Conducting a literature review
Introduction to Social Research Methodology
Why review the literature?
Whatever your topic, you are almost certainly not the first to investigate it. Employee turnover, customer satisfaction, engagement, leadership — each of these has decades of published research behind it. Before designing any study of your own, you need to know what others have found on your general area, and often on your specific issue. You will draw on their work to construct your own project, and you will also need to distinguish your work from theirs. Ignoring the literature means reinventing wheels at best, and repeating known mistakes at worst. If forty years of research consistently shows that supervision quality explains more turnover than pay does, a company such as Meridian should design its study — and spend its budget — accordingly.
Figure 1 shows the two things you do with earlier research: you draw on it — what is known becomes the base of your project — and you distinguish your work from it, by saying what you add. Ignoring the literature leads to reinvented wheels and repeated mistakes.
A literature review is a critical examination of existing research relating to the phenomena of interest, and of the relevant theoretical ideas. In the natural sciences, the review typically sets out an agreed body of knowledge on which the researcher then builds. In the social sciences, agreement is rarer: there are usually competing approaches and competing accounts of the same phenomenon. A social-science review therefore always conveys the sense of what is contested — not only what we know so far (what research discoveries have been made), but how we know it (what research approaches have been taken). It is an argument about the state of knowledge, not a list of everything ever written on a subject.
Figure 2 contrasts the two settings: in the natural sciences the review sets out an agreed base to build on; in the social sciences different disciplines answer the same question — why do employees quit? — in different ways, so the review must convey both what is known and how it is known.
The review serves four connected purposes for your project:
| Purpose | What it means |
|---|---|
| Establish context | Understanding how the topic has been studied and how knowledge of it has developed over time |
| Avoid reinventing the wheel | Learning what is already settled, so effort goes where it is genuinely needed |
| Strengthen methodology | Inheriting tested measures, designs, and warnings about approaches that fail |
| Identify trends, debates, and gaps | Locating the gap in knowledge where your own research question belongs |
Figure 3 draws Creswell and Creswell’s image of the review as the bridge between a topic and a researchable question, held up by the four purposes in the table.
Creswell and Creswell treat the literature review as the bridge between a topic and a researchable question: it is by engaging with what is known that a vague interest becomes a precise question worth asking. The review is an early stage of the research process, but it is not self-contained. It feeds every stage that follows — the concepts and theories that frame the study, the research questions sharpened against existing knowledge, the design and measurement instruments borrowed and adapted from previous studies, and the interpretation of findings, which mean little until compared with what others have found. You will keep returning to the literature throughout a project; the review is a living document, not a chapter you finish and forget.
Figure 4 shows the review as an early but not self-contained stage: it feeds the concepts and theories, the research questions, the design and measurement, and the interpretation of findings, and it stays open for the whole project.
One distinction is worth fixing now. A literature review engages with other researchers’ findings and arguments — their published conclusions. Secondary data analysis, by contrast, reuses other researchers’ data: you run your own analysis on datasets that someone else collected. Reviewing tells you what is known; secondary analysis lets you produce new findings without new fieldwork. The two are complementary but different activities, and secondary data — where to find existing datasets and how to judge whether they can answer your question — is the subject of the next session.
Figure 5 places the two activities on another researcher’s study: secondary data analysis starts from their data and runs a new analysis; a literature review starts from their published findings and arguments.
Types of review
Narrative reviews
The traditional form of review is the narrative review: the reviewer reads widely, selects what seems important, and constructs a critical narrative of the field. Selection and interpretation rest on the reviewer’s judgement, and there is no formal protocol governing what is searched or kept. This gives the narrative review real strengths — it is flexible, can follow an argument wherever it leads, and suits broad, theoretical, or loosely defined topics. But it has corresponding weaknesses. It is vulnerable to cherry-picking: consciously or not, reviewers tend to favour sources that fit the view they already hold. And it is hard for a reader to check, because there is no record of what was searched or what was left out.
Figure 6 sets the narrative review’s flexibility against its main weakness. The panel on cherry-picking shows a field of sources, some fitting the reviewer’s view (filled) and some contradicting it (hollow): the dashed loop round the sources actually cited takes in almost only those that fit.
Systematic literature reviews
A systematic literature review is conducted to an explicit, transparent protocol decided before searching begins. Its key features are: a defined research question that the review sets out to answer; a stated search strategy specifying which databases, which search strings, and which date ranges were used; inclusion and exclusion criteria — rules for what counts as relevant, applied consistently to everything found; and documented screening, recording how many sources were found, how many were kept, and why. The aim is replicability: another researcher following your protocol should retrieve substantially the same body of literature. The systematic review originated in medicine, where the stakes of summarising evidence accurately are highest, and is now standard in management research, where the evidence base is large and very mixed in quality.
Figure 7 shows the protocol, fixed before searching begins, as four features, and its aim: two researchers following the same protocol should retrieve substantially the same literature.
Comparing the two
| Narrative review | Systematic review | |
|---|---|---|
| Selection | Reviewer’s judgement | Explicit criteria, set in advance |
| Search | Undocumented, flexible | Documented strategy, stated databases |
| Replicability | Low | High — the protocol can be repeated |
| Bias risk | Cherry-picking | Reduced, though not eliminated |
| Best for | Broad or theoretical terrain | Focused, answerable questions |
For a semester project, a fully systematic review is beyond the available time. The practical advice is to borrow its discipline: state your keywords, name the databases you searched, and write down your criteria for keeping or discarding sources.
Figure 8 repeats the table with an example question for each type — “how should we think about the employment relationship?” suits a narrative review; “does flexible scheduling reduce turnover in retail?” suits a systematic one — and the three habits to borrow for a semester project.
Preparing for the review
Before searching, make three decisions — in writing. First, define your research question or focus: a review of “turnover” will drown you, whereas a review of why frontline retail employees quit will not. Second, establish inclusion and exclusion criteria — for example: peer-reviewed sources published since 2010; empirical studies of retail or service-sector employees; English-language publications. Third, decide on sources and time frames: which databases you will search and how far back you will go. These three decisions constitute your protocol, and they turn searching from wandering into method.
Figure 9 lays out the three decisions as cards; written down together, they are your protocol.
Searching the literature
Where to search
Scholarly literature is indexed in academic databases such as Web of Science, Scopus, JSTOR, ScienceDirect, and EBSCO Business Source, usually accessed through your institution’s library. Google Scholar is free, publicly available, and vast, and is the practical starting point for most student searches. Theses, dissertations, and institutional repositories are also worth checking — they are often the only place where recent, specialised work appears. Two complementary strategies exist for searching this material: term searching, in which you query databases with keywords, and citation searching, in which you follow the references of sources you have already found. A thorough search uses both.
Figure 10 groups the places to search — academic databases, Google Scholar, and theses, dissertations and repositories — and the two strategies, term searching and citation searching.
From concepts to keywords
Begin by listing the key concepts in your research question, and then list synonyms and related terms for each. Different authors name the same phenomenon differently, and a search finds only the words you actually type.
| Concept | Synonyms and related terms |
|---|---|
| Employee turnover | staff turnover, attrition, quit rates, retention, intention to leave |
| Retail workforce | frontline employees, service sector, sales staff, store employees |
| Causes | antecedents, determinants, predictors, drivers |
Keywords usually appear in the abstract of a relevant article. Tying a keyword to the title produces fewer hits, but if the keyword is in the title, the piece is very likely to be relevant.
Figure 11 shows each concept with its synonyms underneath. Note the third column: academic authors write “antecedents of turnover intention”, not “causes of turnover”. The bottom row compares searching in the abstract (a broad first pass) with searching in the title (a precision pass).
Boolean operators
Databases understand a small logical vocabulary, and three operators do most of the work:
| Operator | Effect | Example |
|---|---|---|
| AND | Narrows — both terms must appear | turnover AND retail |
| OR | Broadens — either term may appear; used to bundle synonyms | turnover OR attrition OR "quit rates" |
NOT (- in Google Scholar) |
Excludes a term | turnover -financial removes asset-turnover papers from finance |
| Quotation marks | Search an exact phrase | "employee turnover" avoids hits where the words appear separately |
Operators can be combined with parentheses: ("employee turnover" OR attrition) AND retail.
Figure 12 draws each operator as a Venn diagram: the shaded area is what the search returns. AND keeps only the overlap; OR keeps both circles; NOT removes one circle from the other. In the parentheses example, the shaded area is the part of either name for the phenomenon that lies inside retail.
Building and refining a search string
Start from your concepts, bundle the synonyms for each concept with OR, and join the concepts with AND.
Figure 13 shows the structure a search string aims at: one block per concept, synonyms joined by OR inside each block, and the blocks joined by AND.
No first search string is right; the skill lies in refinement. For Meridian’s question — why retail employees quit — the iteration might look like this:
| Step | Search string | Result |
|---|---|---|
| 1 | employee turnover |
Millions of hits — hopelessly broad |
| 2 | "employee turnover" AND retail |
Better, but swamped by US supermarket studies |
| 3 | ("employee turnover" OR "staff attrition") AND retail AND (causes OR antecedents) |
Focused on explanations rather than descriptions |
In Figure 14, each attempt notes what changed from the one before and gives a verdict on its results.
Treat searching as iteration: run the search, inspect the first pages of results, adjust, and run again. If there are too many results, narrow — add an AND concept, restrict keywords to titles, or limit the date range. If there are too few, or they are off-topic, broaden — add OR synonyms, drop the most restrictive term, or extend the dates. Use the database’s filters (publication date, peer-reviewed only, subject area, language). And record every search — the string, the database, the date, and the number of hits. This record is what makes your review checkable, and it saves you repeating work you have already done.
Figure 15 shows the loop — run, inspect, adjust, run again — the adjustments for too many or too few results, the database filters, and the search log with its four columns.
Citation searching
Every good source you find is a doorway to others, in two directions. Searching backwards (often called snowballing) means reading the source’s reference list: the studies it builds on are probably relevant to you too. Searching forwards means using the cited by function in Google Scholar to find the studies that later built on it; forward searching also reveals impact, since heavily cited work has demonstrably shaped the field. A handful of good recent articles, combined with citation searching in both directions, can map a literature faster than keyword searching alone. Review articles, and the editorial introductions to edited collections and special issues of journals, are particularly valuable shortcuts: someone expert has already surveyed the terrain, summarised the debates, and gathered the key references for you.
Figure 16 places a good source on a time line: searching backwards follows its references to older work; searching forwards follows “cited by” to newer work that built on it.
Evaluating and reading sources
The hierarchy of credibility
A search returns a mixture of material, and a literature review must weigh its sources, not merely collect them. For academic purposes, a rough hierarchy of credibility runs as follows:
| Rank | Source type | Quality control |
|---|---|---|
| 1 | Peer-reviewed journal articles | Vetted by independent experts before publication |
| 2 | Scholarly books and chapters | Reviewed by editors and academic publishers, less formally |
| 3 | Grey literature (consultancy, industry, government reports) | No academic review; methods often undisclosed |
| 4 | Quality press | Journalistic standards, but written for news value |
| 5 | Blogs, corporate content, Wikipedia | No systematic quality control |
The hierarchy is a starting point, not a verdict: weak journal articles exist, and so do excellent reports. But the burden of proof shifts as you descend the table — the lower the source, the more scrutiny it needs before you rely on it.
Figure 17 draws the five tiers with the arrow of scrutiny beside them: the further down, the heavier the burden of proof. Wikipedia is for orienting yourself and mining references, never for citing.
What peer review does — and does not — guarantee
Before a journal accepts an article, independent experts scrutinise its theory, methods, and conclusions; weak work is filtered out or sent back for revision. Peer review is the reason journal articles anchor an academic literature review. But it is a quality floor, not a truth guarantee. Reviewers cannot re-run the study, so errors and occasionally outright fraud pass through. Journals vary enormously in rigour — at the bottom of the market sit predatory journals that will publish almost anything for a fee while claiming to be peer-reviewed. And findings can be superseded: always check whether later work confirmed or challenged a study before leaning on it. Citation counts are a useful supplementary signal — influence is evidence of scrutiny — but they must be read with care, since recent work has not had time to accumulate citations, and some work is heavily cited only in order to be criticised.
Figure 18 shows the review process, the three reasons it is a quality floor rather than a truth guarantee, and how to read citation counts: they measure attention, not agreement.
Grey literature in management research
In management, much of the most current evidence appears outside academic journals: consultancy studies from firms such as McKinsey or Deloitte, industry-association reports, and government statistics. This grey literature matters — it is current, practically oriented, and often built on proprietary data that academics cannot access. But it carries dangers. It is not peer-reviewed, and its methods are frequently undisclosed. It may serve a commercial interest: a consultancy report on turnover may exist chiefly to sell the consultancy’s retention services. Samples and definitions are sometimes chosen to flatter the headline finding. The working rule: use grey literature for context and currency, but rest your arguments on academically reviewed work. Of every non-academic source, ask: who produced this, for whom, and what do they gain from its conclusions?
Figure 19 sets the value of grey literature against its dangers and draws the working rule: grey literature supplies context and currency; academically reviewed work carries the argument.
Questions to ask of any source
For each publication you are considering keeping, ask a standard set of questions: What question or problem is the author addressing? What are the key concepts, and how are they defined? What theories, models, and methods are used — does the research apply established frameworks or take an innovative approach? What are the results and conclusions? How does the publication relate to the rest of the field — does it confirm, add to, or challenge established knowledge? And what are its strengths and weaknesses? Babbie’s rule of thumb applies throughout: read as a sceptical consumer of research — every claim invites the question how do they know that?
Figure 20 sets out the six questions as a checklist, with Babbie’s question behind them all.
Reading efficiently: abstract-first triage
You cannot read everything in full, and you should not try. Triage each candidate source in stages, discarding at the first failure. Stage one: read the title and abstract — is the source actually about your question? This takes a minute. Stage two: read the introduction and conclusion — what was found, and does it matter for your purposes? This takes perhaps five minutes. Stage three: read the full text closely, taking structured notes — but only for the survivors of the first two stages. Most sources exit at stage one, and that is the system working, not a shortcut. A Google Scholar search on retail turnover might return 400 plausible-looking hits; triage reduces them to perhaps 25 worth reading properly. For those survivors, record notes under standard headings — question, method, sample, findings, limitations, and relevance to your question — so that the material is comparable when you come to write.
Figure 21 draws triage as a funnel: 400 plausible hits enter at the top, most leave at the first stage, and perhaps 25 come out at the bottom to be read in full and noted under standard headings.
Organising, synthesising and citing
Organising what you find
As sources accumulate, the task shifts from finding to connecting. Look for: trends and patterns — do approaches or findings become more or less prominent over time?; themes — which questions and concepts recur across the literature?; debates and contradictions — where do sources disagree, and why?; pivotal publications — which influential theories or studies changed the direction of the field?; and gaps — what is missing, under-studied, or methodologically weak? Contradictions between studies are not a nuisance to be smoothed over. Fields develop unevenly, so assess the methods and sample sizes of conflicting studies, and treat unresolved conflicts as candidate gaps. The gap you identify is the justification for your own research question.
Figure 22 draws a literature as a map of sources from earlier to later and marks the five kinds of connection on it: a trend, two themes, a contradiction, a pivotal publication, and a gap that points to your research question.
Structuring the review
Choose an organising principle before you write. Four standard structures exist:
| Structure | Organised by | Best when |
|---|---|---|
| Chronological | The development of the field over time | The field has clear turning points and key debates |
| Thematic | Topics and sub-questions | Distinct aspects of the topic need separate treatment |
| Methodological | Research approach | Sources from different fields or methods yield different conclusions |
| Theoretical | Competing frameworks | Rival theories, models, and definitions explain the same phenomenon |
Thematic organisation is the most common choice for management topics — a turnover review, for example, might be organised into sections on pay, supervision quality, engagement, and labour-market conditions.
Figure 23 gives each structure a small schematic. Whatever you choose, never organise the review source by source — one paragraph per article is not a structure.
Synthesis over summary
The cardinal sin of literature reviews is the sequence of summaries: “Smith found X. Jones found Y. Brown found Z.” A good review does not simply recount what other people have said; it synthesises. The pattern to follow: establish the themes that emerge from the literature; discuss those themes, especially conflict and difference in how they are understood; and draw everything together to establish the agreed state of knowledge on the topic — if there is one, which in the social sciences there often is not — and the nature of the disagreements and arguments. Organise paragraphs around ideas, not authors, with several sources per paragraph connected by comparison and contrast. From the positions available in the literature, take a position yourself; sometimes your research will consist precisely in comparing positions in order to judge between them. The review should conclude with the specification of your research questions and an account of how they relate to the existing literature — the gap analysis that motivates your study.
Figure 24 contrasts a sequence of summaries, organised by author, with the five moves of a synthesis, and gives one sentence that shows what synthesis sounds like.
Citing and avoiding plagiarism
Record full citation details — author, year, title, outlet, page numbers — the moment you find a source; reconstructing them weeks later wastes hours. Cite every source you use, not only those from which you quote directly: using someone’s ideas or findings without crediting the original source is plagiarism, exactly as copying their words is. Paraphrase properly — restating an idea in your own words and citing it; changing three words in someone else’s sentence is not paraphrase. You need not cite sources you read during the research but did not use in the write-up. Use a reference manager: Zotero is free, works on all platforms, and formats citations automatically. The standard referencing style in this programme is APA. The detailed mechanics of referencing styles come later in your degree — this semester, the habit of recording and crediting sources matters more than the formatting.
Figure 25 shows the workflow from finding a source to citing it, and answers the question “does it need a citation?” case by case.
Conclusion
A literature review is critical engagement with what is already known — what we know and how we know it — not an annotated reading list. The systematic review’s discipline of a stated question, a documented search, and explicit inclusion criteria is a standard worth borrowing even for modest student projects. Searching is a craft of iteration — concepts, keywords, Boolean strings, refinement — with every step recorded so that the search can be checked and extended. Sources must be weighed rather than counted: peer review anchors the hierarchy of credibility, while grey literature adds currency and managerial context but should never carry the argument alone. Read by triage, organise by theme, write by synthesis, and cite everything you use. With the review complete, a research team knows what the field already says about its problem; the next question — taken up in the following session — is whether existing data can be reused before any new data is collected.