7.2.1, 7.2.2, 7.2.3, 7.2.4, 7.2.5 - Researching children ethically

Observing child behaviour

An observation becomes useful evidence when researchers specify what counts as behaviour and how it will be recorded.

A participant observer takes part in the activity, such as joining a supervised play session; a non-participant observer watches without joining. Participation can provide contextual understanding but may change the interaction and make systematic recording harder. Non-participation allows greater attention to recording but may miss a child's perspective. These choices are separate from overt observation, in which people know they are observed, and covert observation, in which the observation is concealed. Covert observation can reduce reactivity, but raises serious consent and privacy problems, especially for children. It is not a default solution to those problems.

A behaviour schedule turns an idea such as cooperation into an operational definition. “Offers a toy to another child without taking one in return” is more recordable than “is nice”. Event sampling can tally every defined occurrence in a fixed period; time sampling records behaviour at specified intervals. A tally provides quantitative frequency data. Field notes describing the sequence and context provide qualitative data. Counting offers alone misses whether the recipient wanted the toy, so the method must match the question.

For an illustrative authorised observation, researchers might record toy offers during a ten-minute free-play period. They should use non-overlapping definitions, practise coding and have two observers independently code the same permitted recording. Agreement supports inter-observer reliability; discussing differences can improve the schedule. Agreement does not ensure validity if both observers consistently record the wrong construct. Equal observation time matters because six offers in thirty minutes cannot be directly compared with four in five minutes.

A researcher must also consider the observer's presence, unobserved behaviour and interpretation. A child may perform differently when watched. Looking only at selected play periods cannot establish a permanent attachment type or diagnose autism. Consent, child assent, privacy and a distress-stop rule remain necessary even if the behaviour seems ordinary.

Questionnaires and interviews

Reports from children, parents or carers reveal experiences that direct observation alone cannot show, but questions and sampling shape the answers.

A closed question offers fixed choices or a rating scale, producing readily comparable quantitative data. “How easy was the morning handover, from 1 to 5?” needs clear endpoint meanings. An open question, such as “What helps at handover?”, allows the respondent to choose their own account, giving qualitative detail but requiring careful analysis. An adult's report of a child's experience is evidence of the adult's perspective; it is not identical to the child's own view.

A structured interview uses the same questions in the same order, improving comparability but restricting follow-up. An unstructured interview follows a flexible conversation, allowing unexpected themes but reducing uniformity and making interviewer influence harder to control. A semi-structured interview combines a common question guide with neutral probes such as “Could you explain that further?” It is suitable when researchers need both comparable ratings and contextual accounts.

Sampling also affects conclusions. Recruiting only volunteers at one nursery can omit dissatisfied families who left or families unable to use that nursery. A broader sampling frame and invitations at different times can reduce some imbalance; they do not eliminate non-response. Researchers should record who could participate and who actually did, rather than claim that a convenience sample represents all parents.

Social desirability occurs when respondents give an answer that presents them favourably: a parent may under-report stress to avoid appearing uncaring. Demand characteristics occur when cues about the research purpose influence responses: a child may give the answer they think the researcher wants. Private responding, neutral wording, clear reassurance that there are no right answers, and an interviewer independent of the care provider can reduce these effects. They cannot guarantee complete honesty.

Children's language, memory and attention must be considered. Short age-appropriate questions and accessible response formats help, but repeated suggestive questioning can alter an answer. Asking “How much did you enjoy your lovely nursery?” builds an answer into the question. “What was your time at nursery like?” allows a different view. Pilot questions with appropriate consent and review whether respondents interpret them as intended.

Cultural and developmental comparisons

A group difference is interpretable only when the groups and the measure are genuinely comparable.

Cross-cultural research compares settings with different cultural practices. Using the same Strange Situation sequence provides a standardised procedure, but separation can have different familiarity or meaning across settings. Translation of interview questions must preserve meaning rather than only words; local expertise can identify assumptions a researcher overlooks. Age, family circumstances and care experience should be considered before attributing a difference to culture. A national label is not a complete account of any family's practice.

A cross-sectional design compares different age groups at one time. Comparing four-year-olds and seven-year-olds is quicker than following one cohort, and avoids years of attrition. However, group differences might reflect their different schooling or historical experiences rather than age-related development. It does not show that the younger participants will change into the older group's pattern.

A longitudinal design follows the same participants over time. This can link earlier care experiences with later development and show change within a child. Attrition can distort the sample if families with greater difficulty are more likely to leave, and repeated testing may create practice effects. Longitudinal sequence establishes that one recorded variable preceded another; unmeasured common causes still prevent automatic causal conclusions. Li's care-quality study illustrates these strengths and limits.

A meta-analysis combines quantitative study findings. Comparable procedures, age ranges and outcome definitions matter; overlapping samples should not be counted twice. Van IJzendoorn and Kroonenberg's synthesis found secure attachment modal across aggregated groups, but more variation within countries than between them. This challenges a simple national stereotype and is compatible with a universal attachment system operating through varied experience. It cannot prove universality in unsampled settings or separate nature from nurture by itself. A large combined sample cannot repair a measure that has a different meaning in different settings.

Participation, protection and rights

Children have a right to be heard as well as a right to protection, so ethical research must attend to both.

The United Nations Convention on the Rights of the Child (UNCRC) was adopted in 1989. Article 3 makes children's best interests a primary consideration; Article 12 concerns freely expressing views and giving them due weight with age and maturity; Article 16 concerns privacy; Article 19 concerns protection from harm. These principles guide research decisions, but the treaty itself is not a laboratory consent form. Official Convention text.

Excluding children automatically “for protection” can silence experiences that adults misunderstand. Inviting them without safeguards can expose them to pressure, distress or disclosure. The solution is meaningful participation with suitable support: explain the activity in accessible language, use communication aids where needed, invite questions and offer a choice that can be refused. Guardian permission is generally needed for a child study, alongside the child's informed agreement or assent appropriate to their understanding. A parent's permission does not override a child's refusal or visible distress. Assent should be checked throughout, not only at the start.