03 · What You Need to Know
Which Term Variations Can Change What Your Search Finds?
Start With the Reason Variations Matter
A free-text search depends, at least partly, on the language contained in searchable database records. Authors do not always express the same concept identically. One article may use the full name of a concept, another its acronym, and another a regional spelling.
For example, literature concerning randomized controlled trials may contain randomized or randomised. Literature on cognitive behavioral therapy may use the full expression, cognitive behavioural therapy, or the abbreviation CBT. If the database does not automatically connect those forms, searching only one can miss records containing another.
This is one part of the broader task of identifying the different terms researchers use for the same concept. Acronyms and spelling variants deserve particular attention because they may look like minor typographical differences while functioning as separate searchable strings.
Acronyms and Full Terms Can Retrieve Different Records
An acronym is formed from the initial letters or other parts of a longer expression. Research literature is full of them: AI, ADHD, CBT, STEM, MOOC, RCT, and thousands of discipline-specific examples.
When both the acronym and expanded form are commonly used, consider searching both. An abstract may introduce the full expression and then use the acronym, but titles, author keywords, records with limited abstracts, and other bibliographic fields may not contain both forms.
A simplified concept block might therefore use:
The important qualification is that acronyms can be ambiguous. The same short form may represent several unrelated concepts, particularly across disciplines.
Short Acronyms Can Produce a Great Deal of Noise
Consider AI. In contemporary computing and education literature, it commonly represents artificial intelligence. Elsewhere, the same letters may occur as another abbreviation, part of a name, or text that the database interprets differently.
The shorter the acronym, the greater the potential for ambiguity. Two- and three-letter abbreviations deserve particular scrutiny.
Do not include an acronym merely because it exists. Search it separately and inspect what it retrieves. If most results concern unrelated meanings, you may need to combine it with another concept, search it in a more appropriate field, use proximity where supported, or decide that the additional recall does not justify the noise.
Watch Out
An acronym can increase recall while sharply reducing precision. Before adding a short abbreviation with OR, test it independently and inspect whether the database results actually use it for your intended concept.
Abbreviations Are Not Always Acronyms
The practical search issue extends beyond acronyms. Researchers use shortened forms that may not be pronounced as words or formed in exactly the same way. Examples include initialisms, shortened technical names, measurement abbreviations, diagnostic labels, organization names, and methodological shorthand.
For retrieval purposes, the grammatical distinction is usually less important than usage. Ask whether relevant authors use the abbreviated form independently enough that records could be missed if you search only the expanded form.
If the answer is yes, it may deserve a place in the free-text strategy.
British and American Spellings Can Matter
Academic literature is international, and spelling varies. Familiar pairs include:
-
behavior and behaviour;
-
organization and organisation;
-
randomized and randomised;
-
pediatric and paediatric.
For a sensitive search, relevant spelling variants should be considered. Cochrane's current methodological guidance explicitly includes spelling variants among the free-text elements that may need to be identified when developing database search strategies.
Whether you need to type both forms depends on the search platform. Some databases may perform stemming or other lexical processing. Others may treat the forms separately, particularly when exact fields or syntax are used. Do not assume that because two spellings are obviously equivalent to a human reader, the search engine necessarily treats them as equivalent.
A Wildcard Can Sometimes Capture Spelling Variants Efficiently
Some databases provide a wildcard character that can replace one or more characters within a word. This can sometimes capture regional spelling variations without writing separate OR terms.
For example, a database whose syntax supports a single-character wildcard might allow a pattern conceptually similar to:
Never copy wildcard syntax from one database into another without checking its documentation. The same symbol can behave differently, and some platforms restrict where wildcards can appear.
Truncation Can Capture Several Word Endings at Once
Truncation is useful when the beginning of a word remains stable while its ending changes. A common example is a search such as random*, which may retrieve forms such as random, randomized, randomised, randomly, and others, depending on the database.
Cochrane explicitly identifies truncation and wildcards as techniques that can help capture free-text variation. They can make a search considerably more compact.
However, truncation is not simply shorthand for rigor. If the stem is too short, it may retrieve many words that have nothing to do with your concept. The appropriate truncation point therefore needs to be tested.
The mechanics of these operators deserve separate treatment because their symbols and behavior vary substantially among platforms. When you need to construct them precisely, check how phrase searching, truncation, and wildcards work in the database rather than assuming universal syntax.
Singular and Plural Forms May or May Not Need Separate Terms
Suppose you search adolescent. Will the database also retrieve adolescents? There is no database-independent answer.
Search platforms may use automatic term mapping, stemming, lemmatization, or other processing that captures some grammatical variants. Exact searching or particular field syntax may alter that behavior.
Before writing adolescent OR adolescents everywhere, run a small test. Compare the results for each form and inspect the database documentation. If the platform already captures the plural reliably in the way you are searching, adding the second term may contribute nothing.
Hyphenation and Spacing Can Create Additional Forms
Terms may appear with a hyphen, without one, or as separate words. Examples include expressions such as e-learning, elearning, and occasionally e learning, or web-based and web based.
Again, database behavior varies. Some search systems normalize punctuation. Others may process phrases differently depending on quotation marks, field restrictions, or indexing rules.
If the distinction could affect retrieval, test representative variants rather than assuming punctuation is ignored.
Capitalization Usually Deserves Less Attention, but Verify Unusual Cases
Many major academic search systems are not case-sensitive for ordinary keyword searching, so entering separate uppercase and lowercase versions is generally unnecessary. Searching CBT and cbt separately would therefore often add no value.
However, database-specific syntax and specialized search systems can behave differently. As with other search mechanics, consult the platform documentation when capitalization appears consequential rather than treating a general pattern as a universal rule.
Controlled Vocabulary Can Reduce Some Dependence on Textual Variants
Free-text variation is one reason controlled vocabularies exist. If several expressions refer to the same indexed concept, a standardized subject heading can provide another route to the relevant records.
This does not mean subject headings eliminate the need for acronyms and spelling variants. As explained when comparing subject headings with keywords, free-text searching remains important for recent records, emerging terminology, non-indexed content, and concepts not represented adequately by the vocabulary.
For searches where comprehensiveness matters, Cochrane recommends combining appropriate controlled vocabulary with free-text terms and explicitly considering variants such as synonyms, spelling forms, acronyms, truncation, and proximity operators.
Automatic Search Processing Can Make Some Explicit Variants Redundant
Not every database executes exactly the characters you typed. Search systems may automatically map terms, normalize words, identify variants, or apply other transformations.
PubMed is a useful example. Its Automatic Term Mapping can modify or add terms to an untagged query, including synonyms and other variants. The behavior can also change when particular syntax is used.
Consequently, a variant that matters in one search environment may be redundant in another. This is one reason a search strategy should be developed and tested for a specific database and interface.
It also explains why the same search syntax does not work identically across research databases.
Do Not Confuse Linguistic Variation With Conceptual Variation
This distinction is critical.
Behavior and behaviour are spelling variants. Artificial intelligence and AI may be a full term and its abbreviation. Those are primarily linguistic differences.
By contrast, artificial intelligence, machine learning, and generative AI are related concepts, but they are not interchangeable expressions for exactly the same thing. Adding all of them with OR may expand the conceptual scope of the search.
Term variation
A different textual form used to represent substantially the same intended concept, such as an acronym, spelling variant, or alternative word form.
Concept variation
A broader, narrower, overlapping, or neighboring idea that may change what qualifies for retrieval.
Before adding a variation, ask what changed: the spelling, or the meaning? This simple check can prevent a search from expanding beyond the research question.
More Variants Are Not Automatically Better
A comprehensive search should consider meaningful linguistic variation, but exhaustively enumerating every conceivable form can make a strategy difficult to understand, translate, test, and maintain.
Some terms add unique relevant records. Others merely reproduce results already captured by another term or operator. A short acronym may contribute thousands of irrelevant records while adding almost nothing useful.
Test the marginal contribution of uncertain variants. If removing a term makes no meaningful difference to relevant retrieval, you have evidence that it may be unnecessary. Search strategy development is iterative, and Cochrane explicitly describes terms as something that should be modified in light of what has already been retrieved.
