Observation and Surveys

Watching, and Asking

Observational research studies behavior as it happens, without manipulating any variable. Naturalistic observation watches behavior unfold in its real-world setting, with no researcher intervention — observing children's play behavior on an actual playground, for instance — maximizing realism but sacrificing control. Structured (or controlled) observation instead brings behavior into a standardized setting, like a laboratory playroom with fixed toys and a fixed time limit, trading some realism for greater consistency and easier comparison across participants. A recurring threat in both is observer bias: an observer's expectations subtly shaping what they notice or how they interpret it, addressed by training multiple independent observers and checking their inter-rater reliability — the degree to which independent observers agree.

Surveys and questionnaires instead ask people to report on their own thoughts, feelings, or behaviors, often using standardized Likert scales (e.g. "strongly disagree" to "strongly agree" on a 1-5 or 1-7 point scale) for easy quantification across large samples. Surveys are efficient but face their own distinctive biases: social desirability bias is the tendency to answer in a way that looks favorable rather than answer honestly (under-reporting drug use, over-reporting exercise), and wording effects occur when subtle changes in question phrasing shift how people respond, even when the underlying question is meant to be the same.

Common pitfall: assuming a well-designed survey simply captures what people "really" think or do. Self-report always passes through the filter of memory, self-presentation, and the exact wording used — which is exactly why researchers cross-check surveys against behavioral or observational data whenever possible.