Research is often imagined as the stage where an author goes out, gathers facts, then returns to write. Serious authorial research is more complicated. Research changes the question, the structure, the argument, the characters, the examples, the tone of certainty and sometimes the entire reason a book exists.
An author may begin convinced that a problem has one cause and discover five. A novelist may research a profession simply to make a character credible and end up finding the book’s central conflict. A historian may enter an archive expecting confirmation and instead find a contradiction that rearranges the narrative. A science writer may discover that the most interesting part of a subject is not what experts know, but where experts still disagree.
Good research therefore does not merely decorate writing with authority. It gives the work resistance. It forces the author to encounter what they did not already know.
This article belongs to the How Authors Work series. It extends How Authors Find Ideas, How Authors Build Voice and How Authors Revise and Rewrite by examining how writers build a trustworthy relationship with evidence.
The Short Answer
Authors research by moving through a recurring loop:
- define the working question;
- map what is already known;
- identify the strongest available sources;
- collect material from books, papers, documents, archives, interviews, field observation, data and other relevant evidence;
- record provenance so every important claim can be traced back;
- compare sources rather than trusting the first persuasive one;
- look deliberately for contradictions and counterexamples;
- distinguish fact, interpretation and speculation;
- synthesise patterns across material;
- revise the project when the evidence changes the question;
- verify the final claims before publication.
The deepest rule is simple: research should be capable of changing the author’s mind.
1. Research Begins With a Question, Not a Search Box
A search box can retrieve material. It cannot decide what matters.
Before gathering sources, the author needs a working question. It need not be final. It needs only enough structure to guide attention.
“Education” is too broad. “Why do some students remember vocabulary during practice but fail to retrieve it under exam pressure?” creates a research path. “Aviation” is too broad. “How do airlines build resilience when efficiency pushes them toward tighter schedules?” gives the author something to test.
A research question reduces noise. Without it, the author can spend weeks accumulating information that never becomes useful.
2. The First Question Is Usually Too Simple
Early questions are valuable because they begin movement, but they often assume more than the author realises.
“Why did this policy fail?” already assumes failure. “Why are young people reading less?” assumes the measurement is correct and the category “reading” is stable. “Why did this company succeed?” may ignore survivor bias or the possibility that success depends on the chosen metric.
Research improves when the author learns to inspect the question itself.
3. Turn the Question Into Subquestions
Large projects become manageable when divided into smaller unknowns.
- What happened?
- When did it happen?
- Who was involved?
- What mechanisms might explain it?
- What evidence supports each explanation?
- What evidence contradicts it?
- What changed over time?
- What does the standard account leave out?
- What can be known confidently?
- What remains uncertain?
Subquestions prevent the author from treating one large mystery as though it were a single fact to be found.
4. Build a Knowledge Map Before Deep Research
Before diving deeply into sources, map the terrain.
What fields touch the subject? Which institutions matter? What terminology do specialists use? What historical periods are relevant? Which debates already exist? Which datasets, archives, agencies, journals, books or professional communities may contain useful evidence?
This orientation phase prevents an author from searching only with the vocabulary they began with. Better terminology produces better retrieval.
5. Learn the Vocabulary of the Field
Every domain has distinctions that disappear in everyday language.
A medical term may separate symptoms that non-specialists group together. A legal field may distinguish possession, ownership and control. An engineer may use “failure” differently from a journalist. A historian may distinguish a primary source from a later interpretation.
Learning the field’s vocabulary is not about sounding technical. It allows the author to ask more precise questions.
6. Separate Primary Sources From Secondary Sources
A primary source comes directly from the period, event, person, experiment or system being studied: a letter, interview, court record, dataset, photograph, lab result, meeting minute, diary, artefact or first-hand observation.
A secondary source analyses, interprets or synthesises primary material: histories, reviews, essays, textbooks, documentaries, explanatory articles and many academic publications.
Neither category is automatically superior. Primary sources provide direct evidence but can be biased, incomplete or misleading. Secondary sources can offer context and comparison but may reproduce inherited errors.
Strong authorial research uses both and understands their different jobs.
7. Tertiary Sources Are Useful for Orientation
Encyclopedias, introductory guides, broad reference works and summaries can help an author learn the landscape quickly.
The mistake is treating orientation material as the final evidentiary layer for important claims.
Use summaries to discover names, dates, terminology and source trails. Then follow those trails toward stronger evidence.
8. Follow Citations Backward
A valuable source often points toward older and stronger sources.
If an article cites a study, read the study. If a history book quotes a letter, locate the letter if possible. If a report repeats a statistic, trace the number to the dataset or institution that produced it.
Backward citation tracing helps the author distinguish original evidence from repeated retelling.
9. Follow Citations Forward
Older research may have been challenged, refined or superseded.
Ask who cited the source later. Did subsequent work confirm it? Did methods improve? Did new evidence change the interpretation?
Research is a conversation across time. Reading only the earliest influential work can freeze the author inside an outdated debate.
10. Search for the Source Behind the Statistic
Numbers acquire authority simply by being numerical.
Authors should resist that effect.
Who measured the quantity? How? What population was included? What was excluded? Was the number estimated or directly counted? What year does it describe? Has the methodology changed?
A number without provenance is not yet secure evidence.
11. Read Methods, Not Just Conclusions
Research findings depend on how the research was done.
Sample size, selection method, measurement design, definitions, control groups, survey wording, model assumptions and missing data all affect interpretation.
An author does not need to become a specialist statistician for every project. But if a major claim rests on a study, the author needs enough methodological understanding to know what the study can and cannot support.
12. Distinguish Observation From Inference
“The temperature rose three degrees” is an observation if measured reliably.
“The machine failed because of overheating” is an inference unless the causal mechanism is established.
Authors create stronger work when they mark the boundary between what was observed and what is being interpreted.
13. Distinguish Fact From Interpretation
A historical document may record that a meeting happened on a certain date. Why the meeting mattered may remain interpretive.
An interviewee may state what they believe motivated a decision. That belief is evidence about the interviewee’s perspective, not automatic proof of everyone’s motives.
This distinction lets authors write confidently without pretending that every claim has the same evidentiary status.
14. Distinguish Interpretation From Speculation
Interpretation is reasoned from evidence. Speculation goes beyond what the available evidence can establish.
Speculation is not always forbidden. It can be useful when clearly labelled, especially in reconstructing missing history or exploring possible mechanisms.
The danger is presenting speculation with the grammar of fact.
15. Build a Source Hierarchy
Not every source deserves equal weight.
- Direct evidence: records, datasets, first-hand observations, original publications.
- Authoritative synthesis: strong scholarship, institutional reports, specialist reviews.
- Expert interpretation: credible specialists explaining the evidence.
- General explanation: journalism, textbooks, educational summaries.
- Discovery material: blogs, forums, social media, informal discussion.
Discovery material can be valuable because it reveals leads and lived experience. But major factual claims should move upward toward stronger evidence wherever possible.
16. Authority Is Not the Same as Accuracy
Prestigious institutions can make mistakes. Experts can disagree. Widely cited books can contain errors.
Authority is a useful signal, not a substitute for checking.
When a claim is important, triangulate across evidence rather than relying on reputation alone.
17. Consensus and Minority Views Need Different Treatment
Some debates have a strong expert consensus. Others remain genuinely open.
An author should not create false balance by presenting a fringe view as equally supported. Nor should they hide legitimate disagreement simply because one interpretation is more common.
Research includes understanding the shape of disagreement.
18. Search for the Strongest Counterevidence
If the author hopes a claim is true, confirmation bias becomes a practical research risk.
Deliberately search for the strongest evidence against the preferred explanation. Find the best contrary case, not the weakest.
If the argument survives strong opposition, confidence becomes more defensible. If it fails, the research has done its job.
19. Contradictions Are Research Assets
Two good sources disagreeing is not merely an inconvenience.
The disagreement may reveal different definitions, different time periods, different incentives, incomplete records or genuine uncertainty.
Instead of choosing quickly, ask why both sources can appear credible while reaching different conclusions.
20. Absence Can Be Evidence, But Carefully
If a record never mentions a person, event or practice, that absence may matter.
But absence is difficult to interpret. The record may be incomplete. The institution may not have considered the information worth preserving. A category may not yet have existed.
Authors should avoid turning silence into certainty without understanding the archival conditions that produced the silence.
21. Archives Are Designed Systems, Not Neutral Containers
Archives reflect decisions about what was created, collected, classified, preserved, digitised and made accessible.
That means archival research has two subjects at once: the documents and the system that determined which documents survived.
A skilled author asks not only, “What is here?” but “Why is this here, and what might never have entered the archive?”
22. Photograph Documents With Context
When archives permit photography, capture enough metadata to identify the source later.
A photograph of a fascinating page is nearly useless months later if the author cannot remember the collection, box, folder, date or document title.
Research systems fail most often not when information cannot be found, but when found information cannot be traced back.
23. Interviews Are Evidence and Relationships
Interviews provide information that documents may not contain: tacit knowledge, memory, interpretation, motive, emotion and professional practice.
But interviews are not simple extraction. They involve trust, power, expectation and context.
The author must understand what was agreed: on the record, off the record, background, anonymous or otherwise restricted.
24. Ask Open Questions Before Leading Questions
“Tell me what happened next” allows the interviewee to reveal their own structure.
“Were you angry because management ignored you?” supplies both emotion and cause.
Leading questions can be useful later for testing a specific hypothesis, but open questions often produce more surprising material at the beginning.
25. Ask for Examples, Not Only Opinions
“The system is inefficient” is an opinion.
“Walk me through the last time the system failed” can produce observable detail.
Examples allow the author to see mechanisms rather than only conclusions.
26. Ask Experts What Non-Experts Usually Get Wrong
Specialists often know where public explanations oversimplify a field.
Questions such as “What does the public usually misunderstand?” or “Which explanation sounds right but is technically wrong?” can reveal important distinctions quickly.
The author should still verify the answer, but it provides a valuable research direction.
27. Ask What Would Change the Expert’s Mind
This question reveals epistemic boundaries.
An expert who can explain what evidence would alter their view is often giving the author a map of the real debate.
It also helps distinguish strong knowledge from professional habit.
28. Memory Interviews Need Verification
People remember meaning and sequence imperfectly.
Two sincere witnesses may remember the same event differently. That does not necessarily mean one is lying.
Where factual precision matters, compare memories with records, dates, photographs, messages and other testimony.
29. Fieldwork Shows What Documents Hide
Processes often look different on paper than in practice.
A logistics manual describes the official workflow. Watching a warehouse may reveal improvisation, workarounds and tacit rules. A school policy describes a procedure. Sitting in a classroom shows how human timing alters it.
Field observation can therefore reveal the difference between designed systems and lived systems.
30. Observation Needs a Protocol
“I went there and looked around” is not automatically reliable observation.
What was the author looking for? At what times? From which position? How long? Were participants behaving differently because they knew they were being watched?
Even informal fieldwork improves when the author records conditions systematically.
31. Write Field Notes Quickly
Memory converts detail into summary fast.
Record sensory details, sequences, quotations, spatial relationships and surprises while they remain available.
Later, separate raw observation from interpretation so the author can remember which part was actually seen and which part was inferred.
32. Photographs Are Evidence With Frames
A photograph shows something real, but only within a selected frame and moment.
What lies outside the image? Was the scene staged? What happened before and after? Who made the photograph, and for what purpose?
Visual evidence needs contextual reading just as written documents do.
33. Maps Contain Arguments
Maps can appear objective because they are spatial.
But every map selects boundaries, labels, scale and categories. A transport map emphasises connectivity. A political map emphasises jurisdiction. A historical map may impose later borders onto earlier worlds.
Authors should read maps as designed representations, not transparent reality.
34. Objects Can Be Primary Sources
Material culture can reveal information absent from text.
Wear patterns, repair marks, manufacturing methods, size, material choice and design can reveal use, scarcity, status or adaptation.
An object is not self-explanatory, but it can ask questions documents never ask.
35. Data Needs Definitions Before Analysis
What counts as a school dropout? A household? A crime? A delay? A death caused by a disease? A successful project?
Definitions can change across institutions and years.
Before comparing numbers, make sure the categories are actually comparable.
36. Denominators Matter
A raw count can mislead.
One hundred incidents may be many or few depending on the population, exposure time, number of journeys or amount of activity.
Authors should ask “out of how many?” whenever a number is being used rhetorically.
37. Correlation Is Not Automatically Causation
Two variables moving together may indicate a causal relation, a shared cause, reverse causation or coincidence.
Authorial research becomes stronger when it asks what mechanism could connect the variables and what evidence would distinguish competing explanations.
38. Averages Can Hide Distribution
An average improvement can conceal some people improving greatly and others getting worse.
An average income can conceal inequality. An average delay can hide rare but severe failures.
Authors should inspect variation when variation matters to the story.
39. Time Series Can Reveal Different Stories
A single-year statistic freezes a system.
Looking across time can reveal whether a change is new, cyclical, gradual, sudden or part of a longer trend.
Historical depth prevents authors from treating every contemporary pattern as unprecedented.
40. Comparison Requires Comparable Cases
Authors often compare countries, schools, companies or communities because contrast makes explanation easier.
But cases can differ in size, history, institutions, geography or measurement.
Good comparative research explains both the similarities that make comparison useful and the differences that limit it.
41. Notes Need Provenance
The most dangerous research note is a perfect sentence whose source has been forgotten.
Record author, title, date, page, URL or archive identifier when the note is made, not months later.
Provenance turns notes into auditable material.
42. Separate Quotations From Paraphrases in Notes
Research notes can accidentally blur source language with the author’s own words.
Mark direct quotations clearly. Record paraphrases in a different format. This reduces accidental plagiarism and makes later checking easier.
43. Record Why a Source Matters
A saved article without context becomes a digital pile.
Add a short note: supports mechanism A; contradicts timeline; useful example of implementation gap; possible interview lead; strong definition; weak methodology but important historical influence.
This converts storage into research memory.
44. Build a Claim Ledger
For large nonfiction projects, maintain a list of important claims and the evidence supporting each one.
- Claim.
- Source.
- Strength of evidence.
- Counterevidence.
- Uncertainty.
- Where the claim appears in the manuscript.
A claim ledger makes final fact-checking far easier and reveals weak sections before publication.
45. Build a Timeline When Time Matters
Human memory and narrative instinct both compress chronology.
A timeline can expose impossible sequences, gaps, causal assumptions and events that occurred in a different order than the story initially suggests.
For history, biography, investigations and complex fiction, timelines are structural tools as much as research tools.
46. Build a People Map
Large stories contain networks.
Who knew whom? Who reported to whom? Who funded whom? Who opposed whom? Who shared institutions, schools, companies or families?
A relationship map can reveal hidden structure that chronological notes do not.
47. Build a Source Gap List
Authors often collect what is easy to find and overlook what is missing.
Maintain a list of unresolved gaps: missing year, absent perspective, unsupported mechanism, unclear number, unavailable interview, disputed date.
The gap list turns uncertainty into a work queue.
48. Synthesis Is Not Summarising Sources One by One
Research becomes writing when the author stops organising material by where it came from and starts organising it by what it means.
Instead of: Source A says this, Source B says that, Source C says something else.
Ask: what pattern appears across all three? Where do they disagree? What mechanism could explain the difference? Which evidence is strongest?
Synthesis is relational thinking across sources.
49. Group Notes by Question, Not by Source
Once enough material exists, reorganise it around the project’s questions.
This helps the author see where several independent sources converge, where only one source supports a major claim, and where evidence remains thin.
50. Look for Mechanisms
A pattern becomes more explanatory when the author can describe how one state leads to another.
What incentives operate? What information moves? What physical process occurs? What institutional rule changes behaviour? What psychological pressure alters a decision?
Mechanisms turn correlation into a testable story of causation.
51. Look for Boundary Conditions
When does an explanation stop working?
A teaching method may help beginners but not experts. A policy may work in dense cities but not rural regions. A business model may succeed while capital is cheap and fail when financing conditions change.
Boundary conditions make arguments more precise and less ideological.
52. Separate Necessary Conditions From Sufficient Conditions
A factor can be necessary without being enough by itself.
Water may be necessary for plant growth, but water alone is not sufficient. Trust may be necessary for teamwork, but trust alone does not guarantee performance.
This distinction prevents authors from turning one important factor into a complete explanation.
53. Search for Selection Effects
Who enters the dataset, archive, institution or story?
Studying only successful companies cannot explain failure well. Interviewing only people who remained in an organisation misses those who left. Preserved letters may overrepresent literate elites.
Selection determines which reality becomes visible.
54. Survivorship Bias Produces Attractive Stories
Successful people can explain what they did, but many unsuccessful people may have done similar things.
An author researching success should look for comparable cases that disappeared, failed or were never documented as prominently.
Otherwise the research may mistake surviving traits for causal traits.
55. Institutions Produce Their Own Records
Annual reports, press releases, policy papers and official histories are valuable sources, but they are also institutional self-representations.
Ask what the document was designed to accomplish. Inform investors? Satisfy regulators? Defend a policy? Preserve institutional memory?
Purpose shapes language.
56. Media Reports Are Leads and Contemporary Records
Journalism can provide chronology, interviews, public reaction and leads to documents.
But reporting occurs under deadline, editorial and access constraints. Early reports of fast-moving events can be incomplete.
Authors using media archives should compare outlets, follow corrections and distinguish contemporaneous knowledge from later reconstruction.
57. Social Media Can Reveal Lived Experience and Distortion
Social platforms provide direct traces of language, reaction, networks and community experience.
They also contain bots, performance, selection effects, virality incentives and context collapse.
A viral post is evidence that a post went viral. It is not automatically evidence that its view is typical.
58. Forums Can Be Excellent Discovery Tools
Specialist forums and community discussions often surface practical problems formal publications ignore.
They can reveal terminology, edge cases, frustrations and insider heuristics.
Use them to discover questions and leads. Verify consequential claims elsewhere.
59. Search Engines Rank, They Do Not Judge Truth
Search results are ordered by systems optimised for relevance, authority signals, engagement and other ranking factors—not by a perfect truth detector.
An author should treat search as retrieval infrastructure. Evaluation begins after retrieval.
60. Artificial Intelligence Can Accelerate Discovery
Generative AI can help produce search terms, explain unfamiliar concepts, compare terminology, summarise notes, generate interview questions and identify possible counterarguments.
That makes it useful during orientation and synthesis.
But AI-generated language is not automatically a source. It can produce false citations, incorrect claims or confident simplifications.
61. Use AI to Generate Leads, Then Verify the Leads
A productive workflow is:
- ask for possible terminology, institutions, source types or debates;
- treat the answer as a hypothesis about where to look;
- retrieve the actual sources;
- verify claims against those sources;
- record the real provenance rather than citing the generated summary.
The model expands the search field. The author retains evidentiary responsibility.
62. AI Summaries Can Hide the Very Detail the Author Needs
Summaries compress.
Compression removes caveats, methodological detail, anomalies and rhetorical context—the exact features that may matter most in serious research.
Use summaries to orient. Read originals for claims that matter.
63. Research Ethics Begin Before Publication
Research choices affect people long before a manuscript is released.
Interview subjects may disclose private information. Communities may be identifiable through detail. Vulnerable people may face consequences from exposure.
Ethical research asks what consent means, what promises have been made and what risks the project creates.
64. Consent Is Context-Specific
A public statement can often be quoted differently from a private conversation. A casual remark at dinner is not automatically an interview. A child cannot be treated exactly like an adult source.
Authors need clear agreements when the material is sensitive.
65. Anonymity Must Be Realistic
Removing a name does not guarantee anonymity if the surrounding details identify the person.
Job title, location, date, family relation and distinctive event can reconstruct identity.
When anonymity matters, revise the entire identification pathway, not only the name.
66. Quotation Creates Responsibility
Quoted language carries unusual force because readers treat it as direct evidence.
Verify wording. Preserve context. Do not splice statements in ways that create a meaning the speaker did not express.
Accuracy is ethical as well as technical.
67. Research Can Become Exploitative
Writers sometimes enter communities only when they need material.
That can create extraction: other people supply risk, trauma and labour while the author receives narrative value.
There is no universal formula for preventing this, but awareness of power, reciprocity and consequence should shape research design.
68. Research for Fiction Still Needs Responsibility
Fiction is invented, but the worlds it borrows from are real.
Authors researching professions, cultures, illnesses, disasters or historical violence should distinguish imaginative freedom from careless stereotype.
Research gives fictional invention more constraints—and therefore more credibility.
69. Stop Researching When Research Becomes Avoidance
Research can feel productive forever.
There will always be another book, archive, interview or paper.
The author needs a threshold: enough material to draft the next section responsibly, with a clear list of uncertainties to return to later.
Drafting reveals what research is still necessary. Endless pre-research cannot.
70. Research and Writing Should Alternate
A strong workflow often moves in cycles:
- research enough to understand the territory;
- draft;
- discover gaps;
- research the gaps;
- revise;
- fact-check;
- research final uncertainties;
- stabilise the text.
Writing is diagnostic. It reveals what the author still does not understand.
71. Build Research Into the Structure
Research should not appear as a pile of facts poured into a chapter.
Each fact should answer a question, create a problem, support a mechanism, challenge an assumption or make a world concrete.
The structure determines when evidence becomes meaningful.
72. Do Not Show the Reader Everything You Learned
Authors frequently overfill manuscripts because research was expensive.
But the reader does not owe every note a place.
Research creates authority partly through what the author chooses not to display. The invisible research supports confident selection.
73. Facts Need Narrative Placement
An accurate fact introduced at the wrong time can interrupt a story or confuse an explanation.
Move information to the moment the reader wants or needs it.
Research becomes storytelling when evidence is sequenced around curiosity and consequence.
74. The Strongest Detail Is Often the One That Carries More Than One Job
A good researched detail can establish place, reveal character, explain a system and foreshadow consequence at once.
Authors should prefer details that carry structural weight over details that merely demonstrate how much research was done.
75. Footnotes, Endnotes and References Have Different Reader Effects
Different forms handle provenance differently.
Academic work may cite densely. General nonfiction may use endnotes. Journalism may link sources. Essays may name sources in prose. Fiction may acknowledge research separately.
The citation system should support both readability and traceability.
76. Fact-Checking Is the Last Research Pass
Before publication, return to every consequential factual claim.
- Names.
- Dates.
- Places.
- Quotations.
- Numbers.
- Technical terms.
- Chronology.
- Legal or medical assertions.
- Causal statements.
- Attribution.
Research that was sufficient during drafting may need stronger verification before publication.
77. Check the Surrounding Sentence, Not Just the Fact
A sentence can contain a correct statistic but make an unsupported causal claim around it.
Fact-checking must examine the rhetorical use of the evidence, not only whether the number itself is correct.
78. Research Notes Should Survive the Book
Keep enough research organisation that claims can be revisited after publication.
Questions may arise from editors, readers, translators, lawyers or later editions.
A defensible research archive protects the future life of the work.
79. Corrections Are Part of Responsible Authorship
No research process eliminates every error.
When an important error is found, responsible authors and publishers correct it where the medium allows and preserve a transparent record where appropriate.
Credibility comes partly from how mistakes are handled.
80. For Students: Research Is More Than Copying Information
Student research often fails because the task is treated as collection rather than investigation.
- Write one clear question.
- Find three different source types.
- Record where every note came from.
- Identify one point on which sources agree.
- Identify one point on which they differ.
- Write what you still do not know.
- Draft your own explanation before copying more information.
Research becomes learning when students must compare, judge and explain.
81. For Teachers: Teach Source Evaluation as a Habit of Mind
Instead of asking only whether a source is “reliable,” ask students:
- Who produced it?
- For what purpose?
- What evidence does it contain?
- What does it omit?
- How current is it?
- What would confirm or challenge it?
- What kind of claim is it strong enough to support?
This teaches judgment rather than checkbox evaluation.
82. A Thirty-Minute Research Drill
- Choose one question.
- Write what you currently believe.
- Find one source supporting it.
- Find one source challenging it.
- Trace one important claim back to its original source.
- Write what changed in your understanding.
- List the next two questions the research created.
The goal is not information volume. It is epistemic movement.
83. A Research Triage System
When a project becomes overloaded, classify research needs:
- Red: a central claim lacks support.
- Amber: an important example, timeline or mechanism remains incomplete.
- Green: interesting context that may enrich but is not necessary.
Research red first. Green material can wait.
84. Questions Authors Ask During Research
- What exactly am I trying to know?
- What am I assuming?
- What is the strongest source for this claim?
- Who disagrees?
- What evidence would change my mind?
- What definition is being used?
- What is the denominator?
- What changed over time?
- Whose perspective is missing?
- What was not preserved?
- What part is observation?
- What part is interpretation?
- What part is speculation?
- What is the mechanism?
- What are the boundary conditions?
- What information do I have only because one source keeps being repeated?
- What have I learned that changes the structure of the work?
85. Common Research Myths
Myth: More sources automatically mean better research.
Quality, independence and relevance matter more than raw quantity.
Myth: Primary sources tell the truth directly.
Primary sources are direct evidence, but they still have perspective, motive and limits.
Myth: Experts always agree.
Some questions have strong consensus. Others remain contested. The author must learn which is which.
Myth: Once you have enough facts, the book writes itself.
Research provides material. Authorship still requires selection, structure, explanation and judgment.
Myth: Research should confirm the proposal.
The strongest research may overturn the proposal entirely.
86. A Systems View of Authorial Research
Research can be understood as a knowledge-building system:
- Input: curiosity, prior knowledge and a working question.
- Search: source discovery across documents, people, places, data and archives.
- Evaluation: authority, method, provenance, context and bias.
- Comparison: agreement, contradiction, counterexample and uncertainty.
- Synthesis: mechanisms, patterns, timelines, relationships and boundary conditions.
- Writing: evidence is selected and sequenced for a reader.
- Verification: claims are checked before release.
- Return path: the finished work generates new questions, corrections and future research.
The return path matters. Publication does not end inquiry. A good book changes what the author now knows how to ask.
87. Why Research Matters
Research protects writing from becoming a closed loop of the author’s prior beliefs.
It brings the world into the work as resistance.
That resistance can make fiction more credible, nonfiction more defensible, essays more surprising and explanation more precise.
Research does not eliminate imagination. It gives imagination better constraints.
88. The Long Answer
How do authors research?
They begin with a question and then learn how badly the question was formed. They map the field, learn its vocabulary, trace claims backward to stronger evidence and forward into later debate. They compare primary and secondary sources, interview people without confusing memory with fact, visit places where written systems become lived systems, inspect numbers before trusting them, preserve provenance, build timelines and relationship maps, and search deliberately for the evidence that could prove them wrong.
They organise notes around questions rather than collections. They distinguish observation, inference and speculation. They let contradictions remain long enough to become useful. They stop researching when additional material no longer changes the structure of the problem. Then they return during revision and fact-checking, because writing reveals uncertainties that reading alone never exposed.
The goal of research is not to make the author sound informed.
It is to make the work harder for reality to knock down.
Return to How Authors Work, revisit How Authors Find Ideas, continue through How Authors Build Voice and How Authors Revise and Rewrite, or explore the related How Scientific Research Works guide for the research system specific to scientific inquiry.