The research onion is a six-layer framework developed by Saunders, Lewis and Thornhill that helps you design a coherent, defensible research methodology by working from the outside in: from your underlying philosophy down to the practical data-collection techniques. You “peel” it layer by layer — philosophy, approach to theory, methodological choice, strategy, time horizon, and finally techniques and procedures — and each decision logically constrains the next.
Use the research onion when you are writing the methodology chapter of a dissertation and need to justify why you chose your methods rather than simply describing what you did. It turns a long list of disconnected choices into a single, traceable line of reasoning that examiners can follow and challenge.
What is the research onion?
The research onion is a visual metaphor introduced by Mark Saunders, Philip Lewis and Adrian Thornhill in their textbook Research Methods for Business Students, first published in 2003 and now in its eighth edition. It depicts the research design process as a series of concentric layers, like the rings of an onion. The outermost layer represents the most abstract decision — your philosophical stance — and each successive inward layer becomes more concrete and operational, until you reach the centre: the actual techniques you use to gather and analyse data.
The power of the model is not the picture itself but the discipline it imposes. To reach the data-collection core, you must literally peel away each outer layer first. That sequence forces methodological coherence: your sampling strategy should follow from your research strategy, which should follow from your methodological choice, which should follow from your approach to theory, which should follow from your philosophy. When an examiner asks “why did you run a survey rather than interviews?”, the onion gives you a structured, defensible answer rather than “it seemed easiest”.
Layer 1 — Research philosophy
The outermost layer is your research philosophy: the set of beliefs you hold about the nature of reality (ontology) and what counts as acceptable knowledge (epistemology). It shapes every decision that follows, because it determines what kinds of questions you think are worth asking and what kinds of evidence you trust. For a full treatment, see our dedicated guide to research philosophy.
Saunders presents four main philosophies (the eighth edition adds postmodernism, but these four dominate student dissertations):
- Positivism — reality is objective and measurable; the researcher is detached; knowledge comes from observable, quantifiable facts and testing hypotheses. Typical of natural-science-style and large-scale quantitative work.
- Interpretivism — reality is socially constructed and subjective; meaning is what matters; the researcher engages with participants to understand their lived experience. Typical of qualitative work.
- Realism (critical realism) — a middle position: a real world exists independently of our knowledge of it, but our access to it is filtered through perception and theory, so observations must be interpreted critically.
- Pragmatism — the research question drives everything; if mixing positivist and interpretivist tools answers the question better, do so. This is the natural home of mixed-methods research.
How you choose: ask what you fundamentally believe knowledge is. If you want to measure, test and generalise, you lean positivist. If you want to understand meanings and context in depth, you lean interpretivist. If your question genuinely needs both, pragmatism gives you permission to combine them.
Layer 2 — Approach to theory development
The second layer concerns the relationship between theory and your data — whether you start with theory and test it, or start with data and build theory. There are three options, covered in depth in our guide to inductive and deductive reasoning:
- Deductive — you begin with existing theory, derive hypotheses, and collect data to test (confirm or refute) them. Top-down. Sits comfortably with positivism.
- Inductive — you begin with observations and data, then build patterns and theory up from them. Bottom-up. Sits comfortably with interpretivism.
- Abductive — you move back and forth: surprising data prompts a plausible explanation, which you then test against further data. A pragmatic blend, increasingly recommended for real research where theory and data inform each other iteratively.
How you choose: if there is a strong existing theory you want to test in a new context, go deductive. If the area is under-theorised and you want to generate fresh insight, go inductive. If you expect to refine theory as you go, abductive is honest about that process.
Layer 3 — Methodological choice
The third layer decides whether your study is quantitative, qualitative, or a combination — and how many methods you use. Our comparison of quantitative vs qualitative research unpacks the difference; the onion frames the structural options:
- Mono-method — a single data type and technique (e.g. one questionnaire, analysed statistically; or one set of interviews, analysed thematically).
- Multi-method — more than one technique, but all within the same paradigm (e.g. a survey and structured observation, both quantitative).
- Mixed methods — both quantitative and qualitative data, integrated. Sub-types include mixed-method simple (kept separate) and mixed-method complex (integrated throughout).
How you choose: match the data type to your question. “How many / how often / is there a relationship?” points to quantitative. “Why / how do people experience X?” points to qualitative. Questions with both numeric and experiential dimensions point to mixed methods.
Layer 4 — Research strategy
The fourth layer is your strategy: the coherent plan of action that links your philosophy and approach to concrete data. Saunders lists seven principal strategies. The right one depends on your question, your philosophy and your access to data:
- Experiment — manipulate an independent variable and measure the effect on a dependent variable under controlled conditions. Strong for causal questions; positivist and deductive.
- Survey — collect standardised data from a large sample, usually by questionnaire, to describe or find relationships. Economical and generalisable.
- Case study — an in-depth, multi-source investigation of one or a few cases within their real-life context (associated with Robert Yin). Strong for “how” and “why” questions.
- Ethnography — immersive study of a culture or group in its natural setting over time, through participant observation.
- Grounded theory — systematically build theory from data through iterative coding and constant comparison (Glaser & Strauss).
- Action research — cycles of planning, acting, observing and reflecting with participants to solve a real problem and create change.
- Archival (and documentary) research — analyse existing administrative records, documents and secondary sources as the primary data.
How you choose: let the question lead. Causal → experiment. Patterns across many cases → survey. Deep contextual understanding → case study or ethnography. Building new theory → grounded theory. Improving a practice while studying it → action research. Working with existing records → archival.
“A case study is an empirical inquiry that investigates a contemporary phenomenon in depth and within its real-world context, especially when the boundaries between phenomenon and context may not be clearly evident.” (Source: Yin, 2018)
Layer 5 — Time horizon
The fifth layer is the time horizon: a simple but important choice about whether you capture data at one moment or across time.
- Cross-sectional — a snapshot at a single point in time. Most student dissertations are cross-sectional because of time and resource limits.
- Longitudinal — data collected from the same subjects at several points over time, to study change and development.
How you choose: if your question is about “the state of things now” or relationships at one moment, cross-sectional is right and efficient. If it is genuinely about change, growth or cause-and-effect over time, you need a longitudinal design — but be realistic about whether a dissertation timetable allows it.
Layer 6 — Techniques and procedures (the core)
At the centre of the onion are the techniques and procedures — the operational detail of how you actually collect and analyse data. This is the layer your reader can replicate, so it must be specific. It covers two things:
- Data collection — primary vs secondary data; the instrument (questionnaire, interview guide, observation schedule, experiment protocol); and your sampling. Decide your sampling method here: probability sampling (random, stratified, systematic) for generalisable quantitative work, or non-probability sampling (purposive, snowball, convenience) for qualitative work.
- Data analysis — the analytic technique that matches your data: descriptive and inferential statistics for quantitative data; thematic analysis (Braun & Clarke), content analysis or grounded-theory coding for qualitative data.
How you choose: the outer five layers have already narrowed your options. A positivist, deductive, quantitative, survey, cross-sectional study almost dictates a structured questionnaire, probability sampling and statistical analysis. An interpretivist, inductive, qualitative, case-study design points to semi-structured interviews, purposive sampling and thematic analysis. If the core does not flow naturally from the outer layers, one of your earlier choices is probably wrong.
How to peel the onion for your dissertation
Work strictly from the outside in. Each answer constrains the next, which is exactly the point — it keeps your methodology internally consistent. Follow these steps:
- State your philosophy. Decide whether you are positivist, interpretivist, realist or pragmatist, and justify it against your research question.
- Fix your approach to theory. Choose deductive, inductive or abductive — consistent with your philosophy.
- Set your methodological choice. Quantitative, qualitative or mixed; mono-, multi- or mixed-method.
- Select your strategy. Pick one (or a justified combination) of the seven strategies that fits your question and access.
- Choose your time horizon. Cross-sectional or longitudinal — be realistic about your timetable.
- Specify techniques and procedures. Nail down your instrument, sampling and analysis in replicable detail.
- Check coherence. Read your six choices as a single sentence. If any layer contradicts another, revise — do not paper over it.
If you want a wider menu of strategies and data-collection options to slot into the inner layers, see our overview of the different research methods for a dissertation.
The research onion at a glance
| Layer | Question it answers | Main options |
|---|---|---|
| 1. Research philosophy | What do I believe counts as valid knowledge? | Positivism, interpretivism, realism (critical realism), pragmatism |
| 2. Approach to theory | Do I test theory or build it? | Deductive, inductive, abductive |
| 3. Methodological choice | What type of data, and how many methods? | Mono-method, multi-method, mixed methods (quant / qual) |
| 4. Strategy | What is my overall plan to gather data? | Experiment, survey, case study, ethnography, grounded theory, action research, archival |
| 5. Time horizon | One snapshot or change over time? | Cross-sectional, longitudinal |
| 6. Techniques & procedures | How exactly do I collect and analyse data? | Instrument, sampling, data collection, data analysis |
A full worked example: mapping one study through all six layers
The best way to learn the onion is to see one study peeled completely. Below, a single education-psychology dissertation is traced through every layer, showing how each choice flows from the last.
Layer 1 — Philosophy: Positivism. Anxiety is treated as a measurable construct (via a validated anxiety scale); the researcher stays detached and tests an objective effect.
Layer 2 — Approach: Deductive. Existing theory predicts mindfulness lowers anxiety, so the student derives a testable hypothesis: H₁ — mean post-programme anxiety scores are lower than pre-programme scores.
Layer 3 — Methodological choice: Mono-method quantitative. A single numeric instrument — a 0–40 anxiety scale — is used.
Layer 4 — Strategy: Experiment (a pre-test/post-test design). The mindfulness programme is the manipulated variable; anxiety score is the outcome.
Layer 5 — Time horizon: Longitudinal (short-term). The same 30 students are measured at two points — before and after the four-week programme.
Layer 6 — Techniques & procedures: Probability sampling is impractical, so the student uses convenience sampling of 30 volunteers; the instrument is the validated scale; analysis is a paired-samples t-test.
Worked calculation (the core layer in action): The 30 students’ change scores (pre minus post) have a mean reduction of 4.0 points with a standard deviation of 5.0. The t-statistic is:
t = mean difference ÷ (SD ÷ √n)
t = 4.0 ÷ (5.0 ÷ √30)
√30 = 5.477, so SD ÷ √30 = 5.0 ÷ 5.477 = 0.913 (the standard error)
t = 4.0 ÷ 0.913 = 4.38
With df = n − 1 = 29, the critical value at p = 0.05 (two-tailed) is about 2.045. Because 4.38 > 2.045, the result is statistically significant: the student rejects the null hypothesis and concludes the programme reduced exam anxiety.
Notice how layer 6 (a t-test on convenience-sampled scores) is the inevitable consequence of the five outer layers — that is the onion working as intended.
Strengths and limitations of the research onion
The model is deservedly popular, but it is a teaching scaffold, not a law of nature. Know what it does well and where it stops.
Strengths:
- Forces methodological coherence by linking abstract beliefs to concrete techniques.
- Gives a ready-made structure for the methodology chapter — one heading per layer.
- Makes your choices easy to justify and easy for examiners to follow and challenge.
- Helps you spot contradictions early (e.g. an interpretivist philosophy with a hypothesis test).
Limitations:
- It was written for business and management students, so some strategies fit other disciplines awkwardly.
- Real research is rarely as tidy or one-directional as concentric rings suggest; choices often iterate.
- Students can treat it as a box-ticking exercise, naming a layer without genuinely justifying it.
- It under-represents some traditions (e.g. participatory or arts-based methods) and the abductive, back-and-forth reality of much fieldwork.
Common mistakes to avoid
- Choosing methods first, philosophy last. Picking “a survey” and then reverse-engineering a philosophy to fit produces incoherent designs. Peel from the outside in.
- Mismatched layers. Claiming interpretivism but then testing hypotheses statistically, or claiming positivism but running three unstructured interviews, signals you have not understood the framework.
- Naming without justifying. Writing “this study is positivist” with no reason. Every layer needs a sentence explaining why, tied to your question.
- Over-claiming the time horizon. Calling a one-off survey “longitudinal” because data collection took three weeks. Longitudinal means repeated measures of the same subjects.
- Ignoring feasibility. Designing an ambitious longitudinal ethnography you cannot finish in a dissertation timetable.
Struggling to peel the onion for your methodology chapter?
Our academics help you choose a coherent philosophy, strategy and method — and write a methodology chapter examiners trust.
Putting it all together
The research onion is, at heart, a decision tree dressed as a vegetable. Its discipline — start abstract, end operational, and never let an inner choice contradict an outer one — is what separates a methodology chapter that merely lists methods from one that justifies them. Peel it carefully, write one section per layer, check that all six choices read as a single coherent argument, and your methodology will be both defensible and replicable. That coherence is exactly what examiners reward.
