How to Do a Literature Search Part 1: Search Keywords | Searching for a Literature Review

A literature search starts long before typing terms into a database: it begins with building a keyword set that matches how researchers actually describe the same idea in different ways. The core workflow laid out here is to (1) check whether a similar question has already been answered by existing systematic reviews, (2) translate the research question into structured components like Population/Intervention/Comparison/Outcomes (PICO) or Setting/Phenomenon/Design/Evaluation/Research type (SPIDER), and then (3) expand each component into a web of alternative keywords—synonyms, related phrases, abbreviations, and word stems—so the eventual search can capture more relevant papers.

The first move is a scoping search: look for prior systematic reviews that are close enough to reveal the field’s key papers, knowledge gaps, and even better search strategies. Even when those reviews don’t perfectly answer the exact question, they can still point to foundational studies and suggest inclusion/exclusion logic and keyword choices worth borrowing. This step also helps clarify what kind of review format fits the discipline. In some areas—computer science is cited as an example—related work sections often resemble a more traditional literature review that aims for an overview. In health and medical sciences, systematic reviews are more common, with explicit objectives, search protocols, inclusion/exclusion criteria, and study-quality evaluation.

A key distinction is that a traditional literature review is typically more open-ended: it may not require explicit objectives, a documented search protocol, formal inclusion/exclusion criteria, or a structured assessment of study quality. That openness can make literature reviews less repeatable and more vulnerable to reviewer bias when summaries and “gaps” reflect personal judgment. Following a systematic process—without necessarily labeling it as a full systematic review—can still help keep the search and selection logic transparent and more reproducible.

Once the research question is set, the transcript recommends converting it into a concise template using PICO or SPIDER. PICO breaks the question into Population, Intervention, Comparison, and Outcomes; SPIDER uses Setting, Phenomenon of Interest, Design, Evaluation, and Research type. The practical payoff is that the template turns a potentially long question into a small set of core search concepts.

From those core concepts, the next step is keyword expansion. Each main term should generate alternatives: synonyms and variant spellings (e.g., “marathon runner” alongside “marathoner” or “runner”), abbreviations, and acronyms. Stemming is also emphasized—reducing words to a shared root so a search can match multiple word forms (e.g., “computation,” “computationally,” and “compute”). The transcript also notes that fields often use different labels for the same idea, such as injury terminology in running research, so capturing those variations matters.

The result is a keyword set built for recall: a small number of main concepts combined with many OR-linked alternatives. The search itself is treated as the next phase and is separated into a follow-up video, after the keyword list is ready.