03 · What You Need to Know
What Does "Enough Search Terms" Actually Mean?
There Is No Correct Number of Search Terms
A concept with stable terminology may require only a few expressions. Another may have several disciplinary names, acronyms, historical labels, regional spellings, technical terms, and controlled-vocabulary representations.
Consequently, rules such as "use five keywords per concept" or "include at least ten synonyms" are not reliable measures of search quality.
The appropriate vocabulary depends on the concept and the literature.
Counting terms
Asks how many words or expressions appear in the search strategy.
Assessing coverage
Asks whether the strategy adequately represents the meaningful ways the intended concept appears in the literature.
The second question is methodologically more useful.
A Search Term Should Have a Job
Every term in a developed strategy should have a plausible retrieval function.
It might:
- represent a common synonym;
- capture an acronym or abbreviation;
- cover a spelling or morphological variant;
- represent older terminology;
- provide access through a controlled-vocabulary term;
- capture terminology used by another discipline;
- recover known relevant records that other terms miss.
If you cannot explain what a candidate term contributes, that is a reason to test it before adding it permanently.
More Terms Usually Increase Opportunities for Retrieval, Not Necessarily Useful Retrieval
Within one concept block, alternatives are commonly connected with OR:
This is why adding synonyms usually increases recall but can reduce precision.
The objective is not maximal expansion. It is useful expansion.
Think About Marginal Contribution
A useful way to evaluate another search term is to ask what it contributes beyond the terms already present.
Suppose your existing concept block retrieves 8,000 records. You add another synonym and the combined block retrieves 8,012.
The additional term contributed at most 12 records to that database search. Those 12 records are its unique or marginal retrieval relative to the existing block.
Now inspect them.
If eight are clearly relevant and otherwise missed, the term may be extremely valuable despite adding very few records. If all 12 are irrelevant, the term contributes no apparent benefit. If they are duplicates in meaning or outside the concept, the term may be unnecessary.
The important number is therefore not how much the total result count changes. It is what useful material the change contains.
A Term That Adds Only One Record Can Still Matter
Do not turn marginal contribution into a numerical rule.
In a systematic review, one additional term that retrieves one otherwise missed eligible study can be important. That study could affect the synthesis or reveal a part of the evidence base that other terminology fails to capture.
Conversely, a term that adds 5,000 records may be poor if nearly all of them concern another concept.
Retrieval volume
How many additional records the term contributes.
Retrieval value
Whether those additional records contain relevant evidence that would otherwise be difficult or impossible to retrieve.
Stopping decisions should emphasize retrieval value.
Test Candidate Terms Against the Existing Concept Block
Suppose your existing block is:
("online learning" OR e-learning OR "online education")
You are considering adding virtual learning.
Rather than simply adding it, compare:
Existing block "online learning" OR e-learning OR "online education"
Candidate term "virtual learning"
Unique contribution Search for records retrieved by the candidate that are not already retrieved by the existing block, where the database's search-history functions allow this comparison.
Inspect Are the unique records about the same intended concept? Are any clearly relevant?
Decide Keep the term if it adds useful access to the concept; reconsider it if the unique contribution is negligible, irrelevant, or conceptually different.
This approach turns vocabulary expansion into an empirical decision rather than a synonym-collecting exercise.
Relevant Articles Should Eventually Stop Revealing Important New Terminology
During early search development, every few relevant articles may expose another useful expression. As the vocabulary improves, genuinely new terminology should become less frequent.
You may inspect additional relevant records and repeatedly encounter terms already represented in your strategy. New expressions may appear, but testing shows that existing terms already retrieve the records containing them.
That pattern is a practical sign that vocabulary coverage is becoming mature.
It should not be interpreted as formal proof that every possible term has been found. Rather, it suggests diminishing returns from continued terminology mining.
Do Not Confuse This With Formal Saturation
Researchers sometimes use the language of saturation to describe the point at which additional searching yields little new terminology or evidence.
The analogy can be useful, but it should be used cautiously. There is no universally accepted quantitative "search-term saturation" threshold at which a bibliographic strategy can be declared complete.
Database coverage, indexing, terminology, and unknown relevant studies prevent that kind of certainty.
It is safer to describe diminishing marginal contribution or stabilization of terminology than to claim that all possible search terms have been exhausted.
Controlled Vocabulary Provides Another Coverage Check
If the database uses a thesaurus or subject-heading system, inspect the concept there before deciding that your vocabulary is complete.
A controlled vocabulary may reveal:
- the preferred indexed term;
- entry terms or synonyms;
- broader concepts;
- narrower concepts;
- related concepts;
- historical indexing information.
You do not need to add every term displayed in the thesaurus as a free-text synonym. The relationships help you check whether an important conceptual route has been overlooked.
Do Not Add Every Entry Term Mechanically
A thesaurus may contain several entry terms leading to one preferred descriptor. Some may be useful free-text expressions; others may be uncommon in the literature you are searching or already captured through another search feature.
Likewise, broader and narrower terms are not automatically synonyms.
Before adding a thesaurus term to an OR group, ask whether authors actually use it for the intended concept and whether it retrieves useful records beyond the existing strategy.
Author Keywords Can Help Confirm Vocabulary Coverage
Inspect author keywords from relevant records.
If the same cluster of terms repeatedly appears and your strategy already represents those expressions or their equivalent retrieval routes, that supports the view that the vocabulary is stabilizing.
If a recurring author keyword is absent from your strategy, test it.
Do not assume that every author keyword belongs in the search. Authors may list secondary concepts, methods, populations, theories, or contextual terms that are not alternative expressions for your concept.
Titles and Abstracts Remain Important Vocabulary Evidence
A candidate term becomes more persuasive when it repeatedly appears in titles or abstracts of clearly relevant studies.
Titles are especially useful for identifying terminology authors consider central. Abstracts provide richer variation and may reveal technical, informal, historical, or discipline-specific language.
The process of finding the different terms researchers use for the same concept should eventually feed into the stopping decision: once additional relevant records mostly repeat terminology already represented, further expansion may have diminishing value.
Check Acronyms, but Do Not Keep Them Automatically
Acronyms can retrieve records that do not contain the expanded expression in the fields you search. They can therefore make important contributions.
They can also be extremely ambiguous.
Test the acronym independently. Examine its unique contribution to the concept block. If it adds relevant records, retain it. If its unique retrieval is dominated by unrelated meanings and it adds no useful evidence, it may not deserve a place in the final strategy.
This is particularly important for short abbreviations, as discussed when deciding whether acronyms and spelling variations should be included.
Spelling Variants May Already Be Captured Automatically
Before adding behavior and behaviour, or singular and plural forms, check how the database processes terms.
Some platforms automatically include spelling or grammatical variants. Others require explicit terms, truncation, or wildcards.
A variation that contributes nothing because the database already handles it automatically does not make the search more comprehensive. It merely makes the string longer.
Search strategy length is a poor surrogate for methodological seriousness.
Truncation Can Make Several Explicit Terms Redundant
Suppose a tested truncation expression safely captures several morphological forms.
Writing every form separately may add no retrieval value:
educate OR educated OR educating OR education
might be replaceable by an appropriate truncated expression in a database where the syntax and resulting word forms have been verified.
The point is not to minimize the number of visible terms. It is to avoid mistaking redundant enumeration for improved coverage.
Historical Terminology Can Be Essential Even When It Adds Few Modern Records
If your search covers a long historical period, older terminology may retrieve records that contemporary terms do not.
A historical term might contribute almost nothing to recent literature while being essential for studies published several decades earlier.
Evaluate its contribution within the period where it was actually used.
This is why concepts that changed names over time require a different kind of vocabulary coverage check. A term should not be dismissed merely because its overall retrieval count is small.
Different Disciplines Can Require Different Terminology
An interdisciplinary search may appear complete within one database while remaining incomplete across another disciplinary literature.
Education, psychology, medicine, computing, sociology, and other fields may use different terminology for overlapping phenomena.
Before stopping, inspect relevant records in the major disciplinary databases included in your search. If a recurring term appears in one literature but not another, test whether it needs to be incorporated into the common free-text strategy or only into the database-specific translation.
You Do Not Need Every Database to Use an Identical Term List
Some terminology is database-specific because of disciplinary coverage or indexing.
A term may be useful in PsycInfo but contribute only noise in a computing database. A subject heading in MEDLINE has no direct function in Scopus. A platform may automatically handle a spelling variation that another requires explicitly.
Consistency should exist at the level of the concept, not necessarily at the level of an identical visible keyword list.
This follows from the broader principle that a search strategy should be translated rather than copied between databases.
Known Relevant Records Are One of Your Best Stopping Checks
If you have a set of clearly relevant articles, test whether the search retrieves them in databases where those records are indexed.
If important known records are missing, investigate why before deciding that the vocabulary is complete.
The missing article may reveal:
- a synonym you omitted;
- a historical term;
- a subject heading you did not use;
- an acronym;
- a field restriction that is too narrow;
- a phrase or proximity condition that is too restrictive.
Once known relevant records are consistently retrievable through defensible routes, that provides some reassurance that further vocabulary expansion may have lower marginal value.
Known-Item Retrieval Does Not Prove Completeness
This safeguard has limits.
If all your known articles come from one research tradition, a search can retrieve every one of them while missing another terminology tradition entirely.
Use known records from different journals, years, disciplines, authors, and terminology where possible.
Think of known-item testing as a way to detect obvious omissions, not as proof that the search has found every possible linguistic route.
Look at What New Terms Retrieve That Existing Terms Do Not
The most informative stopping test is comparative.
For a candidate term T and an existing concept block C, you are interested conceptually in:
This use of set difference is different from using NOT to exclude unwanted research from the final search. You are using it temporarily to audit what one candidate term adds.
Do Not Judge a Candidate Only by the First Page of Results
If a term retrieves thousands of records, its top-ranked results may look highly relevant while its unique contribution to your existing strategy is poor.
Conversely, a rare term may look unimportant in isolation but retrieve several unique relevant studies.
Evaluate the term relative to the concept block rather than judging it solely from its standalone search.
Redundancy Is Not Always Bad
Two terms may retrieve almost the same records. That does not automatically mean one should be removed.
Redundancy can provide robustness. One term may capture a record in a different database, historical period, or field where the other is absent. A spelling variant may be redundant in one platform but necessary in another.
The decision to remove a redundant term should therefore consider the complete multi-database strategy, not just one result count.
But Redundancy Has Costs
Every additional term increases complexity.
A longer strategy takes more time to:
- translate across databases;
- peer review;
- debug;
- update;
- report;
- rerun;
- interpret when something behaves unexpectedly.
Complexity also creates more opportunities for syntax errors, unintended truncation, ambiguous terms, and translation inconsistencies.
A term with no meaningful retrieval contribution is therefore not completely harmless merely because it does not reduce recall.
There Is a Difference Between Harmless Redundancy and Harmful Noise
Some unnecessary terms simply duplicate records already retrieved. Others actively reduce precision by adding large amounts of irrelevant literature.
| Candidate Term Behavior |
Likely Interpretation |
Possible Decision |
| Adds unique relevant records |
Provides useful additional access to the concept |
Usually retain |
| Adds mostly duplicate relevant records |
Potentially redundant but may provide robustness |
Evaluate across databases and fields |
| Adds almost nothing |
Little marginal retrieval value |
Consider omitting unless there is another rationale |
| Adds many irrelevant records and no unique relevant records |
Reduces precision without apparent retrieval benefit |
Usually reconsider or remove |
| Adds a different conceptual literature |
Likely conceptual drift rather than synonym expansion |
Remove from the concept block unless scope genuinely includes it |
| Adds historical or discipline-specific relevant records |
May be important despite low overall volume |
Retain where that coverage matters |
Do Not Keep Adding Terms to Compensate for a Different Search Problem
If relevant records are missing, the problem may not be vocabulary.
A title-only restriction may be too narrow. An exact phrase may exclude legitimate wording. A proximity distance may be too tight. A subject heading may not be exploded. A date filter may remove older research. The database itself may not cover the relevant discipline.
Adding another synonym will not fix those problems.
Before expanding the vocabulary further, diagnose whether the issue is actually a field, syntax, indexing, filter, or database-coverage problem.
Search-Term Expansion Cannot Rescue the Wrong Concept
Suppose the concept block itself is poorly defined. You can add 30 synonyms and still have a poor search.
Likewise, if a candidate term belongs to a broader or neighboring concept, adding more related terminology may gradually transform the question.
The discipline developed in improving a search without changing the question applies directly here. Search-term accumulation should stop before vocabulary expansion becomes conceptual expansion.
Do Not Stop Merely Because the Search Returns a Manageable Number
A result set of 500 records may be comfortable to screen. That does not establish that the vocabulary is adequate.
If a known relevant article is missing because its terminology is absent from the strategy, another search term may still be necessary.
Manageability is a practical consideration, but it should not substitute for retrieval adequacy.
Do Not Continue Merely Because the Search Returns a Large Number
The reverse is also true.
A search returning 10,000 records does not necessarily need more synonyms. If the concept vocabulary is already well represented, additional terms may simply increase screening burden.
When a search already returns thousands of results, diagnose precision problems before expanding the terminology further.
The Purpose of the Search Changes the Stopping Threshold
A quick exploratory search and a systematic review do not require the same level of vocabulary development.
| Search Purpose |
Reasonable Stopping Orientation |
| Quick background search |
Stop when enough high-quality literature has been found to understand the topic adequately |
| Scoping or mapping search |
Develop broad terminology capable of representing the range of the literature and major conceptual variants |
| Systematic review |
Continue until important retrieval routes have been tested carefully and additional terms offer little meaningful gain relative to the need for high sensitivity |
| Known-item or targeted search |
Stop once the specific information need or target literature has been located adequately |
The higher the consequence of missing eligible evidence, the stronger the justification needed for stopping vocabulary development.
For Systematic Reviews, Search Development Is Iterative
Cochrane's current Handbook describes search strategy development as an iterative process. Search terms may be modified after examining records already retrieved, and relevant reports can reveal additional terminology.
This supports an important principle: the initial term list should not be treated as sacred.
At the same time, iteration does not imply endless expansion. Once repeated testing produces little additional relevant retrieval, the strategy can reasonably stabilize.
Peer Review Can Reveal Terms You Missed
For consequential searches, another searcher can identify omissions that are difficult to see after working with the same strategy for hours.
The PRESS guideline for peer review of electronic search strategies includes subject headings and text words among the elements that should be assessed, alongside Boolean and proximity operators, spelling, syntax, and limits.
An information specialist may notice a missing synonym, inappropriate subject heading, overly broad term, or unnecessary redundancy.
If a search is intended to support a high-stakes evidence synthesis, peer review can therefore provide another useful stopping check before the strategy is considered final.
Search Translation Can Reopen the Vocabulary Question
You may think the term list is complete until you translate the strategy into another database.
Then the destination thesaurus reveals a different preferred term, or relevant records use disciplinary language absent from the original database.
That does not mean the first search was necessarily poor. Different databases expose different terminology.
If the new term is genuinely useful across the concept, consider whether it should be propagated back to the other database strategies. If it is specific to one disciplinary context, it may remain database-specific.
Document Why Important Terms Were Added or Removed
For a substantial review, you do not need to publish an autobiography of every synonym considered.
But consequential vocabulary decisions can be worth recording, especially when:
- a broad term was removed because it introduced conceptual drift;
- an acronym was retained because it retrieved unique relevant records;
- historical terminology was added for older literature;
- a spelling variant was omitted because the platform handled it automatically;
- a candidate synonym was tested and contributed no useful unique retrieval.
This creates an audit trail for search development and helps future updates avoid repeating the same experiments.
Eventually, Screening Becomes More Efficient Than Vocabulary Expansion
There is a point at which another hour spent brainstorming synonyms may save almost no screening work and retrieve no meaningful additional evidence.
If the strategy already represents the major terminology, known relevant records are retrievable, controlled-vocabulary routes have been considered, candidate terms repeatedly add little unique relevant material, and the remaining results are plausibly within the conceptual neighborhood of the question, further vocabulary expansion may no longer be the best use of effort.
At that point, screening is doing the job it is supposed to do.
A search strategy does not need to distinguish perfectly between every relevant and irrelevant record. If it could, systematic reviewers would have considerably fewer spreadsheets.
Stopping Is a Judgment Supported by Evidence, Not a Proof of Completeness
No practical database search can demonstrate that no undiscovered useful term exists.
The stopping decision is therefore probabilistic and methodological. You accumulate evidence that the vocabulary is sufficiently mature:
- major terminology sources have been examined;
- relevant records repeatedly use terms already represented;
- controlled-vocabulary concepts have been checked;
- historical and disciplinary terminology has been considered where relevant;
- known relevant records are retrievable;
- new candidate terms contribute little unique relevant retrieval;
- additional terms increasingly add redundancy, ambiguity, or noise.
When those signals converge, continuing indefinitely is unlikely to improve the search proportionately.
