Hello everyone, welcome to this research methods tutorial about observations. You should know this by now, but as a minimum, you just need to take notes from the slides which have the pause button. Okay, that's essential. You can read the rest yourself. So, in this tutorial, we're going to be looking at covert and overt observation.
We'll take a look at observation sampling methods, which are time sampling and event sampling. We'll also look at the different types of observation which we can use, so naturalistic, controlled and participant observation. We'll talk about how we assess the reliability of observations. And lastly, we'll look at how we can improve the reliability of observations. First, let's look at the differences between covert and overt observations.
Observations just involve watching people, so we can either hide while we're watching them so that they don't know they're being observed, Or we can just tell them that we're watching them and be open about it. So think of the word covert, sounds a little bit like covered or concealed, and overt, think of open. One of the benefits of using a covert observation is that it's very high in validity. Because people don't know that they're being watched, they won't change anything in their behaviour. So the observed behaviour will be very, very natural.
However... Covert observation is a little bit like spying on people. It's quite unethical because the participants haven't given consent. They don't know that they're taking part in an observation study.
We can reverse these for overt observations though. They're more ethical than covert because people are able to give consent to be observed. However, they're low in validity because people know they're being watched so they might change their behaviour or act differently. Observations can generate a lot of data, so we need to find a way of narrowing down the amount of data that we've got from our study.
If you think about watching children playing in a playground, you want to see how aggressive they are. There's far too much behaviour going on there for you to keep a record of every single event that a child carries out. So we use sampling methods in observation studies just to limit the amount of observations that we have and to take a small sample of the overall behaviour. And there are two ways that we can do that.
One is called event sampling. So what we would do here is keep a count each time a specified behaviour occurs. So we might have a checklist of behaviours that we're looking for.
In the aggression example, we could have kicking, punching, biting on our checklist. Every time we see that child carrying out one of those behaviours, we keep a tally of that. OK, so that's event sampling.
This is really good if we're only recording occasional behaviours. so things that don't happen a lot of the time hopefully but if there's a lot going on it's easy to miss behaviors if there's lots happening at once so say there's a fight and a child is doing all of those things all at once it's difficult to record all of those behaviors we might miss out on important things there the other type of sampling method that we can use is called time sampling So this is where we only record observations at specific time intervals. So we might make an observation every 30 seconds, and we'll only record what the child is doing at the point of 30 seconds. And then we'll count to another 30 seconds and record what the child is doing then.
So this is quite helpful in reducing the number of observations in a set time period. It gives us less data to work with. However, the observed behaviour might not be representative.
if it occurs in between the recording times. So if we're watching for aggression in children, they might not be doing anything aggressive when you get to your set recording point, but they might do in between those points. So you're not seeing a representative sample of their behaviour.
OK, so now let's look at the different types of observations that we can use. So the first one is naturalistic observation. So this is where we're observing behaviour which is... spontaneously occurring in the participant's own natural environment.
So for example we could observe children on a playground, that's where they would be anyway, so that's what they're used to. Or we could observe an animal's behaviour in its own natural habitat. This is a good method to use because it's got high ecological validity if the study is covert, so if the children aren't aware that they're being observed.
they're not likely to change their behaviour and because they're in their natural environment anyway we can generalise that more to real life. However, because it's in a natural environment the researcher doesn't have much control over any extraneous variables so there might be other things going on which would be on the researchers control and that can pose a bit of a problem. Another type of observation that we can use is called a controlled observation.
So this is where we're observing behaviour under conditions which are controlled by the researcher, so maybe in a laboratory. An example of this could be observing parents and children interacting, maybe through a one-way mirror. So they might not know they're being observed, but they might also sometimes know that they're being observed. Ainsworth's strained situation study is a good example of a controlled observation.
These are quite good to use because they have high level of control over the environment, so the researcher can dictate what is in the room, how the observation is carried out, so they're easy to replicate. However, because the participants might know that they're being observed because they've been brought into a lab for the observation, it can cause demand characteristics, so the behaviour that is observed might not be completely natural. The last type of observation that you'll need to know about is called participant observation. This is where the researcher becomes involved in the everyday life of the participants, either with or without their knowledge.
For example, a researcher might try to join a cult to observe their behaviour of cult members from the inside. They probably wouldn't tell the rest of the cult that they were researching them, but in some cases they might do. These kind of studies tend to be very high in ecological validity if the observation is covert, because the behaviour is very, very natural. They're also quite good because they can give very detailed and in-depth knowledge about a particular group. However, it's very difficult to record observations promptly and objectively, especially if the researcher hasn't told the group that they are observing them.
They'll need to try and find somewhere private to make their observations so that they don't get found out. It's very, very difficult to replicate these studies because you won't be able to participate in exactly the same group or the same activities. It's important in observational studies that we make sure the observations of each person observing are consistent or reliable.
So one way to do this is to use two or more observers watching exactly the same thing and see how much they agree. on each event. So here is a table of observational results.
So we're looking at three different behaviours here and we've got two people observing them. Watching the same child, observer one has recorded ten incidents of kicking. but observer 2 has only observed three incidents of that behaviour. So with punching, there's not much agreement, and with shoving, there's not much agreement either. To calculate the inter-rater reliability, we would look at adding up the total agreements and then dividing by the total number of observations.
So to find the agreements, it's generally the smallest number, so here they've agreed on three, but observer 1 has recorded seven more instances of that behaviour. Here again they've agreed on three types of behaviour, but observer 2 has recorded two more incidents of that behaviour. Here they've agreed on one, but observer 1 has recorded one more incident of that behaviour. So for this example, they've agreed on seven of their observations out of a total of 24, which gives us 0.29.
Now we look at anything above 0.80 as being high inter-rater reliability. means is that the two observers are agreeing and recording the same data. Here there's a lot of disagreement so we would say that this observation is quite unreliable. You won't have to actually calculate interrelator reliability in an exam, but you could be asked how it could be assessed or how to improve the reliability of observations.
So one of the things that we can do is to clearly operationalise the behavioural categories, to say what we're counting as a shove or what we're counting as a kick, for example. That way anyone researching and doing the observation knows exactly what behaviour to look out for. Another way that we can make observations more reliable is by training observers in observational techniques.
So showing them how to do time sampling or how to do event sampling will make it much more reliable. And lastly, it's always a good idea to try and pilot an observation study just to see if the behavioural checklist and behavioural categories cover all of the eventualities. So if there's any behaviours that the observers notice which aren't on the checklist, they might need to be added in. This is another really good way to improve the reliability of an observational study.
And that's it for this tutorial. Thanks very much. I'll see you again soon. Bye.