What is the difference between a narrative literature review and a systematic review?

The distinction is often taught as a hierarchy, with systematic reviews at the top and narrative reviews as the thing you do when you cannot manage a real one. That is misleading. They answer different kinds of question and are not substitutes.

A systematic review works when the question is narrow enough to be specified precisely: does this intervention change this outcome in this population. The protocol is fixed before searching, the search is exhaustive within stated bounds, screening is done against explicit criteria and usually by two people, and everything is reported so the process could be repeated. The strength is that the selection of evidence is not a matter of your judgment.

Munn and colleagues set the two apart as a list of dimensions rather than a matter of degree. Their comparison marks the traditional literature review as having no a priori protocol, no PROSPERO registration, no peer-reviewed search strategy, no standardized extraction form, no mandatory risk-of-bias assessment, and no synthesis into summary findings, while the systematic review is marked yes on every one of them (Munn et al., 2018, p. 6). If you cannot answer yes to those six, you have written a narrative review, whatever the chapter is titled.

A narrative review works when the question is about a field rather than an effect: how has this concept been theorized, what assumptions do these traditions share, where is the argument stuck. That work requires judgment about relevance and cannot be reduced to inclusion criteria. Applying a systematic protocol to it would not make it more rigorous, only more expensive.

The same authors are direct about what that judgment costs you. A review resting on the author's prior knowledge and experience is, in their assessment, subjective and not unbiased, exhaustive, or systematic, though it can still give a useful overview of a field (Munn et al., 2018, p. 5). Pautasso qualifies the split differently. Review formats sit on continua rather than in fixed categories, and the form you choose should follow the material, the target journal, the time available, and the number of coauthors (Pautasso, 2013, pp. 2-3). Note also that pooling results statistically is a meta-analysis, which can sit inside a systematic review but is not another name for one.

Most doctoral literature review chapters are narrative, and that is appropriate. The chapter's job is to establish that your research question is worth answering, which is an argument, not a pooled estimate. What a narrative chapter still owes the reader is honesty about how the sources were found, so nobody has to guess whether the coverage is representative or convenient.

When does a scoping review make more sense than either?

When you want to map what exists rather than answer a question. Scoping reviews suit fields where the literature is scattered across disciplines or the concepts are still unsettled, and they report the shape and volume of the evidence rather than pooling a result from it.

They are also the right choice when you suspect a systematic review would be premature. If the primary studies use incompatible measures or definitions, pooling them produces a number that means nothing, and saying so with a scoping review is the more useful contribution.

Scoping reviews follow a protocol and are reported against a checklist, so they are not a softer option. They are a different question with comparable discipline.

Do I need PRISMA for a thesis chapter?

Only if you are presenting the chapter as a systematic or scoping review. PRISMA is a reporting standard for those designs, and its 2020 statement and flow diagram apply there. For a narrative chapter, describing your search honestly is enough and a flow diagram would misrepresent the method.

Some supervisors ask for a flow diagram in a narrative chapter because it looks rigorous. It is worth pushing back gently. A diagram reporting records identified, screened, and excluded implies a protocol-driven process, and presenting a judgment-based selection in that format overstates what you did.

The checklist shows how much reporting the design actually carries. PRISMA 2020 replaced the 2009 statement with 27 items across seven sections, plus item-by-item guidance, a separate abstract checklist, and revised flow diagrams (Page et al., 2021, pp. 1-2). Those items ask for a risk-of-bias assessment for every included study, study-level statistics with their precision, the synthesis results, heterogeneity and sensitivity analyses, the certainty of the evidence, protocol and registration details, competing interests, and where your extracted data and analytic code can be found (Page et al., 2021, p. 8). Almost none of that applies to a chapter whose sources were chosen by judgment, which is the concrete reason a flow diagram misreports a narrative review.

What you can do instead is a short methods paragraph: the databases, the terms, the date range, the date of the last search, and the criteria you applied. That gives the reader the same information without claiming a design you did not use.