Sociology › Section B · Sociological research › Topic B.1
Approaches to sociological research
Anyone can have an opinion about society; sociologists have to back theirs with evidence that others can check. This topic covers the two broad approaches to research, the steps of the scientific method, and the tools that turn a hunch into a testable claim.
From hunch to evidence
Everyone carries theories about society: that crime is rising, that young people read less, that money cannot buy happiness. Some of these beliefs are right, and many are wrong. Sociological research replaces impressions with systematic evidence, gathered by stated methods so that other researchers can check, repeat or challenge the results. No topic is off limits, from the workplace and the family to sport, religion and online life. What makes a study sociological is not its subject but its disciplined approach: a clear question, careful measurement and conclusions that follow from the data.
Two broad approaches
Sociologists inherit two traditions from the discipline's founders. The scientific, or positivist, approach, going back to Comte and Durkheim, models research on the natural sciences: it states hypotheses, measures variables and looks for general patterns and causes. The interpretive approach, going back to Weber's idea of Verstehen, tries to grasp social life as the people living it experience and explain it. Interpretive research is systematic but usually does not test a hypothesis set in advance; researchers may adjust their questions as they learn, and their findings are often descriptive or narrative. Many studies combine both, measuring how common something is and then exploring what it means.
The scientific method in six steps
The scientific approach usually follows six steps. 1. Ask a question narrow enough to study, yet broad enough to matter: "Is society happy?" is too vague, while the habits of one small class are too narrow to tell us much. 2. Review the literature, the existing research on the topic, to learn what is known, find gaps and avoid repeating others' work; every source used must be cited. 3. Form a hypothesis. 4. Design and carry out a study. 5. Analyse the data and draw conclusions. 6. Report the results so others can scrutinise and repeat them. Research is a cycle: results raise new questions, and a hypothesis that is not supported still adds knowledge.
Variables and hypotheses
A variable is a characteristic that can take different values, such as income, education or hours of sleep. A hypothesis predicts a relationship between variables, often in an "if, then" form: if neighbourhood unemployment rises, then property crime will rise. The independent variable is the presumed cause, here unemployment; the dependent variable is the presumed effect, here property crime. Identifying which is which matters, because the direction can run either way: greater self-esteem might improve school grades, or better grades might raise self-esteem. A hypothesis must state the expected relationship, not merely name two topics.
Operational definitions
Abstract concepts must be made measurable. An operational definition states exactly how a concept will be observed and measured. "Social class" might be measured by household income, by occupation or by years of education; "religiosity" by how often someone attends services or by how important they say faith is in their life. Different choices can produce different results, so researchers must state their definitions clearly. If one study counts a C as a "good" grade and another requires a B+, their findings cannot be compared directly. Shared, explicit definitions make studies repeatable.
Reliability and validity
A good measure is both reliable and valid. Reliability is consistency: the measure gives the same results when repeated under the same conditions, or when used by different researchers. Validity is accuracy of meaning: the measure captures the concept it claims to capture. The two are independent. Counting how often people attend religious services can be highly reliable, yet it may not validly measure personal faith, since some people attend for family or social reasons. A measure can be reliable but not valid, as a bathroom scale that always reads two kilograms heavy is; a measure that is not reliable cannot be valid.
Worked example: building a testable hypothesis
A city asks whether its after-school programmes reduce youth crime. 1. Question: do teenagers who attend after-school programmes have fewer police contacts? 2. Hypothesis: if a teenager attends a programme regularly, then they will have fewer arrests. 3. Variables: the independent variable is attendance; the dependent variable is arrests. 4. Operational definitions: "regularly" means at least three afternoons a week over a school year, recorded by sign-in sheets; "arrests" means entries in police records for people aged 13 to 17. 5. Reliability check: sign-in sheets are kept the same way at every site. 6. Validity check: arrests record police activity, not all offending, so the report should say so.
Correlation and causation
A correlation exists when two variables change together. Correlation alone does not prove causation. A third variable may drive both, producing a spurious relationship: the more firefighters sent to a fire, the more damage is done, but the size of the fire explains both. To argue that one variable causes another, researchers need three things: a correlation; the right time order, with the cause coming before the effect; and the elimination of alternative explanations, such as third variables. Much sociological data can show the first two easily but struggle with the third, which is why careful researchers word their conclusions cautiously.
Worked example: family dinners and grades
A survey finds that teenagers who eat dinner with their families most nights have higher grade averages. 1. Variables: dinner frequency is the proposed independent variable, grades the dependent variable. 2. Correlation: present. 3. Time order: unclear from a single survey; struggling students might withdraw from family meals. 4. Third variables: household income or parents' work schedules could affect both dinners and grades. 5. Next steps: compare families of similar income statistically, or follow the same students over several years to see whether changes in dinners come before changes in grades. 6. Conclusion: the survey supports an association, not yet a cause.
Samples and generalisation
Researchers rarely study a whole population, so they study a sample. In a random sample every member of the population has an equal chance of selection, which lets findings be generalised within a margin of error. A biased sample can mislead however large it is. In 1936 the Literary Digest magazine mailed about 10 million straw-poll ballots, received over 2 million back, and predicted that Alf Landon would defeat Franklin Roosevelt. Roosevelt won by a landslide. Its mailing lists leaned towards better-off households, and those who chose to reply differed from those who did not; historians still debate which problem mattered more.
Common mistakes to avoid
First, treating interpretive research as unsystematic: it is rigorous, but it seeks understanding rather than testing a fixed hypothesis. Second, naming two topics without stating their relationship. Third, mixing up independent and dependent variables. Fourth, assuming a reliable measure must be valid. Fifth, reading causation into a correlation without checking time order and third variables. Sixth, trusting a sample because it is large: the Literary Digest had millions of replies and was still wrong.
Key takeaways
Sociological research uses systematic, checkable evidence. Positivist research tests hypotheses and seeks general causes; interpretive research seeks Verstehen, understanding from participants' viewpoint. The scientific method runs from question, literature review and hypothesis to study design, conclusions and reporting. Hypotheses link an independent variable (presumed cause) to a dependent variable (presumed effect). Operational definitions make concepts measurable; reliability is consistency, validity is measuring the right thing. Causation needs correlation, time order and ruling out third variables. Random samples allow generalisation; biased samples mislead.
Flashcards
Flashcard
Research modelled on natural science: hypotheses, measurement, general causes
Test your understanding
Question 1 of 5
Hypothesis: "The more hours students work at paid jobs, the lower their grade averages." What is the independent variable?
Question 2 of 5
A survey item about political interest gives very consistent answers on repeated occasions but actually reflects how much people watch television news. The item is:
Question 3 of 5
Towns with more churches have more bars. The most likely explanation is that:
Question 4 of 5
Why was the 1936 Literary Digest poll wrong despite millions of replies?
Question 5 of 5
Which description best fits the interpretive approach to research?
