Surveys use structured questions to collect self-reported data from many participants quickly and inexpensively. Question wording, response format (Likert, forced-choice, open-ended), and order can dramatically affect responses through acquiescence bias, social desirability bias, and question order effects. Standardized questionnaires with established reliability and validity are preferred over ad hoc measures. Surveys are strong on breadth but weak on depth and cannot determine causation.
Critique a poorly designed survey — identify leading questions, double-barreled items, and ambiguous wording — then rewrite it. Test your revised survey on classmates and note unexpected interpretations.
Surveys are the workhorse of descriptive research in psychology because they let you collect self-reported data from many people quickly and cheaply. But the speed and scale that make surveys attractive also make them treacherous: poor question design systematically biases responses, and those biases are invisible in the final dataset unless you know what to look for. Your prerequisite in descriptive research methods gives you the framework; your work on operational definitions tells you why precise question wording matters; and your background in sampling helps you think about who is actually answering your questions.
The most fundamental design challenge is that survey questions are not neutral windows onto mental states — they are stimuli that produce responses. The same underlying attitude can generate dramatically different responses depending on how the question is phrased, what comes before it, and what response options are offered. Leading questions embed assumptions ("How often do you exercise?" presupposes you exercise at all). Double-barreled items ask two questions at once ("I find this class interesting and useful" — what does agreement mean when someone finds it useful but boring?). Acquiescence bias is the tendency to agree with statements regardless of content — a problem for True/False or agree/disagree formats that can be partially addressed by including reverse-keyed items. Social desirability bias inflates reports of admirable behaviors (voting, helping others) and deflates reports of stigmatized ones (drug use, racist attitudes).
Response format shapes what gets measured. Likert-type scales (Strongly Disagree to Strongly Agree) are the most common format for attitude measurement, but the number of response options, whether to include a midpoint, and whether to label all points or only the endpoints all affect responses. Forced-choice formats (choose between two options) avoid acquiescence bias but may feel artificial. Open-ended questions produce richer data but require coding for analysis. No format is universally best — the choice depends on the construct, the population, and whether you need numbers for statistical analysis or narrative for understanding.
Question order effects illustrate how much context shapes response. A question about general happiness asked *after* a question about recent relationship problems will receive lower ratings than the same question asked first — the relationship question primes negative affect and makes it accessible when the happiness question arrives. This is not randomness; it is a predictable psychological mechanism (the availability heuristic) operating on your survey instrument. Randomizing item order within sections is a practical defense against the worst order effects. Your prerequisite in sampling matters here too: even a beautifully designed survey generalizes only to the population your sample represents, and convenience samples (introductory psychology students, online platform workers) limit the conclusions you can draw about broader populations.