Section 2.5: Extraneous variables
From Research Methods in Psychology
Contents |
[edit] Introduction
THE THIRD TYPE OF VARIABLE THAT WE NEED TO CONSIDER ARE EXTRANEOUS VARIABLES. THESE ARE VARIABLES THAT AFFECT THE DEPENDENT VARIABLE, ALONG WITH THE INDEPENDENT VARIABLE. IF WE WANT TO GET ACCURATE AND ‘TRUE’ RESULTS FROM OUR EXPERIMENTS WE NEED TO TRY TO REMOVE OR MINIMISE THE EFFECTS OF EXTRANEOUS VARIABLES. WHEN YOU READ ABOUT OTHER PEOPLE’S RESEARCH, YOU SHOULD SEE IF YOU THINK THAT THERE ARE ANY UNCONTROLLED EXTRANEOUS VARIABLES IN THEIR RESEARCH.
The third type of variable is the trickiest for you to deal with, and this type is a pain for other researchers – these are the extraneous variables. Extraneous variables are additional variables, which, along with the IV, affect the DV. When you design studies, you always need to try to take extraneous variables into account, and either minimise them or remove them.
Removing the effect of extraneous variables is called controlling, and the specific steps that you take are known as controls. (Note that this is a second, and different, use of the word control, from the one that we saw earlier, which referred to the control group).
If there are extraneous variables, which you have specifically dealt with, and ensured that they will not have any effect, then these variables are referred to as controlled variables. In addition, when you read about studies, you should try to determine whether there were any uncontrolled extraneous variables were not accounted for.
As we saw earlier in this chapter, extraneous variables can have two types of effect: random and systematic.
Random extraneous variables affect all of your participants randomly – each person’s score on a dependent variable will be increased or decreased a little because of this, but on average the effects cancel each other out, and are therefore not all that problematic. Random extraneous variables increase the amount of variation, and obscure the effect of the independent variable, so we would like to reduce them as much as possible, however we cannot hope to eliminate random extraneous variables. Systematic extraneous variables are much more of a headache. A systematic extraneous variable affects one level of the IV, but does not affect the other level of the IV. If we find a difference between our two (or more) groups in an experiment, we cannot be sure that the differences were not due to extraneous variables. Similarly, if we do not find any difference, it may be that there would have been a difference, but the extraneous variables have obscured its effect. In this section, we will consider the different types of extraneous variables, and look at some of the ways that extraneous variables can be controlled. We will look at four types of extraneous variables: situation variable, participant variables, participant effects and experimenter effects.
[edit] Situation Variables
| Crucial Concept: Situation variables are anything about the situation that may affect the performance of your participants. |
Situation variables are dangerous, because if you are not careful, they will exert a systematic effect on your dependent variable. If you are carrying out an experiment in which you decide to give two groups of participants a memory test, you have several options about how to test these two groups, some of which are listed below:
- Ask everyone to arrive at the same time. Test the organised group first, and then, when you have finished, test the disorganised group.
- Ask one group to come in the morning and one group to come in the afternoon.
- Ask both groups to come at the same time, on different days.
- Test both groups on the same day, at the same time, in different rooms.
What would you do? Decide, before you read on. Below, we will identify some of the problems associated with each of those methods.
- If you ask all of your participants to arrive at the same time, one group will have to sit around and get bored while the other group does the task. Because that group has been sitting around getting bored, the groups will no longer be similar, you will have the ‘fresh’ group and the ‘bored’
- If you test people on the same day, the time that you test them may have an effect. If you test both groups in the morning, the earlier group may be more tired, and the later group (as lunchtime approaches) may be hungry. If you test in the afternoon, the first group may be suffering from a drop in arousal as they digest their food. If you test them at the end of the day, the second group may be lethargic and wanting to go home.
- You could test both groups on different days at the same time – this would avoid the problems that we identified above. However, here you may encounter a systematic variable that is something to do with the day of the week. When I started lecturing there was a popular student night on a Tuesday night. The lecture on Wednesday morning had fewer people in it than other days, and those attending were noticeably less enthusiastic than usual.
- Testing in different rooms solves some of the problems that we identified above. However, the rooms may differ in size, atmosphere, temperature, lighting, and other environmental variables, and any of these might have an effect on results. Moreover, you may have to use different experimenters who may have a different an effect on the way that the participants react. (Although this effect is less likely to be a problem in this type of cognitive psychology experiment.)
OK, let’s stop being so negative, and have a look at what we can actually do. One approach is to test everyone in the same place, at the same time. To do this, you have to have some method of presenting the material individually. You might give out written materials, or present spoken materials through headphones. However, testing everyone at the same time might be difficult or in some cases impossible.
A different approach is to try to cancel out the effects of any extraneous variables, to ensure that the effects of any extraneous variables are consistent across groups, this ensures that they will not have a biasing effect on the experiment. Thus, if you have two experimenters, you could arrange that each experimenter could run half of each group. Or, half of each group could be tested in the morning, and half in the afternoon. Or, half of each group can be tested in one room, and the other half in a different room. By splitting the groups in this way you can ensure that, if the extraneous variables that we mentioned do have an effect, they would an equal effect on each condition.
[edit] Participant Variables
Everybody is different. (If you talk with your friends at university who study other subjects about the things that you have learned in your studies of psychology, they may be cynical and say “Huh. How can you say that? Everyone is different.” To stop this happening, before you say anything, you should say “Of course, everyone is different, but generally …” )
If you want to carry out research using ‘a person’ or ‘an adolescent’ or ‘a 45 year old female’ or ‘a student’ or ‘a person has been diagnosed with autism,’ you cannot simply go and find one of these people – because every person that you choose may be different from all of the other people in that category.
This leads to two issues, the first is that we must select a large enough group of participants, and we must select them in an appropriate fashion, to ensure that they are representative of all the people that we are interested in – we'll deal with this issue in Chapter 6. Second, the fact that people are different means that we have an additional extraneous variable – a participant variable. The effect of this extraneous variable may be either systematic, or random. We must try to ensure that the effect is random rather than systematic (this is very important), and we must try to ensure that the effect of the extraneous variable is as low as possible. We will look at three ways of dealing with, or controlling, this extraneous variable
The first method is not to study two groups of people at all, but rather to study the same people twice – doing this is called a repeated measures design. We discuss the pros and cons of this design later on in this chapter.
The second method you might use is to create two matched groups of participants, for example if you were studying the effects of organisation of material on learning, you could pre-test people’s memory, and then pair each participant with a second participant with a very similar score. In this way, you would ensure that the two groups were very similar in respect of memory. This design is called a matched groups design, and is similar to the repeated measures design. Both the matched groups and the repeated measures designs are examples of related groups designs. Although the idea of matched groups seems to overcome certain problems, the great practical difficulty comes in knowing that you have really matched individuals properly. The matched groups design also takes much longer and adds greatly to the inconvenience for the researcher, and more importantly, the inconvenience for the participants (who are frequently giving their time as volunteers).
The third and the most commonly used method of controlling for individual differences is to randomly assign participants to groups. Although this may seem crude, as long as your sample sizes are not very small, it is likely to be an effective method. It is very important that you really do randomly assign participants to conditions. You should be aware that many things that appear at first sight to be random turn out not to be random.
Before reading on, look at the following descriptions of methods of assigning participants to conditions, and decided which of these methods are truly random, and which are not.
- Participants who need to leave early are in Group 1, other participants are in Group 2.
- First person to arrive are assigned to Group 1, second person is assigned to Group 2, third person in group 1, etc.
- First 10 people to arrive are assigned to Group 1, next 10 people to arrive are assigned to Group2.
- Participants asked to stand in a large group, then you draw a line approximately down the middle of the group dividing the participants into two groups.
- All participants placed in alphabetical order of surname. Participants in the first half are placed in Group 1, participants in the second half are placed in Group 2.
Which of those methods did you decide were truly random? The answer is none.
| Crucial Tip: Students use the word ‘random’ inappropriately more than almost any other word. Never use the word random when you don’t really mean random. |
There are only two types of ways of assigning people to groups randomly. The first is to put the names of all participants in a hat, stir them very thoroughly and draw them out. The second is to toss a coin for each person. (The hat and the coin do not have to be real, they can be computerised, or constructed using random number tables but the process must be the exact equivalent of using either a hat or a coin.) (Actually, there are more ways, such as block randomisation, but that's more complex. Let's say there are two simple ways)
Although randomisation would seem to be a rather crude method of dealing with participant variables, it is effective for ensuring that all groups are equivalent -especially when dealing with larger numbers. From the point of view of researchers and participants, it is very and easy to administer.
| Crucial Tip: When you carry out experiments of your own, you need to randomly assign participants to conditions, and you also need to describe how you randomly assigned participants to conditions, in your report of the experiment. |
[edit] Participant Effects
In psychology, we have a special problem that is not encountered in other sciences. In chemistry, astronomy or biology the things that we are studying will usually sit still and allow us to investigate them. A frog, a star or a chemical reaction will carry on doing what they were doing, regardless of whether anyone is watching or not. This is not the case in psychology; as soon as you know that you are in a psychology experiment, and are being investigated, you will start to behave differently. This makes psychology a much more difficult than many other sciences.
When people find out that they are in a psychology experiment, thoughts such as:
- ‘What is the psychologist really studying?’
- ‘Am I behaving in a way that would be considered ‘normal’?’
- ‘Am I spoiling the experiment for them by behaving in this way?’
In addition, because of these thoughts, people will behave differently. This phenomenon was recognised by Orne, who wrote an influential and important paper in 1962, called “On the social psychology of the social psychology experiment.” (Although Orne focused particularly on social psychology, the implications of his paper relate to many areas of psychology). Orne said that we psychologists put people into different situations, so that we can study them. What we sometimes forget is that being in a psychology experiment is a special kind of situation, and people may behave differently – people in psychology experiments do things that they would not normally do.
On one occasion, Orne asked some of his friends if they would mind doing five press-ups for him. The people he asked, naturally, said “Why?” Then Orne asked people if they could spare five minutes for a psychology experiment. When people agreed Orne asked them if they would mind doing five press-ups for him. The people said “Where?” The characteristics of the situation that make people behave in strange ways, or ways in which they would not normally behave, are called ‘demand characteristics.’
What can you do to prevent participant effects? One method is to disguise the true nature of the experiment from the participant by providing them with as little information as possible. If they do not know what you are expecting or ‘hoping for’ they cannot change their behaviour to please you in such a way that it will ruin your experiment - there should be no ‘good participant’ effects. This commonly used procedure is called a ‘single blind’ procedure.
If you tell them nothing about the experiment, the participants will be unlikely to work out the hypothesis that you are investigating. However, this lack of knowledge will not stop them from guessing what your hypothesis might be, and responding to that instead – and so their responses might well be even more random and bizarre. So, do not deny them all information. People are naturally suspicious, and people who have any knowledge of psychological research are especially suspicious. If you are conducting an experiment to examine the effects of attractiveness on judgements of intelligence, tell the participants that you are investigating judgements of intelligence. Tell the participants something about the area - just do not tell the participants the precise nature of your hypothesis. (We will come across this issue again when we consider ethical issues within psychological research, in Chapter 6.)
[edit] Experimenter Effects
In this section, we will look at experimenter effects as extraneous variables. Experimenter effects can introduce two kinds of extraneous variable - both systematic, and random. In this section, we will first look at systematic extraneous variables, introduced through experimenter effects, and then at random extraneous variables, introduced through experimenter effects.
[edit] Systematic Experimenter Effects
A further difficulty that we have in psychology is that you and I (the researchers) are also people. And people find it difficult to be objective when there is a desired or expected outcome. When cars driven by two people collide, there is rarely agreement about why the collision occurred. Similarly, researchers may not agree about what happened if they observe a particular behaviour. Imagine two psychologists watching two children at play. One of them, A, has a theory that suggests that the children behave aggressively toward one another. The second, psychologist, B, has a theory that the children are friendly towards one another. One child throws a ball towards another child. What do the two psychologists write in their notes? Psychologist A notes: 'Child throws ball in aggressive attempt to hurt other child.' Psychologist B notes: 'Child throws ball in aggressive attempt to encourage other child to join in game.'
Experimenters, as we have seen, are capable of interpreting the same behaviour differently. Subtle differences in experimenter behaviour can also alter the way that participants behave. A set of instructions read out to participants could be read in different ways to different participants.
Compare:
Please look at the following problems, thank you. Work as QUICKLY as you can. Oh, but there is no time limit. Take as long as you need.
Please look at the following problems, thank you. Work as quickly as you can. There is NO TIME LIMIT. Take as LONG as you need.
In conclusion, how an experimenter behaves or perceives certain situations can have a systematic, or biasing, effect on our results.
On some occasions there may be deliberate and malicious attempts on the part of the experimenter to manipulate the experiment in the direction that supports their ‘pet’ hypothesis. Much more likely that the researcher, while making every effort to be scrupulous and fair, will still give additional cues to participants, without realising that this is what they are doing. (If an experimenter wanted to make a deliberate and malicious attempt to manipulate an experiment, it is much easier to simply change the data.
[edit] Random Experimenter Effects
As we have seen, experimenters can introduce systemic extraneous variables (biases) into their experiments, through their behaviour, or their perception. Less serious, although much more common, is the introduction of a random (non-systematic) extraneous variables. They could make a mistake in recording information or they could make an error in judgement. As we have seen above, an experimenter who is recording information may make errors. These may be judgment errors – they saw an act as aggressive, when it was playful, or they may simply be clerical errors, they ticked the wrong box, or wrote the wrong number down. In terms of behaviour, different experimenters may behave in slightly different ways – one may talk quickly, one may talk more slowly. They may have different regional accents, and this may alter perceptions of the experimenter. If a person has a Cockney accent, the participant may think them less trustworthy than a person with a Northern accent. (Call centres tend to be based in the North of England, because people find Northern English accents more trustworthy.) It is difficult, or even impossible, for a researcher to behave in an identical manner with every participant – and this may have some
[edit] Section Summary
Extraneous variables are variables that affect the dependent variable in our experiment, and may mean that our experimental results mislead us. Extraneous variables come in two types: random extraneous variables, and systematic extraneous variables (also known as biases). Random extraneous variables may have the effect of obscuring our results, but they will not mislead us, because they affect all participants randomly. Systematic extraneous variables are much more problematic, because they bias the results of our experiment. We looked at three different types of extraneous variable: situation variables, participant variables, participant effects and experimenter effects. In any experiment, you should be aware of all of these effects, and attempt to minimise them.
