November 9, 2026

How to Combine Keywords with Boolean Operators for a Literature Search

You have a topic, a deadline, and a search box that returns either three papers or three thousand. Combining keywords with Boolean operators is how you turn that topic into a search string you can defend, rerun, and hand

Written byWisPaper TeamAI Research Workflow Team
Editorial illustration for How to Combine Keywords with Boolean Operators for a Literature Search

You have a topic, a deadline, and a search box that returns either three papers or three thousand. Combining keywords with Boolean operators is how you turn that topic into a search string you can defend, rerun, and hand to a supervisor. The goal is not a clever one-line query; it is a documented, reproducible search that matches your eligibility criteria and produces a set of records you can screen in stages.

This guide walks through the decisions and artifacts you need, using a January project-start example: a student beginning a thesis chapter on a defined question and needing a defensible starting corpus before the first supervisor meeting.

What state are you actually starting from?

Most readers arrive in one of three states. Diagnose yours before touching operators.

  • Topic-only. You can describe the area in a sentence but have no eligibility criteria. Boolean operators will not help yet, because you do not know what should count as a relevant record.
  • Criteria-but-no-terms. You know what studies you want to include and exclude, but your keyword list is thin or borrowed from a single paper's title.
  • Terms-but-no-structure. You have a long keyword list and no logic connecting them, so results are noisy and you cannot explain why a paper appeared.

Cochrane's guidance is to plan searches from eligibility criteria, combine free-text terms with subject headings, use multiple sources, and document the process. That ordering matters: criteria first, then terms, then Boolean structure. If you skip to operators, you will build a precise string around an imprecise question.

Step 1: Turn your question into eligibility criteria

Before writing a single operator, write down what a record must contain to be included and what would exclude it. Keep this as a short artifact — a two-column list is enough.

For the January example, a thesis chapter on a specific intervention might include records that address the intervention, the population, and the outcome of interest, and exclude records that address only adjacent populations or unrelated outcomes. The criteria do not need to be perfect, but they must exist, because every keyword decision downstream is justified by them.

This is also where rigor begins. NIH guidance treats careful assessment of prior knowledge, knowledge gaps, and transparent methods as core parts of rigorous research practice. Your criteria list is the first transparent artifact of the project.

Checkpoint: You can state, in one sentence each, what makes a record eligible and what makes it ineligible.

Step 2: Build concept blocks, not a flat keyword list

Group your keywords into concept blocks, one block per element of your question. Each block holds synonyms, spelling variants, and related free-text terms. This structure is what Boolean operators will connect.

A practical pattern for the January project:

  • Block A — intervention terms: the intervention name plus common synonyms and abbreviations.
  • Block B — population terms: the population label plus related descriptors.
  • Block C — outcome terms: the outcome plus its variants.

Within a block, synonyms are alternatives (OR). Between blocks, concepts must co-occur (AND). Cochrane's recommendation to combine free-text terms with subject headings means your blocks should also note any controlled vocabulary your target database uses, so the same concept is captured under both its free-text and indexed forms.

Checkpoint: Every keyword sits inside exactly one concept block, and each block has at least two terms.

Step 3: Choose operators that match your intent

Boolean operators encode relationships between terms. Match the operator to the relationship you actually mean.

IntentOperatorUse it whenWatch for
Any synonym is acceptableORTerms are alternatives within one concept blockOver-broad blocks that pull in unrelated fields
Concepts must co-occurANDTwo blocks must both be present in a recordStrings so tight they return almost nothing
A term must be absentNOTYou need to remove a clearly irrelevant categoryRemoving relevant records that mention the excluded term incidentally
Terms must appear togetherPhrase (quotation marks)A multi-word term loses meaning when splitDatabases that handle phrase syntax differently
Variant word endingsTruncation (wildcard)A stem covers plurals and variantsStems that also match unrelated words

Two habits keep this table useful. First, nest your OR blocks in parentheses so the AND between blocks applies to the whole block rather than a single term. Second, treat NOT as a last resort applied at the end, because it silently deletes records you may have wanted.

Checkpoint: You can read your string aloud and explain each operator's job.

Step 4: Assemble and test the string

The assembled string follows the shape: (Block A terms joined by OR) AND (Block B terms joined by OR) AND (Block C terms joined by OR).

Test it in stages. Run each block alone to see how much it returns, then combine blocks pairwise, then run the full string. This staged testing tells you which block is responsible for noise or for silence. If a block returns nothing on its own, the problem is the vocabulary, not the operators.

When you are ready to run the search against a scholarly index, you can find academic papers through WisPaper's scholarly search interface and paste your tested string there. Running the same logic in more than one source is consistent with Cochrane's recommendation to use multiple sources; the operators may need small syntax adjustments per database, but the concept blocks stay the same.

Checkpoint: Each block returns results on its own, and the full string returns a set you can scan without immediate despair.

Step 5: Document the search as a reusable artifact

A search you cannot rerun is a search you cannot defend. Record, at minimum: the date, the sources searched, the full string as entered, any syntax changes per source, and the number of records returned.

This documentation is what lets you rerun searches before you finalize the chapter. Cochrane recommends rerunning searches before publication, and PRISMA 2020 provides reporting templates, including flow-diagram templates, for transparent reporting of search and selection methods. Even for a course paper, the same habit pays off: your methods paragraph writes itself from your log.

Checkpoint: A reader could reproduce your search from your log alone.

Step 6: Save, deduplicate, and screen in stages

Once the string returns a workable set, save the records you intend to screen. In WisPaper you can save or upload papers and organize them in the library so the corpus for this chapter lives in one place rather than across browser tabs.

Then deduplicate and screen in stages, as Cochrane describes: title and abstract first, then full text. Keep a running count at each stage so your flow numbers stay consistent. When multiple reports describe one study, link them rather than counting them as separate studies.

Checkpoint: You have a deduplicated record set, a screening log, and counts at each stage.

Worked example: a January project start

Suppose the thesis chapter question concerns an intervention, a population, and an outcome. The student writes criteria, builds three concept blocks, and assembles a nested string. Testing reveals Block C is too narrow and returns almost nothing; adding two outcome synonyms fixes it. The full string now returns a manageable set.

The student logs the string, runs it in a second source, saves records to a library, deduplicates, and screens titles and abstracts. Before the supervisor meeting, the student reruns the search to catch anything new and updates the log. The deliverable for the meeting is not a perfect review — it is a documented search, a screening log, and a clear statement of what the corpus does and does not cover.

If the supervisor asks how the topic fits the wider literature, that is a different task from running the string. WisPaper's Scholar Agent area supports longer research tasks such as literature review and inspiration work, so it can support a longer literature review task once your search has produced a corpus to reason over.

How do you know the search is ready?

Use explicit completion criteria rather than a feeling that results look right.

  • Your eligibility criteria are written and each keyword traces back to one of them.
  • Every concept block has at least two terms, including any controlled-vocabulary forms.
  • The string is nested so OR blocks are grouped and AND applies between blocks.
  • Each block was tested alone, and the full string was tested after combination.
  • The search is logged with date, sources, string, and counts, and can be rerun.
  • Records are deduplicated and screened in stages with counts recorded.
  • You have rerun the search before finalizing and updated the log.

If any line is missing, the search is not ready — but you also do not need to perfect it before the next step. Fix the missing line, then move on.

A tested, documented string is a milestone, not the end. Two natural next moves depend on what your project needs.

If you are starting a fresh project and want the search and its artifacts to live alongside the rest of the work, you can create a new project in the Research Projects area and keep the chapter's search log and screening notes together.

If you have uploaded papers to your library and want to interrogate them directly — for example, to check how a saved paper defines a term in your concept blocks — you can ask questions based on papers uploaded to My Library. That is useful when a keyword's meaning shifts across the corpus and you need to see the variation before finalizing your blocks.

A reusable checklist

Keep this beside your search log.

  1. Eligibility criteria written (include and exclude).
  2. Concept blocks defined, one per question element.
  3. At least two terms per block, including synonyms and controlled vocabulary.
  4. Operators matched to intent: OR within blocks, AND between blocks, NOT sparingly, phrases and truncation where they help.
  5. Parentheses group each OR block.
  6. Each block tested alone; full string tested after combination.
  7. Search logged: date, sources, string, per-source syntax, counts.
  8. Records saved, deduplicated, screened in stages with counts.
  9. Multiple reports of one study linked.
  10. Search rerun before finalizing; log updated.

FAQs

Use NOT sparingly and only for a clearly irrelevant category, applied at the end of the string. Because NOT removes any record containing the excluded term, it can delete relevant records that mention the term incidentally. If a block is noisy, tightening its vocabulary is usually safer than excluding a term outright.