A literature review surveys the published research on one question and organizes it into an argument: what is known, where studies disagree and what is still missing. The eight samples below start with the same three studies written twice, as summary and as synthesis. Then come seven finished excerpts from different fields: four structures (thematic, chronological, methodological, theoretical) and three settings (a section of a research paper, a thesis chapter and a standalone review). A short note after each one explains how it is put together.
Read them side by side to choose a structure for your own material; the table after the first sample says when each one fits. Every study cited in the samples is invented, so the author names and years are placeholders, not sources to look up. The theories, databases and reporting guideline the samples mention are real, and the structure is the part to borrow.
| Sample | Field | Structure or setting |
|---|---|---|
| SampleOpen-plan offices | FieldWorkplace psychology | Structure or settingOne paragraph, written as summary and as synthesis |
| SampleCommunity wind farms | FieldEnvironmental policy | Structure or settingThematic |
| SampleEmoji at work | FieldCommunication | Structure or settingChronological |
| SampleMeasuring writing time | FieldEducation research | Structure or settingMethodological |
| SampleFarmers and soil sensors | FieldAgricultural technology | Structure or settingTheoretical |
| SampleBus arrival prediction | FieldComputer science | Structure or settingRelated-work section of a paper |
| SampleLocal news closures | FieldJournalism studies | Structure or settingOpening of a thesis chapter |
| SampleLanguage-learning apps | FieldApplied linguistics | Structure or settingStandalone systematic review |
Synthesis vs summary, side by side
Summary reports what each source says. Synthesis, in the UNC handout's phrase, is "a re-organization, or a reshuffling, of that information": it sets sources side by side and says what they show together. Both versions below use the same three studies.
Haddad (2018) surveyed 640 employees at four firms that had moved into open-plan offices and found that 58% reported more difficulty concentrating than in their previous offices. Brennan and Okoro (2020) tested 96 participants in a laboratory and found that intelligible speech in the background reduced accuracy on a memory task by 14%, while unintelligible noise at the same volume had little effect. Lindgren et al. (2022) interviewed 30 workers in an open-plan insurance office. Many of them wore headphones and booked meeting rooms to find quiet time.
Nothing in it is wrong. But the reader has to work out how the three studies connect, and the paragraph ends where the third study ends.
Complaints about open-plan offices are usually framed as complaints about noise, but the evidence points to a narrower culprit: speech that can be understood. Most of the 640 employees Haddad (2018) surveyed after their firms moved to open plan said they found it harder to concentrate, and laboratory work suggests why. Intelligible conversation cut accuracy on a memory task by 14%, while equally loud unintelligible noise barely mattered (Brennan & Okoro, 2020). Workers seem to know this. The insurance staff interviewed by Lindgren et al. (2022) asked less for quieter offices than for a way out of other people's conversations, using headphones and empty meeting rooms. None of these studies measured whether such workarounds restore concentration or simply move the problem elsewhere.
Why it works. The first sentence is a claim, not an author, and the paragraph sets up a relationship the summary never states: explanation. The lab study is brought in to explain the survey result: what hurts concentration is speech people can follow, not noise as such. The interviews fit that explanation from the workers' side without proving it, which is why the paragraph says workers "seem" to know. The last sentence names what none of the three studies answered, which is where a review earns its place.
Signs of synthesis
- Paragraphs open with a claim in your words
- Several sources in most paragraphs
- Sources grouped by what they show
- Differences between studies are explained
- Each section ends on what is known and what isn't
Signs of summary
- Paragraphs open with an author's name
- One source per paragraph
- Studies appear in the order you read them
- No sentence compares two studies
- The section ends when the sources run out
Parts and structures of a review
A literature review is not a list of everything you read. The UNC Writing Center's handout describes it as a discussion of published work on a subject that adds no new contribution of its own and usually combines summary with synthesis. Most reviews, whatever the field, have three parts:
| Part | What it does | Inside a paper | Thesis chapter or standalone review |
|---|---|---|---|
| PartIntroduction | What it doesNames the topic and the question, sets the scope, says how the review is organized | Inside a paperA sentence or two | Thesis chapter or standalone reviewSeveral paragraphs, often with how the sources were found |
| PartBody | What it doesGroups the sources and synthesizes each group: where they agree, where they differ and why | Inside a paperA few paragraphs | Thesis chapter or standalone reviewSections with subheadings |
| PartConclusion | What it doesSays what the literature adds up to and where it stops | Inside a paperA sentence that leads to your research question | Thesis chapter or standalone reviewA section of its own, ending on the gap or an agenda |
The real decision is how to organize the body, and it is the one the seven samples below are for. Purdue OWL names four common structures, and the UNC handout adds organizing by publication date or by a broader trend. The structures can be combined: the UNC handout points out that a thematic review can still move through time within each theme.
| Structure | Sections follow | Use it when | Watch for |
|---|---|---|---|
| StructureThematic | Sections followRecurring themes or debates | Use it whenStudies cluster around a few questions | Watch forThemes that are only topics, such as “Methods” or “Results” |
| StructureChronological | Sections followPeriods and turning points | Use it whenThe field's questions changed over time | Watch forA list of studies in date order |
| StructureMethodological | Sections followResearch approaches | Use it whenMethods explain why findings differ | Watch forLosing sight of what the studies found |
| StructureTheoretical | Sections followTheories, models and definitions | Use it whenYour study needs a framework | Watch forDescribing theories without comparing them |
Thematic literature review example
A thematic review gives each section to one question the studies keep returning to. This excerpt is the second of three sections in a review of why communities accept or oppose nearby wind farms. The other two cover landscape and visual impact, and local benefits such as community ownership.
2.2 Fair process: who decides matters as much as what is built
A second body of research shifts attention from the turbines to the decisions behind them. Residents who felt they had been consulted early, and that their objections could change the plan, reported higher acceptance than residents near similar projects who were told after the main decisions were made (Achterberg & Silva, 2016; Moreau, 2019). The effect appears even when the outcome is the same. In a survey experiment, Mbatha (2021) gave 1,800 respondents identical project descriptions that differed only in how the site had been chosen, and support rose by 17 percentage points when the description included a public hearing held before the decision.
The studies disagree about whether fair process can offset a poor outcome. Moreau's case studies of three contested projects found that residents who had been consulted still opposed turbines that dominated a valued view, while Ferreira and Kahn (2022), surveying eleven communities, found that consultation softened opposition even where visual impact was high. The difference may lie in what consultation offered. Moreau's residents could comment on a site already chosen; the communities Ferreira and Kahn studied chose among three. Read together, the studies suggest that consultation raises acceptance when it gives residents real influence over the outcome, and that process and visual impact interact rather than simply add up. No study has followed one community from consultation through the first years of operation, so it is unclear whether acceptance won early survives the turbines turning.
Why it works. The heading states the section's claim instead of naming a topic. Two sources support the claim, an experiment shows it holds when the outcome is fixed, and the second paragraph turns a disagreement into an argument by offering an explanation for it. The section ends on a gap, which hands the reader to the next section or to the study the review introduces.
Chronological literature review example
A chronological review traces how a field's questions changed. Purdue OWL warns against "simply listing and summarizing sources in order," and the fix is to define each period by the question it asked, not by its dates.
2. From penalty to politeness: research on emoji at work
Research on emoji in workplace writing has passed through three phases, and each asked a different question. The earliest studies, mostly of email, treated emoticons as a lapse in professionalism and asked how much they cost the sender. Experiments in this period found small but consistent penalties: recruiters rated applicants' follow-up emails as less competent when they ended with a smiley (Park, 2009; Aldana & Reyes, 2011).
As emoji moved into workplace chat, the question became what they meant at all. Interview and survey studies found that colleagues read the same symbol differently, and that it sometimes displayed differently on different devices (Ostrowski, 2016; Nair et al., 2018). Researchers stopped asking whether emoji were professional and began asking whether they were understood.
The most recent work treats emoji as a way of managing tone. Analyses of chat logs from three companies show that employees add emoji most often to requests and corrections, the messages most likely to sound curt in plain text (Hollis & Danjuma, 2022), and that a thumbs-up reaction now often replaces a short written reply (Tanaka, 2024). This later view has not been tested against the earlier one. No recent study has repeated the competence experiments of the email era, so it is unknown whether readers still penalize the emoji their own colleagues use every day.
Why it works. Each paragraph opens with the question of its period, so the order carries an argument: the field moved from judging emoji to decoding them to using them. Transitions mark the turning points. The gap at the end is one only a chronological view reveals, because it comes from setting the newest work beside the oldest.
Methodological literature review example
Organize by method when the methods explain why studies disagree. The excerpt comes from a review of research on how academics spend their working time.
3.1 How studies measure writing time
Estimates of how much time academics spend writing vary more with the method used than with the people studied. Studies that ask faculty to estimate their weekly writing hours in a single survey report the highest figures, typically six to nine hours a week (Grant & Obi, 2015; Delacroix, 2019). Time-diary studies, in which participants log their activities every day for one or two weeks, report about half as much (Marsh, 2017; Yilmaz et al., 2021). Diaries are less exposed to recall errors and to the wish to appear productive, but they cover fewer weeks, and both diary studies sampled the middle of the semester, when teaching crowds out research.
Keystroke logging gives a third picture. Software that records when a document is open and being typed in counted 2.4 hours of active writing a week among 112 doctoral students (Fenwick & Al-Sayed, 2023), less than either of the other methods, partly because it misses reading, planning and writing by hand, which diary keepers count as writing. The three methods measure different things: what writers believe they do, what they recall doing that day, and what their keyboards record. A study comparing all three within one sample would show how far apart they really are. None has been published, and reviews that average across methods (Kim, 2020) combine three different quantities.
Why it works. The opening sentence states the pattern that justifies organizing by method. Each method gets its result and its blind spot, and no method is declared the winner. The closing criticism of an earlier review is specific and fair: it names what the averaging hides.
Theoretical literature review example
A theoretical review compares the theories a field uses and builds the framework for your own study. Here the two theories are real: the technology acceptance model, which Fred Davis set out in 1989, and the diffusion of innovations theory from Everett Rogers's book of that name, first published in 1962. The studies that apply them are invented.
2.3 Why growers adopt, and abandon, soil sensors
Two theories recur in research on farmers' adoption of digital tools. The technology acceptance model (Davis, 1989) explains the intention to use a technology through two beliefs: that it will improve the user's work (perceived usefulness) and that it will take little effort (perceived ease of use). Studies of soil-moisture sensors support the first belief more than the second. In a survey of 410 vegetable growers, perceived usefulness predicted the intention to keep using sensors, while ease of use mattered only for growers over 60 (Okonjo & Weiss, 2019).
Diffusion of innovations theory (Rogers, 1962) looks past the individual to the innovation and the community around it. Of Rogers's five attributes of an innovation, trialability and observability recur most in this literature: growers who could borrow a sensor for a season, or see one working on a neighbor's field, adopted at higher rates (Pereira, 2020; Lund & Achebe, 2022). Each theory brings forward a side the other leaves in the background: the acceptance model, a grower's own beliefs about the tool; the diffusion model, the tool's features and how word of it travels through a community.
Neither, as these studies apply it, explains abandonment, which two surveys report as common: about a third of growers had stopped using their sensors within three years (Okonjo & Weiss, 2019; Haugen, 2023). The studies measured the decision to adopt, not the decision to stop. This study therefore combines the two theories, using the diffusion attributes to explain first adoption and perceived usefulness, measured each season, to explain continued use. It adds maintenance cost, the reason abandoners gave most often (Haugen, 2023), which none of the earlier studies measured.
Why it works. Each theory is stated in a sentence before any study appears, then tested against the evidence. The comparison says what each theory brings and where the studies using them stop, which is the point of a theoretical review. The last paragraph turns that shared gap into the framework the study will use.
Literature review inside a paper
In a journal article the review is short and serves one purpose: to show why your study is needed. In IMRaD papers it is often folded into the introduction. Computer science papers often give it a section of its own, Related Work, organized by approach and closed by a paragraph that positions the paper against those approaches.
II. Related Work
Methods for predicting bus arrival times fall into three groups. Historical-average models predict travel time on each route segment from past trips at the same time of day [1], [2]. They are simple and robust but cannot react to an accident or a closed lane. Filtering approaches, most often based on the Kalman filter, update the prediction as each new GPS position arrives [3], [4] and beat historical averages when traffic departs from its usual pattern. Their accuracy depends on frequent position reports, however, and errors grow on routes where buses report only every minute or two [4]. Machine-learning models trained on large archives of trips report the lowest errors on busy urban routes [5], [6], but each was evaluated on data from a single city, and none reports how much accuracy is lost on another network.
Our method combines the second and third approaches: a learned model supplies the expected travel time for each segment, and a filter corrects it as positions arrive. Unlike [5] and [6], we evaluate it in three cities with different reporting intervals, including one where buses report every 90 seconds.
Why it works. Three approaches, each with a strength and a limit, fit in one paragraph because the section has one job. The limits are chosen to set up the contribution: single-city evaluation and sparse position reports are exactly what the second paragraph addresses. In a psychology or education paper the same moves would sit in the introduction, with author-date citations and the research question in the final sentence. For how a focused question reads, see these research question examples.
Literature review chapter example
In many empirical theses the review is Chapter 2, and it needs its own introduction: the question, the scope, how the sources were found and a map of the sections. This is the opening of a master's thesis chapter.
Chapter 2: Literature Review
When a town's newspaper closes, what happens to its civic life? This chapter reviews research from three fields that answer the question in different ways. Section 2.1 examines studies of news deserts, which document where papers have closed and what, if anything, replaced them. Section 2.2 turns to research that links the loss of local news to lower voter turnout and other measures of civic participation. Section 2.3 reviews the smaller body of work on residents themselves: how they find local information after a closure, and whom they trust.
The review covers work published in English between 2005 and 2025, found through Communication & Mass Media Complete, JSTOR and Google Scholar using combinations of “local news,” “newspaper closure,” “news desert” and “civic engagement,” and through the reference lists of the studies found. It includes peer-reviewed articles, books and foundation reports whose data later studies rely on.
The chapter argues that the three literatures rarely speak to each other. Closure studies count newspapers, participation studies measure outcomes for whole counties, and resident studies sample cities rather than the small towns where closures cluster. As a result, the research can show that closures and lower participation go together, but not how a resident's routine changes when the paper stops arriving. Section 2.4 sets out that gap and the questions it leaves for this thesis.
Why it works. The chapter opens with the question, not with the history of newspapers. The roadmap gives every section a number and a purpose, the search paragraph lets an examiner judge how complete the review is, and the last paragraph gives the chapter a thesis of its own about the literature. To place the chapter in the whole project, see the thesis outline template.
Standalone literature review example
A standalone review is the whole paper, so it has to justify itself: why a new review, and why now. Purdue OWL notes that standalone reviews more often describe how the sources were located and selected. A systematic review goes furthest. The Cochrane Handbook defines it as an attempt to collate all the empirical evidence that fits criteria set in advance, to answer a specific question. Reporting guidelines such as the PRISMA 2020 statement, a 27-item checklist published in 2021, set out what such a review must report.
Introduction (final paragraph)
Language-learning apps reward learners with points, streaks and leaderboards on the theory that game elements keep people practicing, and that practice builds vocabulary. Studies disagree about whether the theory holds once the novelty wears off. Some report large gains in words remembered after a month (Arslan, 2019; Quispe & Moore, 2021); others find that learners chasing streaks favor easy review over new words, which leaves long-term retention no better than with paper flashcards (Duval, 2022). Earlier reviews pooled vocabulary results with grammar and speaking outcomes (Brandt, 2020) or included only laboratory studies (Seo & Varga, 2021). This review asks a narrower question: in studies that followed adult learners for at least four weeks, did gamified apps improve vocabulary retention compared with the same app without game elements or with other self-study methods?
Method (search and selection)
We searched ERIC, PsycINFO, Scopus and Linguistics and Language Behavior Abstracts on March 3, 2026, for studies published from 2012 onward, combining terms for gamification (gamif*, “game element*,” leaderboard*, badge*) with terms for vocabulary learning. We included randomized and quasi-experimental studies with adult participants, an intervention of at least four weeks and a delayed vocabulary test. Two reviewers screened titles and abstracts independently and settled disagreements by discussion. Of 1,284 records, 97 were read in full and 23 met the criteria (Figure 1). The review is reported following PRISMA 2020.
Why it works. The introduction names the disagreement, shows that earlier reviews could not settle it, and ends on a question narrow enough to answer. The method gives a reader everything needed to repeat the search: databases, date, terms, criteria, who screened and how many records survived each step. A course assignment rarely needs this much, but even one sentence on where and how you searched lets a reader judge how complete the review is.
| Setting | Usual length | Introduction | Ends on |
|---|---|---|---|
| SettingSection of a research paper | Usual lengthA few paragraphs | IntroductionA sentence or two, or merged with the paper's introduction | Ends onThe gap and your research question |
| SettingThesis or dissertation chapter | Usual lengthA chapter | IntroductionQuestion, scope, search and a roadmap | Ends onThe gap the thesis fills, leading into the method |
| SettingStandalone review | Usual lengthThe whole article | IntroductionWhy a new review is needed, then its own method | Ends onWhat the field knows and an agenda for research |
Mistakes that turn reviews into lists
- Author-by-author paragraphs. If most paragraphs start with a name and a year, rewrite the first sentence of each as a claim, then bring in the studies as evidence.
- Every study treated as equal. A survey of 30 volunteers and a trial with 3,000 participants should not carry the same weight by default. Say what kind of study each one is when it matters to the argument, as the methodological sample does.
- Too many quotations. The UNC handout advises using quotations sparingly. Paraphrase findings accurately and in your own words, and save quotation for a definition or a phrase that can't be improved.
- Everything you read. Keep only the points that bear on your question. An annotated bibliography is the place for the rest, and its notes become the review's raw material.
- No voice of your own. Begin and end each paragraph with your own sentences, as the UNC handout recommends. The sources support your points; they don't make them for you.
- No gap at the end. "More research is needed" is true of every topic. Name the specific thing no study has done, as every sample above does.
- Tense drift. APA Style recommends the past tense ("Haddad found") or the present perfect ("researchers have argued") for other researchers' work. Pick one pattern and keep it.
In a long draft the author-by-author paragraphs are easy to miss. You can ask the Strokes Assistant to review a section and list every paragraph that discusses only one study; those are the ones to rewrite around a claim.
One more check before you submit: sort your sources by kind. A review built mostly on other reviews and textbooks reads as a review of reviews. The guide to primary and secondary sources explains which is which in each field.
Frequently asked questions
How long should a literature review be?
It depends on the setting. In a journal article it is often a few paragraphs in or after the introduction. In a thesis it is usually a full chapter, and a standalone review is a whole article. Your assignment, your graduate school's handbook or the journal's author guidelines set the actual limit, so check them before you plan.
Is a literature review the same as a systematic review?
No. A systematic review is one kind of literature review. The Cochrane Handbook describes it as collating all the empirical evidence that fits criteria set in advance to answer a specific question, and reporting guidelines such as the 27-item PRISMA 2020 checklist set out what it must include. Course papers usually ask for a narrative review, which selects and interprets the important sources without a fixed search protocol.
What tense should a literature review be written in?
In APA Style, use the past tense or the present perfect for other researchers' work: Haddad (2018) found, researchers have argued. Use the present tense for what the literature as a whole implies. In literature and other humanities fields, writers usually discuss texts and arguments in the present tense. Whichever pattern you choose, keep it consistent.
How many sources should a literature review have?
There is no fixed number. A review inside a course paper may rest on a handful of key studies, while a thesis chapter aims to cover the main work on the question. Ask your instructor or supervisor for their expectation. Coverage of the important studies matters more than the count, and every source should earn its place in the argument.
