A dispersion plot is a visual map of where a searched word or phrase occurs within one or more documents. In a common corpus-linguistic display, each document is represented by a horizontal bar, and the occurrences appear as short marks at their relative positions. The diagram shows whether the word clusters in one passage, recurs throughout a text or is absent from other documents. It does not, on its own, explain the meaning or correctness of the word.
Why does this matter for vocabulary? Because two texts can use the same word the same number of times and still do entirely different things with it. One may concentrate a term in a single explanatory section; another may weave it through a whole argument. A dispersion plot makes that hidden difference visible, helping students decide where to read more carefully and whether repetition is serving a real purpose.
This guide explains what dispersion plots show, how to interpret relative positions, why frequency and distribution differ, when clusters can mislead and how to use the method in Primary and Secondary English, Science, Mathematics, History, GP, translation and professional writing. Original examples and a paper-barcode exercise show how to learn the principle without paid software. Corpus-tool documentation is linked below.
Editorial note: All worked distribution scenarios are explicitly fictional and do not represent measured language frequencies. No corpus search or statistical dispersion score is claimed to have been executed for the invented examples.
Contents: why the location of a word changes the question
Open 46 chapters, examples and cautions
- 1. The plot shows positions, not just counts
- 2. Why vocabulary needs location information
- 3. What the horizontal document bar represents
- 4. A mark is a retrieved match, not a truth claim
- 5. Relative position differs from absolute location
- 6. Frequency and dispersion answer separate questions
- 7. Three fictional distributions with equal counts
- 8. A concentration can indicate a topic section
- 9. Repeated quotations can create a false hotspot
- 10. A dispersed word may be a narrative thread
- 11. A concentrated term can dominate a corpus count
- 12. The range measure and visual distribution differ
- 13. Burstiness is a descriptive idea
- 14. A plot is not a timeline of publication
- 15. A long document and a short document can look alike
- 16. Search rules shape the image
- 17. A missing document is not always irrelevant
- 18. Duplicate documents exaggerate reach
- 19. The plot does not measure positive or negative tone
- 20. Word senses can produce multiple underlying patterns
- 21. Concordance lines turn marks into language
- 22. Word sketches ask a different question
- 23. A plot cannot rank an author’s intelligence
- 24. Primary English: repeated story details
- 25. Secondary English: map a writer’s argument
- 26. Literary analysis: motifs need close reading
- 27. Science: distribution across report sections
- 28. Mathematics: recurring instructions and conditions
- 29. History: trace framing across a long source
- 30. GP: inspect claims and counterarguments
- 31. Translation: identical meanings can occupy different positions
- 32. Professional manuals: find where a warning is given
- 33. Public communication: don’t infer approval from repetition
- 34. Local Singapore varieties need appropriate data
- 35. A visually dramatic cluster can be a formatting artefact
- 36. Avoid using one colour as a complete explanation
- 37. An accessible text summary accompanies a plot
- 38. A classroom paper barcode exercise
- 39. A reproducible investigation note
- 40. Ethics of plotting student writing
- 41. AI visualisations can invent convincing evidence
- 42. A ten-minute lesson with one real sentence set
- 43. Questions that a plot cannot answer
- 44. How distribution supports editing
- 45. Never optimise an essay for a barcode
- 46. The final test is meaningful transfer
One frequency, three very different patterns
The following comparison is fictional practice data. Each invented text contains twenty target-word matches. The patterns differ only by position.
| Fictional document | Matches | Illustrative distribution | Question to investigate |
|---|---|---|---|
| A | 20 | Most marks in the first tenth | Does the introduction explain the term before moving to another topic? |
| B | 20 | Marks distributed along the document | Does the concept recur through the whole argument? |
| C | 20 | Most marks in the final tenth | Is the term mainly introduced in the conclusion or an appendix? |
These patterns demonstrate why raw totals cannot tell the full story. The table does not establish which text is better. Read the actual passages and consider purpose, length, structure and source before interpreting the marks.
1. The plot shows positions, not just counts
A dispersion plot in corpus linguistics visualises where searched words or phrases occur inside one or more texts. In a common barcode-style display, each document is represented by a horizontal bar and each matching occurrence by a short mark at its relative position.
This answers a different question from a frequency table. A count tells us how many occurrences were found; a plot shows whether they are spread across the document, clustered in one section or absent from other documents. Location can reveal patterns worth investigating, but it does not itself explain what the word means.
2. Why vocabulary needs location information
A word appearing twenty times in a single paragraph tells a different story from a word appearing twenty times across an entire book. The first may signal a concentrated topic, quotation or repeated instruction; the second may be part of a theme developed throughout the text.
A student who knows only the total can easily assume the word is equally important everywhere. The dispersion plot exposes distribution, encouraging closer reading of the passages where the occurrences gather. This is a technology of attention: it shows the reader where further interpretation may be most productive.
3. What the horizontal document bar represents
In a familiar concordance-plot layout, a horizontal rectangle represents one document from beginning to end. The left edge corresponds to the start and the right edge to the end. Marks indicate positions where the target expression was detected by the search.
Some tools normalise each bar to a common visual width even when documents contain different numbers of tokens. That helps compare relative positions, but it does not mean all documents are equally long. Read the display options before drawing conclusions about frequency or how much writing lies between two marks.
4. A mark is a retrieved match, not a truth claim
Each plotted mark corresponds to an occurrence returned by the query. If the search looks for the surface spelling charge, a mark could refer to a fee, a battery, an accusation or movement. The plot does not automatically disambiguate the senses.
A responsible researcher clicks or navigates to the original context and checks which meaning is intended. The spatial picture is a map of search results; the sentences themselves supply the linguistic evidence. Without this step, a neat distribution can support a false story about the text.
5. Relative position differs from absolute location
A mark halfway along a normalised bar means the occurrence is near the middle of that document under the tool’s positioning rule. It does not mean the occurrence appears at the same word number in every document. A long novel and a short notice can both show a mark at the midpoint.
To compare absolute positions, document lengths and measurement conventions must be included. The reader should ask whether the research question concerns relative structure—introduction, middle or conclusion—or absolute spacing in tokens, pages or sentences.
6. Frequency and dispersion answer separate questions
Frequency counts matches. Dispersion describes how they are distributed through texts or across documents. Two texts can contain the same number of matches but display very different patterns. One may have marks bunched together; the other may show marks throughout the bar.
A high-frequency expression concentrated in one chapter may be central to that chapter, not to the entire book. A lower-frequency expression appearing in every section may have a different structural function. Neither conclusion can be made confidently without reading the surrounding passages and understanding the document.
7. Three fictional distributions with equal counts
Imagine three wholly invented texts containing twenty matches of a chosen expression. In Text A, all matches occur near the opening; in Text B, they are spread along the length; in Text C, they occur mainly near the end. The totals are deliberately equal.
A dispersion plot would make these distinct patterns visually apparent. The example does not claim that any real author writes this way. It teaches that totals discard positional information, and it invites the next question: what is happening in each part of the text that might explain the distribution?
8. A concentration can indicate a topic section
An instructional article may discuss one technical term intensively in a dedicated section and mention it rarely elsewhere. The plot could show a dense cluster of marks corresponding to that section. That is a plausible hypothesis to test, not an automatic interpretation of the display.
Open several examples from the cluster and check headings, paragraph function and source context. If the section is a glossary or a repeated template, the cluster has a different significance from a sustained argument using the term repeatedly.
9. Repeated quotations can create a false hotspot
A writer might quote the same sentence several times for analysis. Each quotation can generate a match, even if the repeated wording originates from one speaker or earlier source. A visual cluster may therefore reflect quotation rather than the writer’s own preferred vocabulary.
Before inferring stance or style, distinguish the author’s assertions from reported words and copied material. This matters especially in History and literary criticism. The plot can point to a location for investigation, but attribution determines what can be claimed.
10. A dispersed word may be a narrative thread
In a novel or reflective essay, a recurring noun might appear at widely separated moments to connect scenes or develop an image. A dispersion plot can help a reader notice that the motif returns after long intervals.
The reader must still examine whether each occurrence refers to the same thing and whether the repetition is intentional or meaningful. A recurring spelling could have multiple senses. The plot highlights the pattern; close reading establishes its literary significance.
11. A concentrated term can dominate a corpus count
Suppose a collection contains many documents but one technical report repeats a specialist term hundreds of times. Its aggregate frequency may appear high even though most documents never use the term. A document-level dispersion analysis can reveal that concentration.
This is why researchers often ask how many distinct texts contain a word as well as how many total occurrences exist. A high hit count does not establish broad distribution. The distinction helps students avoid confusing an intensely discussed topic with a generally common expression.
12. The range measure and visual distribution differ
A simple range measure can describe how many documents or segments contain a target item. A dispersion plot shows where occurrences lie within the plotted documents. Range captures a form of spread across units; the visual display offers more detail about position and clustering.
Neither method should be mistaken for a complete model of language. Defining the units matters: a document, paragraph and equal-length segment are not interchangeable. Choose a measure or display that matches the research question, and state its limitations.
13. Burstiness is a descriptive idea
Researchers sometimes describe occurrences as bursty when they are concentrated in local stretches rather than evenly spaced. A technical topic may produce such bursts because a passage repeats its key concept while explaining it.
The observation does not by itself prove that an author is emotionally excited, unusually biased or careless. It may simply reflect how chapters are organised. Use a burst as a reason to read the relevant passage, not as a psychological diagnosis.
14. A plot is not a timeline of publication
The horizontal axis usually represents position within a text, not the calendar dates on which words were written, spoken or published. If a corpus contains documents from several years, a separate chronological analysis is required to investigate change over time.
Do not confuse the midpoint of a document with the midpoint of an author’s life or a historical period. A plotted word at 80% means it occurs late in the document under the display’s scaling—not that it appeared recently in real-world time.
15. A long document and a short document can look alike
When bars are normalised to equal width, two texts can show similarly spaced marks despite differing greatly in actual token length. A short article with five matches and a long report with five matches do not have the same normalised frequency.
Consult document lengths and raw counts before comparing density. If the question concerns frequency per thousand words, calculate an appropriate rate. If it concerns where a concept appears in the argument, relative positions may be the right choice.
16. Search rules shape the image
An exact-form query for write may yield a different pattern from a lemma query that includes writes, wrote and written. Case sensitivity, phrase matching, tokenisation and grammatical tagging also affect which occurrences become marks.
A reproducible plot needs a documented corpus, query and display configuration. If the researcher changes the search rule halfway through, the two images no longer describe the same target. A visual comparison is only as sound as the retrieval instructions behind it.
17. A missing document is not always irrelevant
Some interfaces may hide files that contain no matches, while others may display empty bars. This option changes the apparent picture of distribution across the collection. An image showing only documents with matches may exaggerate how widely the target occurs.
Record whether zero-hit documents were included. If the aim is to show a term’s presence across a whole corpus, hiding absence can mislead. For an exercise focusing on internal document structure, excluding empty files may still be reasonable as long as the choice is explicit.
18. Duplicate documents exaggerate reach
Copied webpages, template notices and repeated lesson materials can appear as separate documents in a corpus. A dispersion plot may then show apparently independent bars with almost identical patterns. Those bars may all descend from one original source.
Check source identifiers, dates and content duplication before interpreting breadth. A term repeated in dozens of copies is not necessarily used by dozens of independent writers. Provenance is part of honest vocabulary research.
19. The plot does not measure positive or negative tone
A word can appear throughout a text while the writer alternates between endorsement, criticism, quotation and irony. The marks indicate lexical matching, not whether each sentence is supportive or hostile.
For example, successful could occur in sincere praise or in ‘The plan was not successful.’ Treating both as equal positive sentiment would misread the text. Open the sentences and identify negation, source and communicative purpose before interpreting tone.
20. Word senses can produce multiple underlying patterns
The same spelling may refer to different concepts in different chapters. A book about banking could also use the word bank for a riverbank in an illustration. A single plot of surface occurrences can mix these senses.
If the research aims to trace the financial concept, manually check or use appropriately validated sense-specific methods. A semantic distribution plot requires classification beyond basic matching. The learner should recognise when one visible pattern hides several meanings.
21. Concordance lines turn marks into language
A concordance line presents an occurrence with words on either side. In suitable corpus software, a plot result may allow the user to open or navigate to the matching passage. This connects the broad map to the actual sentence.
The most useful sequence is: observe a pattern in the plot, inspect examples in a concordance or file view, identify the grammatical and semantic function, then revise the hypothesis. Switching between overview and detail is the skill the visual tool is meant to support.
22. Word sketches ask a different question
A word sketch groups a word’s grammatical and collocational partners. A dispersion plot shows where the word or query results appear in the texts. One focuses on relationships among words; the other on distribution through documents.
They can complement each other. A student may first notice that a technical term clusters in a Science section, then inspect which verbs and modifiers occur with it. Neither tool independently establishes the scientific truth of the section’s claims.
23. A plot cannot rank an author’s intelligence
An essay with concentrated uses of one technical term is not automatically weak, and a story with evenly spread descriptions is not automatically sophisticated. Repetition can be necessary for precise reference, and sparse use can result from deliberate style or simple topic differences.
Do not assign grades based on the smoothness of a dispersion pattern. Evaluate whether language serves the writing task, maintains accurate meaning and helps a real reader. The plot is an analytical instrument, not a test of someone’s worth.
24. Primary English: repeated story details
A Primary learner can read a short original story where the word umbrella appears at the beginning, middle and end. The teacher draws a simple bar and places three marks at the corresponding sections. The child explains what the umbrella means in each scene.
This invented exercise makes the visual principle accessible without claiming real corpus findings. The learner then writes a fresh story where an object returns with a changed role. The educational goal is narrative coherence and purposeful repetition, not software fluency.
25. Secondary English: map a writer’s argument
In an expository passage, words associated with the problem may cluster in the introduction while solutions are discussed later. A student can plot selected terms and test whether the text actually follows that structure.
The temptation is to assume that a mark near the conclusion carries the final answer. Instead, read the whole paragraph and distinguish restatement, counterargument and recommendation. The map can direct attention to evidence, but the passage determines interpretation.
26. Literary analysis: motifs need close reading
An image such as a door, journey or shadow may recur in a novel and contribute to theme. A dispersion plot can help locate occurrences across chapters, especially when a long work is difficult to search manually.
The student should compare the scenes and explain how context changes the image’s meaning. Counting the word door does not prove every occurrence is symbolic; many could be ordinary physical descriptions. Interpretative claims require close reading.
27. Science: distribution across report sections
A Science report might use method, results, uncertainty and conclusion in different structural positions. A plot could help students notice how information is organised, provided the chosen corpus contains appropriate reports.
This does not mean a scientifically important concept must appear evenly. An uncertainty statement may occur primarily in discussion, and that could be appropriate. The scientific method and actual evidence determine what the report should say.
28. Mathematics: recurring instructions and conditions
A Mathematics text may repeat terms such as assume, therefore, given and prove as an argument develops. A distribution analysis can reveal where the author introduces premises, uses transitions or revisits a key condition.
Yet the correctness of a proof does not follow from the presence or position of those words. Students must examine the mathematical statements, their definitions and the logical links. The plot is a reading support tool, not a proof checker.
29. History: trace framing across a long source
A long historical report might shift between terms such as resistance, security and reform. A dispersion plot can help locate where each framing appears. The researcher can then inspect its speaker, date and evidential basis.
Avoid turning the word positions into unsupported claims about the author’s motives. The same term may appear inside a quotation or be criticised by the writer. Historical understanding requires corroboration and source analysis in addition to linguistic pattern recognition.
30. GP: inspect claims and counterarguments
An argumentative essay may introduce a claim, discuss objections and revisit its thesis in the conclusion. Searching for central terms can reveal possible structural patterns. A learner can use the plot to check whether important concepts are developed or merely mentioned at one point.
A good GP essay still needs sound evidence, clear reasoning and calibrated conclusions. An evenly distributed keyword does not guarantee those qualities. The visual should prompt better revision questions: where was a term defined, supported, challenged and resolved?
31. Translation: identical meanings can occupy different positions
Translated documents do not necessarily preserve word order or token counts. One language may express a concept through several words or a different grammatical construction. A plot of identical strings in a source and translation therefore does not automatically show conceptual alignment.
If positional comparison is useful, work with carefully aligned passages or independently checked equivalent expressions. Verify that a target-language term refers to the intended concept, rather than assuming every mark represents the same meaning across languages.
32. Professional manuals: find where a warning is given
A safety manual may discuss a hazard in one introductory section and then provide specific precautions later. Searching for the warning term and looking at its distribution can help an editor inspect whether relevant instructions are easy to find.
The plot cannot determine whether the manual meets regulatory or safety requirements. The current authoritative standard and the reader’s actual task remain essential. Language tools can identify passages for review; qualified people must verify consequential content.
33. Public communication: don’t infer approval from repetition
A policy briefing might repeat approved in quotations, proposals and final decisions. A dispersion plot can show where the word appears, but it cannot distinguish a proposal from a completed authorisation without contextual analysis.
Read the exact clause, actor, date and procedural record. This prevents a common language error: converting the appearance of a word into a claim that an event occurred. Vocabulary skill includes knowing when a term’s factual force requires external confirmation.
34. Local Singapore varieties need appropriate data
The distribution of a phrase in one school corpus may differ from its use in local conversation, formal media or international academic writing. A plot of one sample tells us where the matches occur there, not how Singaporeans as a whole use English.
Select sources suitable for the target audience and avoid treating legitimate local expressions as errors simply because they are absent from a foreign corpus. A careful reader asks who produced the texts and what situations they represent.
35. A visually dramatic cluster can be a formatting artefact
Repeated headers, copied glossary boxes or page-navigation text may create artificial clusters of matches in digital documents. If the source extraction includes irrelevant menus or repeated templates, the plot can be distorted before any analysis begins.
Check the raw text and preprocessing. A result that depends on boilerplate removal should say so. Attractive graphs can conceal poor source handling; the integrity of the underlying text matters more than the appearance of the output.
36. Avoid using one colour as a complete explanation
Some plotting tools let users overlay multiple queries or display different series visually. Colours can help distinguish searched items, but they do not make semantically different terms automatically comparable. Overlap may reflect common topics, shared templates or chance.
Write the exact query associated with each visual layer and check the examples behind apparent intersections. If one term is a broad concept and the other a narrow specialist expression, the comparison should recognise those differences instead of treating marks as equivalent units.
37. An accessible text summary accompanies a plot
A visual plot may be difficult to interpret for readers who use screen readers, have colour-vision differences or cannot distinguish tiny marks. A responsible report should summarise the observable pattern in text and provide data values or a readable table when available.
For example, say that the target appears mainly in the final third of the document in the illustrative dataset, rather than relying on a colour-coded bar alone. Do not invent counts the plot does not provide. Accessibility improves the clarity of the underlying evidence for everyone.
38. A classroom paper barcode exercise
Draw three equal-length bars representing three invented short texts. Place marks at each occurrence of a chosen term using relative position. Give all three the same number of occurrences but distribute them differently: clustered early, spread evenly and clustered late.
Ask students to explain what the bar reveals and what it cannot tell them. The answer should mention relative position, the importance of reading the sentences and the lack of evidence about real-world frequency. The paper exercise teaches the reasoning without specialist software.
39. A reproducible investigation note
For each plot, record the corpus or files, text dates and genres, query string, matching rules, document lengths, whether bars were normalised, whether zero-hit files are shown and which examples were inspected. Include a date of access when sources can change.
This record helps another reader understand why the plot appears as it does. It also prevents silent method changes from turning two non-comparable images into a persuasive but misleading story about vocabulary use.
40. Ethics of plotting student writing
A classroom corpus of student essays could reveal where certain words appear, but the underlying documents may contain names, school information and sensitive personal stories. Sharing a plot is not automatically harmless if it identifies individual papers or allows excerpts to be recovered.
Follow permissions, school rules and data-protection requirements. Use invented samples for general demonstrations. If genuine learner writing is legitimately analysed, de-identify and restrict access as appropriate and describe the limited scope of any conclusions.
41. AI visualisations can invent convincing evidence
A generative model can draw a plausible barcode plot without having retrieved real text. Such an image can illustrate the concept but cannot establish where an expression occurs in authentic documents unless a transparent data-processing step produced it.
Label fictional visualisations as teaching examples. If empirical results are needed, run the actual search, save the query and verify the source contexts. A beautiful chart is not a substitute for observed language evidence.
42. A ten-minute lesson with one real sentence set
A tutor can choose a short, appropriately sourced passage and identify a target expression. The learner marks its occurrences on a paper bar, observes the pattern, reads the matching sentences and writes a short explanation of what the pattern might mean.
The time is only an illustrative teaching plan; it is not a measured learning result. If the learner becomes interested in an exception, extend the exercise. The valuable outcome is the habit of moving from a visual clue to a contextual explanation.
43. Questions that a plot cannot answer
A dispersion plot cannot by itself tell us whether a sentence is true, whether a word is offensive, whether a writer intended irony, whether an expression is grammatical in all contexts or whether a student understood the word.
Those questions need dictionaries, grammar analysis, evidence from the document, social context or a direct comprehension task. The correct use of a tool includes recognising its boundary. The plot is a map of positions, not an all-purpose judge of meaning.
44. How distribution supports editing
A writer revising a long article may find that a key term appears only in the opening and conclusion, leaving the middle insufficiently connected. Alternatively, the same term may be repeated mechanically in every line. A plot can suggest where to inspect, but not automatically which revision is warranted.
Read the sentences first. Some repetition is necessary for precision and cohesion; some can be replaced by clear reference or better paragraph structure. The goal is to help the reader follow the actual ideas rather than to achieve a mathematically uniform keyword pattern.
45. Never optimise an essay for a barcode
If students begin inserting target words merely to spread marks evenly across a plot, the tool has displaced the educational objective. Readers need a coherent argument, not engineered recurrence. A strong essay may legitimately introduce specialist vocabulary only where the subject demands it.
Use dispersion analysis diagnostically: it can surface areas worth checking for missing explanations, abrupt topic shifts or excessive repetition. It cannot prescribe an ideal shape for every genre. A pattern is useful when interpreted against the task and reader.
46. The final test is meaningful transfer
After viewing a distribution, close the chart and ask the learner to explain one specific, evidence-supported feature of the text. Then ask how that insight might improve a new sentence, composition or reading strategy.
This is vocabulary as a technology of attention and judgement. A word’s frequency matters, but so does where it appears, what it means there and what role it plays in the document. The best plot leads back to better reading, not further away from the words.
An original five-step dispersion activity
- Choose a short text and one expression that matters to its meaning.
- Mark every exact occurrence under a defined search rule.
- Represent the document with one bar and place marks at approximate relative positions.
- Compare the visual pattern with the actual passages in which the expression appears.
- Write a narrow, defensible interpretation and identify one conclusion the plot cannot establish.
Ten frequently asked questions
1. What does a dispersion plot show?
Where query matches appear within texts or across documents under a defined search and display rule.
2. Is it the same as a word frequency table?
No. A table counts occurrences; a plot provides positional or distribution information.
3. Does one bar always represent the same number of words?
No. Bars may be normalised to similar visual widths even when texts have very different lengths.
4. Can a plot show a recurring motif?
It can locate possible repeated instances, but literary interpretation requires close reading.
5. Does a cluster prove that a term is important?
No. It can reflect a topic section, quotation, template or copied material.
6. Does an empty bar prove the expression is impossible?
No. It indicates no matches under the specified search conditions in that represented text.
7. What is burstiness?
A descriptive tendency for occurrences to cluster in particular stretches rather than appearing evenly spaced.
8. Is the horizontal axis calendar time?
Usually not. It commonly represents relative position inside a document.
9. Can a Primary child learn this without software?
Yes. Marks on paper bars can show the principle using original short passages.
10. What is the learning outcome?
Better questions about a text and more accurate, evidence-based explanations of word use.
Verified reference guides and further reading
- University of Mannheim: Corpus Search Tools and Concordance Plots — how plotted hits show the position and concentration of matches in texts.
- AntConc: Concordance Manual — use concordances and source context to interpret retrieved hits.
- Laurence Anthony: AntConc Official Website — the corpus toolkit associated with concordance and plotting tools.
- Sketch Engine: Concordance Guide — related corpus queries, context views and filters.
- British National Corpus: Corpus Description — why dataset scope and representativeness matter.
- Vocabulary | What Is a Corpus Query? — understand what becomes a plotted match.
- Vocabulary | What Is a Concordance Line? — read examples behind each visual mark.
- Vocabulary | What Is a Corpus Comparison? — compare appropriately selected sources.
- Vocabulary | What Is a Word Token? — know what the search is counting.
- Vocabulary Learning Hub — the complete vocabulary pathway.
A dispersion plot is a way to direct attention back to evidence. It shows where words appear, then asks a reader to examine what they mean in those locations. That is how a visual technique becomes part of a usable vocabulary system rather than another attractive chart.