What an observation is

1. An observation collects data by watching and recording behaviour, rather than asking people about it.
2. Because behaviour is recorded directly, the data do not depend on what a participant remembers or is willing to admit.
3. Counting how often each behaviour happens turns what is seen into quantitative data.
4. Designing one means making three choices: where the watching happens, whether the participants know, and whether the observer joins in.
5. Because one study takes one option from each of those pairs, the three choices give six types in all: naturalistic, controlled, overt, covert, participant and non-participant.

Naturalistic and controlled observations

1. Naturalistic observation watches behaviour in its normal everyday setting and changes nothing about it.
   1. Strength: the behaviour is the real thing, so ecological validity is high.
   2. Weakness: extraneous variables cannot be controlled, and the setting never repeats exactly, so reliability is lower.
   3. It suits an aim about how people behave in ordinary life, such as how politely pupils queue in the school canteen.
2. Controlled observation watches in a setting the researcher has set up, holding some variables steady.
   1. Strength: the same set-up can be run again with a new group, so reliability is high and groups can be compared fairly.
   2. Weakness: the room is artificial, which lowers ecological validity and invites demand characteristics.
   3. It suits an aim that needs a fair comparison, such as whether children share more when an adult is in the room.

Overt and covert observations

1. Overt observation means the participants know they are being observed, because the observer's role is out in the open.
   1. Strength: informed consent can be gained and the right to withdraw explained.
   2. Weakness: being watched changes how people act, which is the observer effect, and validity drops as a result.
   3. It suits an aim where consent has to be gained in advance, such as observing a lesson with the class's agreement.
2. Covert observation means the observer's role is hidden or disguised, so the participants do not know they are being watched.
   1. Strength: there is no observer effect, so behaviour stays natural and validity is higher.
   2. Weakness: there is no informed consent and privacy is at risk, so it is kept to public places where being seen is expected anyway.
   3. It suits an aim about behaviour people would tidy up if they knew, such as whether shoppers hold a door open for the person behind them.

Participant and non-participant observations

1. Participant observation means the observer joins the group and takes part in whatever it is doing.
   1. Strength: being inside the group reveals meaning and context an outsider would miss.
   2. Weakness: recording while taking part is awkward, and getting close to the group brings observer bias, so objectivity falls.
   3. It suits an aim about what a group's behaviour means to the people in it, such as how a football squad encourages each other.
2. Non-participant observation means the observer stays outside the group and only watches and records.
   1. Strength: tallying can be done as the behaviour happens, so the data are easy to count and compare.
   2. Weakness: from outside the group a behaviour can be misread, so a shove between friends gets logged as aggression.
   3. It suits an aim that needs counts of clearly defined actions.

Behavioural categories

1. Behavioural categories split a broad behaviour into separate actions that can be seen and counted.
2. Each category is defined tightly enough that it is clear what does and does not count.
3. A tally goes on a tally chart every time one of those actions happens.
4. Totalling the columns gives frequencies, which can be turned into a bar chart, a percentage or a modal category.
5. A vague category such as "being naughty" is the usual reason two observers end up with different totals.

Structured and unstructured observations

1. A structured observation records only the categories chosen before the watching starts.
2. Strength: every observer uses the same checklist, so the data are quantitative and easy to compare between groups.
3. Weakness: anything not on the list goes unrecorded, so a behaviour nobody predicted is lost.
4. An unstructured observation records everything that happens, which gives qualitative notes with a great deal of detail.
5. Those notes take judgement to summarise, so two observers can pull different findings from the same session.

Inter-rater reliability and the observer effect

1. Inter-rater reliability is how far two observers watching the same behaviour agree on what they record.
2. To check it, both observers watch the same event, tally separately, and then compare their totals.
3. Close agreement shows the categories are working, and wide disagreement shows they need rewriting.
4. Agreement between two observers shows the recording was consistent, but it does not show that the categories captured the behaviour worth counting.
5. The observer effect threatens validity, because once people know they are being watched they may not behave as they usually would.
6. Watching from an unobtrusive spot, giving people time to settle, and training both observers on the same categories are the standard defences.

Choosing an observation

1. An observation suits an aim about what people do, especially where they would not report it accurately.
2. It suits behaviour that is visible and countable, because a tally chart needs actions rather than thoughts.
3. It cannot supply reasons or feelings, because watching someone does not show why they acted.
4. It cannot establish a cause, because no variable is manipulated.

Turning this into marks

1. When a type and a reason are both asked for, keep them as two separate points, one naming the type and one naming what it does to the behaviour recorded.
2. Give the behavioural categories by name, such as hitting, kicking and pushing, because "aggression" does not secure the knowledge mark.
3. State whether the participants knew they were being watched, since that decides whether the observer effect applies.
4. The mark most often lost is naming observer bias when the answer needed the observer effect.