SCIF1000 Chap.4 Finding Scientific Information
Finding Scientific Information
A literature search defines the evidence sample available for later synthesis. This chapter breaks a question into concept blocks, combines synonyms with Boolean logic, chooses search systems by coverage and functionality, tests known relevant records and preserves a reproducible search log.
Precision, recall and workload must be balanced transparently rather than hidden behind a final result count.
Start by separating the question into essential concepts. List spelling variants, broader labels, discipline terms and controlled vocabulary for each block. OR broadens within a concept by accepting alternatives, while AND requires different concepts to appear together.
This visible Boolean structure is easier to test than a long natural-language sentence and easier for another person to reproduce.
Database choice is part of the method. Systems vary in subject and date coverage, record types, field controls, export options and query reproducibility. A discovery engine that finds a few useful papers may still be unsuitable as the only evidence source.
Explain what the selected systems cover and which relevant material may remain outside them.
Test the search against known relevant records and inspect the vocabulary used in strong results. When the search is noisy, identify which term creates the noise before narrowing it. When useful studies are absent, examine synonyms, indexing and field limits.
Change one element at a time and preserve the earlier query so the effect of the revision can be understood.
The search log should include the database and platform, date, exact query, fields, filters, result count and reason for each change. Screening also needs explicit eligibility rules and exclusion reasons. A transparent report does not claim perfect coverage.
It lets another searcher inspect the evidence boundary, rerun the process and decide how an update or additional source could change the sample.
What this chapter covers
- 01
Concept blocks and synonyms
- 02
Boolean operators and field choices
- 03
Database coverage and functionality
- 04
Precision, recall and known records
- 05
Transparent and reproducible search logs
Worked example · free
Search for urban shade evidence
- +1Build intervention, outcome and setting blocks with useful synonyms.
- +1Join synonyms with OR and concepts with AND, then select justified fields and databases.
- +1Record platform, date, exact query, fields, limits, result count and reason for revisions.
Key terms
- Concept block
- A group of synonyms or related labels representing one essential idea in a search question.
- Boolean logic
- Rules using operators such as OR and AND to combine alternatives and distinct concepts.
- Precision
- The proportion of retrieved records that are relevant to the search question.
- Recall
- The proportion of all relevant records in the searchable evidence space that the query retrieves.
Finding Scientific Information FAQ
Why is database choice important?
Search systems differ in disciplinary coverage, date range, indexing, field control, export and reproducibility. The same query can therefore produce a different evidence sample and conclusion.
How broad should my first query be?
Begin with concepts essential to relevance and test known useful records. Add defensible terms or limits after diagnosing noise, rather than inserting every detail from the question immediately.
What makes a search reproducible?
Another person needs the database and platform, date, exact syntax, fields, filters, result count and screening boundary. A final count or screenshot alone cannot reproduce the process.
How should backward and forward citation searching be recorded?
Name each seed article, why it was selected, the direction searched, the database or citation index, the search date and the records added for screening. Keep these routes separate from keyword database results so another person can see how the evidence set grew.
Citation searching follows an existing network and can inherit the seed's topic, language, prestige or publication biases; it is not a neutral replacement for a planned query. Use more than one defensible seed where possible, deduplicate the results, and record whether repeated rounds were performed and what stopping rule ended them.
How can screening decisions introduce bias after a good search?
A broad query cannot protect the review if inclusion rules drift while titles, abstracts or full texts are assessed. Write operational criteria before screening, test them on a small shared sample and record a reason for each exclusion at the relevant stage. When an ambiguous case forces a rule change, document the revision and reconsider earlier records affected by it.
If more than one person screens, compare disagreements to identify terms that are being interpreted differently. The goal is not perfect agreement by authority; it is a visible, consistent boundary that connects the research question to the final evidence set.
Assessment move
Choose one question and build a search log from the first draft onward. Test a small set of known relevant records, then change one term, field or limit at a time. Explain whether each revision aims to improve coverage or reduce noise. Practise identifying which details are true inclusion boundaries and which should remain out of the first discovery pass.
End every search report with the most important retrieval boundary and how you checked it.
