Skip to content

Writing · Beginner

Write an introduction that earns the research question

Build the introduction from relevant evidence to a precise aim.

11 min read · Oct 26, 2026

Key points

  • Explain a specific unresolved issue.
  • Synthesize rather than list sources.
  • End with an answerable aim.
Contents

Build the introduction from relevant evidence to a precise aim.

The Introduction is a chain of reasoning that makes the research question worth asking. It explains a relevant problem, what existing evidence can tell us, what remains uncertain and how the present paper addresses that uncertainty. Its task is to prepare the reader to understand the study, not to display everything the author has read.

A strong Introduction gives a novice enough context without making an expert wait through several paragraphs of broad statements. It also makes a promise that the Methods and Results can keep. If the final aim concerns an association, the opening should not promise proof of a causal mechanism.

Understand the four main functions

Start by writing four plain sentences: the problem matters because something specific depends on it; existing evidence establishes something relevant; a particular uncertainty remains; this paper addresses that uncertainty in a defined way. These functions can occupy four paragraphs, several paragraphs or a short combined section. The number of paragraphs matters less than the logic connecting them.

ICMJE's guidance for medical manuscripts emphasises relevant context and a specific objective. Other disciplines may place greater weight on theory, formal definitions or an engineering requirement. Check the conventions for your article type, but retain a clear connection between background and aim. S01

An Introduction for a review also needs a reason for that review. “Many papers have been published” is insufficient. Explain why readers need a synthesis now: definitions may differ, evidence may conflict, a new method may change interpretation, or an existing review may not cover the population or question you address.

Start at the right level of context

Imagine writing about seedling growth. “Plants are essential to all life on Earth” is usually too broad for a small comparison of light schedules. It creates a scale of importance the study cannot address. Begin closer to the actual decision: comparing light schedules requires a defined growth outcome and consistent measurement conditions.

For social science, do not begin with unsupported claims that an entire generation is anxious, distracted or unprepared. Define the relevant phenomenon and population. For engineering, avoid starting with the claim that a technology is transforming every sector. Explain the specific task, constraint or failure that motivates the evaluation.

Ask whether each opening sentence helps the reader understand why your outcome, comparison or theoretical distinction matters. If it merely sounds important, remove it or replace it with a narrower, supported statement.

Turn reading notes into an evidence map

Before drafting, organise the relevant literature by questions rather than author surnames. Useful headings include what is measured, which populations or systems were studied, which designs were used, where findings agree, and what cannot yet be concluded.

A primary study reports an original investigation. A secondary source discusses or synthesises earlier work. A theoretical framework is a set of concepts and proposed relationships used to explain a phenomenon. A research gap is a specific uncertainty the proposed work can meaningfully address. None of these terms implies that a particular paper is automatically reliable.

For each key source, record the exact claim you intend to use and where the source supports it. Distinguish an author's interpretation from the result itself. If a paper measured self reported confidence, do not describe it as measuring competence. If it tested one material at one temperature, do not cite it as proof about all materials under all conditions.

An evidence map also makes disagreement easier to interpret. Two studies may use the same label for different measurements. Alternatively, they may study different populations or use different comparison groups. Explain such differences before declaring that the literature is contradictory.

Identify a gap that your design can resolve

A useful gap includes a reason the uncertainty matters and a link to the present design. Compare these hypothetical statements.

Weak gap More precise gap What the study would need
More studies on plants are needed Reported comparisons use different measurement times, making growth estimates difficult to compare A common timing and measurement rule
Student mentoring has not been fully explored Confidence and practical skill are treated as interchangeable outcomes in the selected literature A synthesis that separates constructs and instruments
Existing controllers are inadequate Existing evaluations do not isolate accuracy under a shared noise condition A justified benchmark with common inputs and conditions

These are examples of reasoning, not claims about the actual state of those fields. A real manuscript must establish each literature statement from sources it has inspected. Do not convert a plausible gap into a factual claim by placing a citation at the end.

An exact combination of variables being unstudied is not enough by itself. Explain what the combination would teach us. Conversely, replication can be valuable without a novelty claim: it may test whether an effect appears in a new population, under a transparent protocol or with more precise measurement.

Build paragraphs that synthesise

A paragraph usually needs one main point, relevant support, an explanation of that support and a connection to the next point. The supporting evidence may include several papers, but the paragraph should still have one identifiable job. Mensh and Kording discuss coherent organisation at the paper and paragraph levels. S201

A list such as “Study A found this, Study B found that, Study C found something else” forces readers to perform the synthesis. Explain the comparison yourself. You might group findings by measurement, design or conditions. When several studies support a statement, make clear whether they converge on the same estimate, merely the same direction, or a broader conceptual point.

A transition explains why the next paragraph follows from the present one. “However” is useful only when there is a real contrast. “Therefore” requires that the previous evidence supports the next claim. Connecting words cannot repair missing logic.

Worked revision: natural science

All details in this example are hypothetical. It illustrates framing, not an actual literature review.

Weak draft:

Plants are very important and light has many effects on them. A lot of researchers have studied plant growth, but more research is necessary. We used a new lighting system and measured plants. The purpose was to prove that our system was better.

Revised teaching draft:

A comparison of light schedules is interpretable only when the growth outcome and observation period are defined. Seedling height, dry mass and survival describe different aspects of plant development; a study measuring one should not silently substitute it for another. Our classroom project focuses on change in seedling height over a fixed observation period.

The practical question is whether two light schedules produce different mean height changes when other recorded growth conditions are held as consistent as the classroom setting permits. The project is a local comparison rather than a claim that either schedule is universally optimal. We therefore compare the schedules using independently assigned growth units and the same height measurement procedure. The primary objective is to estimate the difference in mean height change; any observations about leaf appearance are exploratory unless specified in advance.

The revision defines the outcome, acknowledges scope and connects the objective to the design. A real version still needs sources for biological background, the actual rationale for the schedules and an accurate account of allocation. The phrase “independently assigned growth units” must correspond to reality. If a whole tray receives one treatment, individual seedlings within that tray do not automatically become independent treatment replicates.

Worked revision: social science

This example is fictional and contains no claim about real mentoring studies.

Weak draft:

Mentoring is essential for student success. Previous research shows many benefits. Yet nobody has studied mentoring properly in our university. We investigated whether mentoring makes students better researchers.

Revised teaching draft:

Research confidence concerns how capable students believe they are of completing research tasks. It differs from demonstrated skill and may depend on which tasks a questionnaire asks about. A study of mentoring therefore needs to identify whether its outcome is confidence, task performance or participation in research activity.

In this proposed observational project, we examine mentoring attendance and reported confidence among students beginning their first independent project. Attendance is not randomly assigned, so students who attend more sessions may also differ in motivation, prior experience or available time. The study can describe the association in this setting, but it cannot by itself determine the effect of assigning additional mentoring sessions. Our primary objective is to estimate the relationship between attendance and the specified confidence score, with adjustment for a small set of justified variables recorded before analysis.

The revision avoids presenting local novelty as the whole rationale. It defines the construct and makes the causal limitation part of the logic. Adjustment means representing other variables in the analysis; it does not guarantee removal of every alternative explanation. A real introduction would cite the relevant theory and measurement literature and explain why the selected variables matter.

For a qualitative version, the aim might instead ask how students describe changes in responsibility and help seeking during mentoring. That question requires an interpretive approach rather than a directional numerical hypothesis.

Worked revision: engineering

Hypothetical revised introduction passage:

Controller comparisons can be difficult to interpret when algorithms are evaluated on different layouts or sensor conditions. In the present teaching problem, the question is how two implementations differ when evaluated against a shared set of cases using the same error definition. The outcome is mean positioning error in millimetres, rather than a general judgement of navigation quality.

Accuracy alone cannot establish suitability for deployment. Completion failures, computation time and energy requirements may alter a practical decision even if average error decreases. This example therefore addresses a deliberately limited reporting question: how to estimate and communicate the average paired error difference. A complete engineering study would justify its benchmark conditions and evaluate the additional requirements relevant to the intended application.

This passage is longer than a slogan because it establishes a real comparison logic. It remains honest about being a teaching example. For an actual technical paper, replace the generic comparison with verified prior methods, a precise unresolved limitation and the contribution of the implemented system. Do not claim to introduce an algorithm when you only compared existing implementations.

Write aims, hypotheses and scope separately

An aim states what the work intends to establish or describe. A hypothesis is a testable expectation. An exploratory analysis examines patterns without claiming that the full comparison was specified before seeing the relevant data. These categories should remain visible throughout the manuscript. APA's quantitative standards explicitly distinguish primary, secondary and exploratory components. S211

Write an aim such as “Estimate the mean difference in completion time between the two interfaces under a shared task protocol.” A hypothesis might predict shorter completion time with one interface. A scope statement might restrict inference to the tested devices, users and task. These sentences perform different functions and should not be collapsed into “We investigated performance.”

For review papers, specify the population or system, concepts, evidence types and time or language boundaries where relevant. If you are comparing theoretical explanations, state the basis for comparison. If the review aims to map measures rather than estimate effectiveness, say so explicitly.

Revise by testing the argument

Read only the first and last sentence of each paragraph. They should form a coherent route to the aim. Then check each factual statement against its citation. Finally, compare the final aim with Methods, Results and Discussion. Remove promises the study cannot keep and add context needed to understand the actual outcome.

Practice: A student writes, “Because no study has examined our exact classroom, this project proves that cooperative learning improves achievement.” Identify two unsupported moves. Model answer: A local absence of research does not by itself establish the importance of the question, and a design has not been described that could support the causal claim. A better aim would specify the classroom, learning measure, comparison and design. The introduction should then explain what the local investigation contributes and what it cannot determine.

Use Template 7 to draft the problem, relevant evidence, specific uncertainty and aim as separate components before joining them into prose.

Sources

S01: ICMJE. Preparing a manuscript for submission to a medical journal

S201: Mensh B, Kording K. 2017. Ten simple rules for structuring papers. PLOS Computational Biology 13(9):e1005619.

S211: American Psychological Association. Quantitative Research Design (JARS Quant).

S200: American Psychological Association. Journal Article Reporting Standards (JARS).