Question and explanation — A research question specifies what is to be investigated; a hypothesis proposes a testable relationship or difference where appropriate. Interpretive projects may begin with open questions rather than fixed predictions. Match the design to the kind of claim being made instead of assuming every sociological study follows one identical sequence.
Concepts and operationalisation — Abstract concepts such as social class, religiosity or integration need a justified connection to evidence. Income, occupational class and self-identification measure different dimensions of class; religious affiliation is not identical to attendance or belief. Explain why the chosen indicator addresses the question and what it omits.
Quantitative and qualitative — Quantitative evidence expresses numerical counts or measurements; qualitative evidence expresses meanings, experiences or contextual material. The distinction concerns the information and analysis, not simply whether research occurs inside or outside a laboratory. Open interview answers can be coded numerically, while documents can be analysed interpretively.
Primary and secondary — Primary data are collected for the current investigation; secondary data already exist from another collection or purpose. A researcher can use both in the same design. Secondary material may allow historical comparison or large-scale analysis, but its original definitions and selection may not fit the new question.
Deduction and induction — Deductive reasoning develops implications from a theory for investigation; inductive reasoning develops patterns or ideas from material. Actual research often moves between them. Induction is not the absence of prior concepts, and a disconfirmed prediction may challenge a theory, a measurement or an auxiliary assumption rather than automatically settle every issue.
Cross-sectional and longitudinal — A cross-sectional study examines a population or cases at a particular period; a longitudinal study follows change over time. Panel studies revisit the same participants, while repeated cross-sections sample a population at different times. Longitudinal evidence can establish temporal order but attrition (participants dropping out of the study) and changing measures can complicate interpretation.
Comparisons and causal claims — An association alone does not show whether one factor causes another, whether causation runs in reverse or whether another process influences both. Comparisons can investigate alternative explanations, but matching or statistical adjustment depends on what is measured and assumed. State the actual reach of the design rather than using causal language automatically.
Samples and quality
Population and frame — The target population is the group to which a study seeks to relate its conclusions. A sample contains the cases studied, and a sampling frame lists eligible selection units where one is available. Incomplete frames, institutional refusal and participant nonresponse can distort coverage even before analysis begins.
Random and systematic samples — Random sampling uses a chance procedure to select eligible units; systematic sampling selects at an interval after a suitable starting point. A random sample does not guarantee perfect balance in a small study. A repeating pattern in the frame can distort a systematic sample, so examine how the selection rule interacts with the list.
Stratified and quota samples — Stratified probability sampling divides a population into relevant strata and selects randomly within them. Quota sampling recruits until planned category totals are met without necessarily using random selection within categories. Matching a population's gender proportions alone does not guarantee representation on every other relevant characteristic.
Opportunity and snowball recruitment — Opportunity sampling uses accessible cases; snowball recruitment uses participants or contacts to identify further potential cases. Snowball methods can support access to hard-to-reach networks but may overrepresent connected people or similar experiences. Explain the access advantage and selection limit in relation to the actual population.
Purposive and theoretical sampling — Purposive sampling selects cases for their relevance to a research aim; theoretical sampling selects further cases to develop or challenge emerging explanation. These approaches pursue informative contrasts rather than automatically estimating population prevalence. Explain the selection logic and the justified reach of the findings.
Representativeness and generalisation — Representativeness concerns how a sample reflects relevant population features; generalisation concerns extending findings beyond the studied cases. Statistical population inference differs from transferring an explanation to a comparable context. A large sample can remain biased, while a small contextual study can still generate useful theoretical insight.
Reliability and validity — Reliability concerns consistency of a procedure or measurement under relevant conditions; validity concerns whether the evidence supports the intended interpretation. A repeated but poorly targeted question can be reliable without validly measuring its claimed concept. Specify which aspect of consistency or interpretation is in question rather than treating the words as interchangeable.
Asking and observing
Questionnaire design — Closed questions can support comparison and efficient coding, but response options can force a poor fit. Open questions allow unexpected meanings but require interpretation and can be harder to compare. Avoid leading, ambiguous and double-barrelled items that ask two things at once, such as “Are lessons enjoyable and useful?”, and pilot both understanding and administration with relevant participants.
Questionnaire response — A high response count is useful only relative to the eligible population and recruitment process. Nonresponse, literacy, digital access, question order and willingness to disclose may affect the results. Distinguish the proportion responding from the representativeness or truthfulness of their answers.
Structured interviews — Structured interviews use standardised wording and sequence, supporting comparison and limiting some interviewer variation. An interviewer can clarify logistics but restricted responses may overlook meanings or unexpected issues. Standardisation improves one kind of consistency without guaranteeing that every participant understands a term identically.
Unstructured and semi-structured interviews — Unstructured interviews follow participants' accounts flexibly; semi-structured interviews combine planned themes with opportunities to probe. They can develop contextual depth and clarify meanings, while rapport, interviewer assumptions and interpretation influence the material. Open conversation is not automatically a direct or complete window into experience.
Group discussion — Focus groups and group interviews use interaction among participants as part of the evidence. They can reveal shared language and contested norms, but conformity, dominant voices and privacy concerns may limit disclosure. Group agreement can be a finding about the interaction rather than proof of a general population consensus.
Observation choices — Participant/non-participant describes the researcher's involvement; overt/covert describes whether the research role is disclosed. These dimensions are distinct, so an overt participant observer and an overt non-participant observer are both possible. Choose a role through the question, access, ethics and evidential trade-offs rather than a simple validity ranking.
Recording observation — A structured observation scheme defines events or categories for systematic recording; unstructured field notes preserve richer context and emerging meanings. Categories must be clear, and recording choices should remain transparent. A count can support comparisons while missing meaning; rich notes can reveal meaning while making selective interpretation harder to assess.
Observer effects and immersion — People may change behaviour when aware of observation, while a participating researcher may develop close loyalties or lose critical distance. Immersion can improve understanding without removing the need for reflexivity (examining the researcher's own influence) and corroboration (checking an account against other evidence). Explain the specific route of influence instead of assuming either outsiders or insiders are always superior.
Experiments documents and statistics
Laboratory and field experiments — Experiments deliberately manipulate a condition and compare outcomes. Laboratory settings can support control, while field experiments operate in everyday contexts with other practical and ethical constraints. A field experiment remains an experiment because of manipulation, not simply because the research takes place outdoors.
Natural comparisons — Researchers can investigate differences produced by policies, events or existing circumstances without assigning the exposure themselves. Such comparisons may permit stronger causal reasoning if their assumptions are justified, but affected and comparison groups can differ in other ways. A naturally occurring contrast is not automatically equivalent to random allocation.
Comparative method — Comparing social contexts can identify patterns consistent with an explanation and challenge a proposed universal claim. Historical or cross-national comparisons must consider differences in definitions, recording, institutions and context. Similar outcomes can arise through different mechanisms, so a table of correlations is not the complete explanation.
Official statistics — Official statistics are produced through state or public-body collection systems for specified purposes. They can provide large-scale repeated measures, but categories, legal definitions, reporting and collection practices shape what is recorded. A crime-recording total and a victimisation survey estimate are different measures, not automatically rival counts of the same known truth.
Documents and authorship — Personal diaries, letters, reports, policy texts and digital material can be sources. Evaluate who produced a document, for whom, why, what survives and how it was selected. An authentic document can still present a selective or misleading account, so authenticity and credibility are distinct questions.
Content analysis — Content analysis examines communication through a defined sample, unit of analysis and coding scheme. It can quantify representation or interpret meanings, but the presence of a portrayal does not by itself show audience exposure or effect. Explain the categories, coder agreement and handling of context or ambiguity.
Digital evidence — Online posts and platform records can reveal patterns and interaction, but access, moderation, algorithms, deleted material and unequal platform use shape the sample. Public visibility does not automatically settle consent, privacy or potential harm. Avoid treating accessible users as a random sample of the whole population.
Ethics interpretation and communication
Theoretical and practical choices — Positivist aims emphasise explaining social patterns through systematic measurement and comparison; interpretivist aims emphasise understanding people's meanings. These aims can influence a preferred type of evidence, but researchers also face access, resources, topic sensitivity and institutional constraints. A chosen method does not prove an exclusive philosophical commitment. Justify how the design serves the particular knowledge claim.
Ethical relationships — Informed and continuing choice, privacy, confidentiality, potential harm and unequal power should be considered throughout a study. Ethical review is a continuing responsibility rather than a form completed once. Hypothetical plans do not certify real participant involvement, and suitable oversight is required for authentic research.
Reflexivity — Reflexivity examines how a researcher's position, assumptions, relationships and decisions shape the investigation. It strengthens transparency without automatically removing bias or making every interpretation equally justified. Link a reflective point to its possible effect on access, evidence or conclusion.
Triangulation and methodological pluralism — Combining methods or sources can examine complementary aspects of a question and test interpretations. Shared assumptions, sample origins or recording definitions can produce agreement without independent confirmation. Explain what each source adds, how contradictions would be handled and whether the extra complexity is justified.
Coding and evidence selection — Qualitative coding groups material into meaningful categories; quantitative coding turns defined responses into numerical variables or counts. Keep an audit trail connecting decisions and claims to the underlying evidence, consider discrepant cases and avoid selecting only supportive extracts. Coding is an interpretive decision even when its outputs are numbers.
Reporting a justified claim — State the question, population, selection, procedure and relevant analysis clearly enough for scrutiny. Distinguish observation from interpretation, association from cause and absence of evidence from evidence of no effect. Discuss limitations that matter to the specific conclusion rather than appending a generic list unrelated to the claim.
Test yourself
50 questions · Random sets of 10. These quick checks support revision; practise longer explanations and justified judgements too.
Mr Kidman's Classroom
