Observations: Watching Behaviour Directly

An observation is what you use when asking people what they do would ruin the answer. Nobody tells you honestly how often they interrupt in a meeting, but you can sit there and count it. The catch is that you can only ever record what you see — never why.

What an observation can and cannot say

Imagine you watch a child in a playground hit another child. You can write down: hit, once, with an open hand, after being pushed. What you cannot write down is “because he is an aggressive child” or “because he was angry”. You did not see anger. You saw a hand move.

This sounds obvious and it is where most marks are lost. Observations give you behaviour, and behaviour has to be linked to the theory afterwards, carefully, with no assumption of cause and effect.

Three questions that describe any observation

Observations get labelled with two or three words at once, and students find that confusing. It is easier if you treat it as three separate questions. Answer all three and you have named the type.

Naturalistic and controlled

A naturalistic observation happens where the behaviour normally happens, with nothing set up: shoppers choosing between two brands, a crowd at a match, children in a playground. There is no IV at all.

You choose it when running an experiment would wreck the very thing you want to see. You cannot randomly allocate children to “have friends” or “be excluded”, so you watch instead.

A controlled observation puts the behaviour in a setting the researcher designed: a room with a one-way mirror, a fixed set of toys, a script for the adult. Because the situation repeats exactly, another researcher can run it again, which is why controlled observations have far better reliability. The price is that the room is not real life, so ecological validity drops.

Covert and overt

The ethical line usually drawn is this: a covert observation is acceptable only where the behaviour would have been visible to any member of the public anyway — a mall, a station, a stadium. Once you are recording behaviour people believed was private, you have a serious problem.

Participant and non-participant

In a participant observation the researcher joins the group and becomes one of them. This gets you inside conversations, jokes and unspoken rules that an outsider would never see, which is a real gain in validity. Two things then go wrong over time: you only see the part of the group you are standing in, and you start to like the people you are studying. That second one has a name — loss of objectivity — and it is a strong evaluation point.

In a non-participant observation the researcher stays outside, often behind a one-way mirror or at a distance. You keep your objectivity and you usually get a better view of the whole group, but you lose the inside detail, and you cannot ask anyone what just happened.

Behavioural categories: the bit that decides everything

Before a single minute of watching, the researchers agree a list of behaviours and write down exactly what does and does not count. Then they tally them. Vague categories are the reason two observers end up with different numbers.

Inter-observer reliability

One person watching alone might simply be seeing what they hoped to see. So two trained observers record the same behaviour separately, and the two sets of tallies are compared. A strong positive correlation between them means the categories were clear and the recording was consistent.

Worked examples