea is **lexical density**. A text can contain many content words relative to function words: > policy failure triggered rapid institutional reform. That is dense. A spoken sentence may use more grammatical scaffolding: > The policy failed, and because of that, the institution changed the way it worked quite quickly. The two sentences differ in density. Diversity asks: > How much lexical variety? Density asks: > How much lexical content is packed into the text? Do not confuse the two. ## Lexical diversity is not writing quality A student could artificially maximise lexical diversity: > happy, joyful, delighted, ecstatic, pleased, cheerful, jubilant… That is a varied list. It is not automatically a good paragraph. Writing quality also depends on: – accuracy; – relevance; – cohesion; – syntax; – evidence; – organisation; – tone; – task fulfilment. Vocabulary is one subsystem. The best vocabulary supports the reader’s job. ## Repetition can be good writing Consider a Science paragraph: > Temperature increased in Trial 1. Temperature also increased in Trial 2. However, the increase was smaller. Would replacing **temperature** each time improve it? > Heat level increased… thermal condition… warmth index… No. The repeated technical noun keeps the variable stable. In scientific writing, lexical repetition can protect reference. The reader should not have to wonder whether three near-synonyms denote the same measurement. ## Repetition can create cohesion A paragraph often needs a **lexical chain**. Example: > The proposal would extend school hours. The proposal has attracted support from some parents, but the proposal also raises questions about transport. This is repetitive. But the head noun: > proposal keeps the topic unmistakable. A controlled revision might use: > The proposal would extend school hours. The plan has attracted support from some parents, but it also raises questions about transport. Now the relationship remains clear because: > proposal → plan → it forms a coherent reference chain. The goal is not maximal variation. It is managed variation. ## Synonyms are rarely perfectly interchangeable A student sees: > important three times and replaces it with: > significant > crucial > vital. That may work. But each word carries different strength and collocational behaviour. Compare: > an important detail > a significant statistical difference > a crucial decision > a vital organ. The words overlap. They are not one interchangeable bucket. Vocabulary diversity without semantic control creates errors. ## Why forced synonym replacement can damage an essay Original: > Social trust is important because trust allows people to cooperate. Forced variation: > Social trust is important because confidence allows people to cooperate. But: > confidence is not necessarily the same concept as: > social trust. The paragraph has become less repetitive and less accurate. This is a central vocabulary principle: > preserve the concept before varying the word. ## Hypernyms and superordinates can vary vocabulary safely Instead of replacing a noun with an approximate synonym, writers can sometimes move up the category hierarchy. Example: > electric bus > vehicle > transport option These expressions are not synonyms. They are related through categorisation. The writer can use them when the broader category remains appropriate. This connects lexical diversity to hypernymy, hyponymy, semantic frames and reference. Vocabulary variation works best when the semantic relationship is explicit. ## Pronouns also reduce lexical repetition Text: > Mei opened the old notebook. Mei found a photograph inside the old notebook. Revision: > Mei opened the old notebook. She found a photograph inside it. The nouns disappear because: > she > it carry reference. This lowers lexical repetition but does not necessarily increase lexical diversity in the same way a synonym would. Cohesion can be grammatical as well as lexical. Students need several tools, not one “synonym rule”. ## Lexical diversity and lexical cohesion can pull in opposite directions A highly diverse text may have weak lexical cohesion. Imagine: > pollution > contamination > dirt > impurity > toxicity These words are related, but they are not identical. If the writer swaps them carelessly, the argument may drift. A cohesive technical article might deliberately repeat: > air pollution many times. The best writing balances: > stable concepts > + precise variation. ## Genre matters Narrative writing can tolerate or reward variation differently from Science. Narrative: > The rain thickened. Water streamed from the roofs. The road shone under the lamps. The writer can move across related images. Science: > Rainfall increased from 12 mm to 28 mm. Here: > rainfall is a measured variable. Replacing it with: > water from the sky would be stylistically varied and scientifically worse. Genre decides what variation is useful. ## Singapore examination writing In PSLE, Secondary English, O-Level/SEC English and GP-style writing, students are often encouraged to avoid obvious repetition. That advice is sensible when repetition comes from limited vocabulary. But the repair should be: > expand meaning choices, not: > replace every repeated noun. A strong student needs to know: – which words must remain stable; – which repetitions can become pronouns; – which phrases can be recast; – which synonyms change register; – which broader or narrower terms remain accurate. That is lexical control. ## Vocabulary diversity grows through reading, not synonym swapping A student who reads across fiction, news, essays, Science, History, biography and argument meets words in different contexts. Over time, the learner acquires: – alternatives; – collocations; – register; – semantic distinctions; – phrase frames. This produces usable lexical diversity. A synonym list alone rarely teaches when an alternative belongs. ## Diversity is partly phraseological Modern vocabulary research increasingly treats lexical ability as more than single-word choice. A writer may vary not only words but phrase frames: > plays a role in > contributes to > is associated with > results from > gives rise to > depends on But these phrases do not mean the same thing. Phraseological diversity is useful only when causal and relational distinctions remain intact. ## Lexical diversity can reveal limited vocabulary — sometimes Suppose a student repeatedly writes: > good > good > good > good across unrelated contexts: > good evidence > good person > good result > good explanation. The repetition may signal that the learner lacks more precise choices: > convincing evidence > considerate person > favourable result > clear explanation. Here diversity improvement and precision improvement align. The teacher should diagnose the semantic function of each **good**. Then teach alternatives by context. ## But repetition can reveal conceptual discipline A History essay repeatedly uses: > sovereignty because sovereignty is the central analytical concept. That may be excellent. Replacing it with: > power > freedom > control > independence could blur the argument. Vocabulary variety is not a higher goal than conceptual stability. ## A useful distinction: avoid accidental repetition, keep deliberate repetition Accidental repetition: > The event was important because it was an important event at an important time. The word repeats because the writer has not developed the thought. Deliberate repetition: > The policy changed incentives. Those incentives changed behaviour. The behavioural change then altered demand. Here repetition builds a causal chain. Ask: > Is the repeated word doing structural work? If yes, keep it. ## Lexical diversity and topic constraints A composition about friendship will naturally repeat: > friend > friendship > relationship > trust. A report about photosynthesis will repeat: > light > carbon dioxide > glucose > plant. A narrow topic constrains the vocabulary space. So diversity scores should not be compared across unrelated genres without caution. ## Function words distort naive comparisons TTR counts all word types unless the method says otherwise. Common function words: > the > a > of > in > to > and are repeated frequently. Some research therefore examines: > content-word diversity separately. Again, the metric design matters. A score is a model of a text. It is not the text itself. ## Lemmas and word forms matter Should: > run > runs > running > ran count as four types? A raw word-form TTR might say yes. A lemmatised analysis may group them under: > RUN. Different choices answer different questions. If the educational question is: > How many distinct lexical families does the student control? lemmatisation may matter. If the question is: > How many different surface forms appear? word-form counts may matter. Measurement follows purpose. ## Proper nouns can inflate diversity A travel composition may mention: > Singapore > Tokyo > Kyoto > Osaka > Shinjuku > Ueno. A naive diversity score increases. Did the learner demonstrate deeper vocabulary? Perhaps not. Named entities can affect metrics without representing general lexical range. This is another reason human interpretation remains necessary. ## Numbers and spelling variants can also distort automated scores A tool may count: > 20 > 21 > 22 as different tokens. Or: > organise > organize as different types. Or treat: > don’t differently depending on tokenisation. Do not treat a computational lexical-diversity score as objective truth without checking preprocessing. ## AI-generated text can look diverse without being precise A language model can produce many surface alternatives. That does not guarantee: – correct reference; – correct claim strength; – correct collocation; – correct disciplinary meaning. Students using AI should therefore evaluate: > Did the vocabulary become more accurate? not only: > Did it become more varied? Diversity can be generated cheaply. Precision remains the test. ## A teacher-friendly diagnosis ### Gap 1: narrow general vocabulary Student repeats broad words: > thing > good > bad > nice > problem. **Repair:** teach context-specific alternatives. ### Gap 2: technical repetition misdiagnosed as weakness Student repeats: > velocity in Physics. **Repair:** keep the stable technical term. ### Gap 3: synonym substitution changes meaning Student replaces: > evidence with: > proof. **Repair:** compare claim strength. ### Gap 4: reference chain breaks Student rotates: > proposal → issue → situation → matter. **Repair:** map which expressions actually refer to the same concept. ### Gap 5: metric obsession Student chases a higher TTR score. **Repair:** evaluate accuracy, cohesion and genre alongside diversity. ### Gap 6: text-length comparison error Teacher compares short and long texts using raw TTR. **Repair:** use same-length samples or length-robust measures. ## A practical vocabulary-variation method Take the repeated word: > problem. Do not immediately find a synonym. Ask: ### 1. Is it the same problem? If yes, repetition or pronoun reference may be best. ### 2. Is the second mention broader? Maybe: > issue. ### 3. Is it a cause? Maybe: > underlying factor. ### 4. Is it a consequence? Maybe: > complication. ### 5. Is it a measurable defect? Maybe: > error > fault > discrepancy. The right alternative comes from semantic role. ## A paragraph revision example Draft: > The traffic problem is a big problem because the problem causes delays. The problem also causes pollution. Poor revision: > The traffic problem is a major issue because the difficulty causes delays. The complication also causes pollution. Better: > Traffic congestion causes delays and increases pollution. The problem is especially severe during peak periods. Why is this better? The writer did not merely vary vocabulary. The writer reorganised the proposition. Lexical diversity improved as a consequence of clearer thinking. ## Lexical diversity and sentence design Repetition sometimes signals a syntactic problem. Draft: > Technology affects students. Technology affects teachers. Technology affects schools. Instead of synonym replacement: > Technology affects students, teachers and schools differently. One structural revision removes unnecessary repetition. Vocabulary and grammar interact. ## A quiet-luxury literary lens A controlled writer does not prove skill by displaying the largest possible vocabulary. The stronger principle is: > notice the exact object, action or relation, then choose the word that fits it. If a character touches the same brass key three times, repeating: > key may be better than rotating: > object > implement > device. Close observation gives vocabulary a reason. Variety should follow perception. ## Internal-link opportunities This article can connect to existing eduKateSG assets: – [Lexical Chunks and Phrase Frames](https://edukatesg.com/2026/08/28/lexical-chunks-phrase-frames-fluent-english/) – [Semantic Prosody in English Vocabulary](https://edukatesg.com/2026/08/29/semantic-prosody-hidden-attitude-vocabulary/) – [Colligation in English Vocabulary](https://edukatesg.com/2026/08/29/colligation-word-grammar-patterns-vocabulary/) – [How to Paraphrase Without Changing Meaning](https://edukatesg.com/2026/08/28/how-to-paraphrase-without-changing-meaning/) – [Adjective Entailment in English Vocabulary](https://edukatesg.com/2026/08/29/adjective-entailment-intersective-subsective-privative-vocabulary/) – [Lexical Entailment in English Vocabulary](https://edukatesg.com/2026/08/29/lexical-entailment-snore-sleep-vocabulary/) – [How Language Works](https://edukatesg.com/2026/08/26/how-language-works/) The distinct reader intent is: > repeated vocabulary → determine whether repetition is limitation or cohesion → improve lexical choice without damaging meaning ## Connections eduKateAI can learn **Lexical diversity ↔ variation:** diversity measures how varied the vocabulary of a sample is, but the interpretation depends on sample length, genre and metric. **Type ↔ token:** a type is a distinct word form; a token is an occurrence of a word. **TTR ↔ text length:** raw type-token ratio tends to decline as texts become longer, so unequal-length comparison is unsafe. **Diversity ↔ sophistication:** using many different words is not the same as using advanced, specific or phraseologically appropriate vocabulary. **Diversity ↔ cohesion:** deliberate repetition can maintain topic continuity and technical reference. **Synonymy ↔ semantic risk:** replacing repetition with a near-synonym can change register, strength, category or reference. **Genre ↔ lexical need:** Science, narrative, Humanities and formal argument require different balances between stable terminology and lexical variation. **Writing ↔ restructuring:** the best reduction in repetition often comes from clearer sentence architecture rather than thesaurus substitution. **AI ↔ evaluation:** automated diversity scores can be informative, but they should not replace semantic and discourse-level judgement. ## Final checkpoint A paragraph repeats: > evidence four times. Should the writer replace every occurrence? Not yet. First ask: – Does each occurrence mean the same evidence? – Is the repetition keeping the argument stable? – Would a synonym such as **proof** strengthen the claim incorrectly? – Can the sentence structure be improved instead? If the learner can answer those questions, lexical diversity has become a writing tool rather than a vocabulary score. ## Research basis This article was informed by the public research pass, including: – research on type-token ratio and the text-length problem; – Covington & McFall’s work on Moving-Average Type-Token Ratio; – McCarthy & Jarvis’s work on MTLD and lexical-diversity measurement; – Kim, Crossley & Kyle, **Lexical Sophistication as a Multidimensional Phenomenon**, *The Modern Language Journal*: https://onlinelibrary.wiley.com/doi/10.1111/modl.12447 – recent research on lexical complexity in English academic writing; – recent research on lexical reiteration and cohesion in EFL writing. The article deliberately avoids universal “good lexical diversity” cut-offs. Diversity is sensitive to text length, genre, tokenisation and the specific measure used.