Experiments

One of the most widely used methods in the study of behaviour has been the experiment. The goal of an experiment is to determine whether a cause-and-effect relationship exists between two variables.  The experiment is an example of quantitative research, which generates numerical data. These data can be statistically tested for significance in order to rule out the role of chance in the results.

Let us say that a researcher wants to find out if noise affects one’s ability to recall information. The aim of the study is to see if one variable (noise) has an effect on another variable (recall of information). The variable that causes a change in the other variable is called the independent variable (IV). This is the variable that the researcher deliberately manipulates. The variable that is measured after the manipulation of the independent variable is called the dependent variable (DV).

A key characteristic of an experiment is the use of controls.  The idea of "control" is that when the researcher manipulates the independent variable, all other possible variables stay the same.  In other words, the procedure must be exactly the same in both groups; the only difference should be the manipulation of the independent variable. 

For example, if we are testing the role of noise in one's ability to recall a list of words, one group would read the list while listening to music.  Another group would read the list in silence.  Otherwise, there should be no other difference between the groups.  Some examples of controls for this study would include:

• The list of words would be the same - the same words, the same font, the same size font, the same order of the words.
• The conditions of the room should be the same.  If one room has a lot of posters on the wall with information, while the other room has bare walls, this could theoretically influence the results.
• The temperature of the rooms should be the same.
• The time of day when the test is taken should be the same.

Both the independent and dependent variables must be operationalized. In other words, they need to be written in such a way that it is clear what is being measured. In the example used above, noise is the independent variable. This could be operationalized as dissonant rock music played at a volume of 100 decibels. An operationalized dependent variable could be the number of words remembered from a list of 30 words. Now we know exactly what the IV is and what you are going to measure in order to support your research question.  Simply stating that your dependent variable is "the results of the study" is not enough as it does not say anything about what is actually being measured.

Another characteristic of experiments is that they are highly standardized.  This means that they have procedures that are written in enough detail that they can be easily replicated by another researcher.

Finally, a true experiment randomly allocates participants to conditions.  With random allocation, participants have the same chance of being assigned to the experimental or the control condition. This lessens the potential for characteristics of the individuals influencing the results.

 Teacher only box

Experiment 1

• Students will give several suggestions.  The problem is that not all of them are ethical or practical.  Typical ways that psychologists increase stress are giving the participants a puzzle that they cannot solve, manipulating their stress hormones through medication, or having them give an oral presentation in front of a group of strangers who give negative non-verbal feedback.
• Some ethical suggestions would be to ask them to respond to a theoretical situation where one of the options would be aggression, playing a video game, using VR, and putting them in a situation where they may push others out of their way, have them play a game with others to observe mild aggression.

Experiment 2

• When doing this question, many students talk about lowering one's self-esteem, but raising one's self-esteem is the more ethical option. They would raise a person's self-esteem by asking them to write a short response to a question and then praising them for their thinking and writing.  Setting up a situation in which they would be complimented - or giving them a score on a test and then saying that they have scored among the highest that the researcher had seen.
• Problem-solving is very broad, so this would need to focus on something specific - e.g. a set of math of problems (the typical student response!), a set of anagrams, a set of word problems, etc.

Hypotheses

Before carrying out an experiment, researchers first read through the research which is available on the research question that interests them.  Their research question is the aim of the study, usually limited to a certain population.  For example, the aim of this study is to determine whether listening to dissonant rock music while studying vocabulary words affects recall in adolescent girls.  After the aim of the study is decided, the researcher formulates a hypothesis. The hypothesis predicts how the independent variable affects the dependent variable.

An experimental hypothesis predicts the relationship between the IV and the DV - that is, what we expect will come out of the manipulation of the independent variable. In this case, we will have two conditions: one condition where participants have to recall words with loud music, and one where the participants recall words with no music. In the second condition, there is no noise. This is called the control condition, because we compare the two conditions—that is, one with noise and one with no noise—in order to see if there is a difference.

An example of an experimental hypothesis could be: Listening to dissonant rock music played at 100 decibels will decrease the number of words that adolescent girls are able to recall from a list of 30 words. In an experimental hypothesis, the IV (listening to dissonant rock music played at 100 decibels) is predicted to have an effect on the DV (the number of words recalled from a list of 30 words).

In experimental research, it is conventional to formulate both a null hypothesis and an experimental hypothesis. The null hypothesis states that the IV will have no effect on the DV, or that any change in the DV will be due to chance.

An example of a null hypothesis could be: Listening to dissonant rock music played at 100 decibels will have no significant effect on adolescent girls’ ability to recall words from a list of 30 words; any change in the individual’s ability to recall a list of words is due to chance.

You may find it strange to make a null hypothesis, but in fact, it makes sense. The researcher wants to reject the null hypothesis to show that the predicted cause-and-effect relationship between the IV and the DV actually exists. Sometimes, however, we have to accept the null hypothesis. This would happen if the results showed no relationship between music and the recall of words. It is important to recognize that psychologists never prove anything - they can only disprove.  Our goal is either to accept the null hypothesis, which means that we have found that there is no relationship between two variables, or reject the null hypothesis, which means that there is some type of relationship between the two variables.

 Teacher only box

1.  The null hypothesis would be The presence of other people has no effect on one's level of risk-taking behaviour. The research hypothesis could be: The presence of five people observing a skateboarder would result in higher levels of risk-taking behaviour than when the skateboarder does not know that s/he is being observed. 

Although the IB usually recommends that we use one-tailed hypotheses for IA, the student could also write a two-tailed hypothesis: The presence of five people observing a skateboarder would result in different levels of risk-taking behaviour than when the skateboarder does not know that s/he is being observed.

2. The null hypothesis would be: There will be no difference in the mean scores on a mathematics exam for students that had a high carbohydrate breakfast and those that had a low carbohydrate breakfast. The research hypothesis could be: Students who have a high carbohydrate breakfast will score higher on a mathematics test than students who have a low carbohydrate breakfast.

Two-tailed hypothesis: Students who have a high carbohydrate breakfast will show a difference in average score compared to those students that had a low carbohydrate breakfast.

3.  The null hypothesis would be: There will be no difference in the mean IB exam performance of students who have a course that is primarily visual vs auditory.  The research hypothesis could be: The regular use of visual materials in the teaching of the IB Psychology course will lead to higher IB exam performance than when no visual materials are used.

Two-tailed hypothesis: The use of visuals in an IB Psychology classroom will have an effect on IB exam performance.

Field experiments

The advantage of a laboratory experiment is that it allows the researcher to control for extraneous variables  - this is, other variables that may influence the results of the study - in this case, the behaviour or cognitive processes of a participant.  But as we know from our study of validity, if we control a situation in an experiment, it may not reflect what happens in real life.  In that case, the study may have low ecological validity.

When we do studies outside of the laboratory, in the "real world", this is called a field experiment.  Field experiments have two key limitations - they cannot control for extraneous variables and they cannot be easily replicated.

There are also ethical considerations with field experiments.  Let's say that you want to do a study on helping behaviour on a city's metro (subway).  You decide that you are going to have someone faint on the metro to see if anyone will help. In one condition, the person in need of help is well dressed, wearing a suit and tie.  In the second condition, he is wearing torn jeans and a t-shirt advertising a hard rock band. 

The concern is that in field experiments it is often not practical to get informed consent. The general rule is that people may be observed in public spaces where they would expect to be observed by others.  The other ethical concern is debriefing. Often the nature of field experiments makes it impossible to explain to the unwary participants what has just happened.

Think about the study scenario described above.  This is similar to a famous study done by Piliavin, Rodin, and Piliavin (1969) which we learn about in the human relationships unit. The passengers were not willing participants in the study.  And when the door to the train opened, they disembarked and were not debriefed. How might it have affected passengers to think that they had seen someone pass out on the metro - potentially from a heart attack or some other terrible problem - and yet they didn't help? Just like with laboratory experiments, it is important that an ethics board approves any field experiments in order to guarantee the protection of participants.

 Teacher only box

1.  Well, this study is problematic on several levels.  I walk my students through CARDUD (Consent, anonymity, right to withdraw, deception, undue stress or harm, debriefing). First and foremost, there was no consent given for this study. Anonymity was not a problem because the identities of the men that came into the bathroom were not known; however, it is a rather serious invasion of one's privacy. Since the men did not know that they were part of an experiment, they would not have the right to withdraw, unless they were debriefed about the experiment. At that point, they would be allowed to withdraw their data. Since they were not told about the study and there was a person next to them that was actually a confederate, this is deception by omission. As for undue stress or harm, it could be the case that when debriefed, they would feel very uneasy about using public toilets in the future. Finally, they should have been debriefed, although it is not clear from the original study that they were...

2. Believe it or not, this study has value.  Think about concerts or big sports events.  When you are trying to make public toilets "efficient," it is important that you do not create a situation where men would take longer to use the facilities.  In addition to wanting to make sure that the people waiting in line are not unhappy because they have to go, we also want people to have free time to buy those snacks and other souvenirs!

Quasi and natural experiments

In both quasi and natural experiments, participants are not randomly allocated to conditions. In quasi-experiments, participants are grouped based on a trait or behaviour.  For example, in a quasi-experiment, you may have two groups attempt to memorize a list of words.  One group is made up of people who have been diagnosed with depression and the other group, the control, would not have such a diagnosis.  As the variable that we want to study - having clinical depression - cannot be randomly assigned, this study is a quasi-experiment. Other typical examples of IVs used in quasi-experiments include gender, culture, and age.

A subset of quasi-experiments is "natural experiments."  Although these two terms are often used interchangeably, a natural experiment usually refers to an independent variable that is environmental in nature and outside of the control of the researcher. Most natural experiments work on a pre-test, post-test design - that is, the behaviour is measured both before and after the variable was introduced.  For example, in 2017 the Czech Republic when they introduced a smoking ban in all bars, restaurants, and workplaces. Researchers may decide to measure the level of clinical depression in the city of Prague before the ban is put in place and then again six months after the ban is in place.  Once again, there is no random allocation to groups, but the researchers would be able to see if there is any difference in mental health after the change in the IV.

Quasi and natural experiments do not show direct causation, but they are able to imply a causal relationship between an IV and a DV.

Evaluating experiments

In an experiment, researchers attempt to control as many variables as possible. However, this is not always easy. Extraneous variables (also called confounding variables) are undesirable variables that influence the relationship between the independent and dependent variables.

Demand characteristics occur when participants act differently simply because they know that they are in a study. They may try to guess the aims of the study and act accordingly.

Researcher bias is when the experimenter sees what he or she is looking for. In other words, the expectations of the researcher consciously or unconsciously affect the findings of the study. Using a double-blind control can help to avoid this. In this design, not only do the participants not know whether they are in the experimental or control group, but the person carrying out the experiment does not know the aim of the study, nor which group is the treatment and which one is the control group.

Participant variability is a limitation of a study when the characteristics of the sample affect the dependent variable. This can be controlled by selecting a random sample and randomly allocating the participants to the treatment and control groups.

One other consideration is artificiality. This is when the situation created is so unlikely to occur that one has to wonder if there is any validity in the findings.

Correlational studies

Very often, an experiment cannot be carried out, but data are collected which show a relationship between two variables - these are correlational studies. The principle in correlational studies is that when one variable changes, another variable changes as well. A positive correlation is when both variables are affected in the same way. As x increases, y increases. For example, the more hours you spend studying, the better you do on exams; or the fewer hours you spend studying, the less well you do on exams. A negative correlation means that as one variable increases, the other decreases. For example, as the number of hours watching television increases, exam scores decrease.

Because no independent variable is manipulated, no cause-and-effect relationship can be determined. For example, a researcher could study the average number of hours that a child watches television and the child's level of aggression. This would be difficult to do as an experiment because it would be unethical. When the data are gathered, the researcher might find that as the number of hours of television viewing increased, so did the level of aggression in the child. This would be a positive correlation. However, it would not be possible to say whether the television viewing caused the aggression, or if it was the aggression that led the child to watch more television. This is called bidirectional ambiguity. It could also be that there is no cause-and-effect relationship at all, but that another variable might be responsible for the behaviour.

 Teacher only box

Cell phones disrupt teen sleep.

This study could be either. If it were an experiment, then teens would have to be asked to record their quality of sleep, either by self-reporting or by using some technology to monitor their sleep.  They would do that first with their regular phone use before bedtime and then again after two weeks with no phone use.

The study could also be correlational where there is no before and after position, but a survey would be given asking them about their phone use and the quality of their sleep. Students should be able to see why the experimental method would lead to more convincing results. However, even with the experimental method, one has to be cautious when drawing conclusions about causation as there are so many other variables that may affect one's level of sleep.

Using Wash your Hands signs in public bathrooms increases the number of hand washers!

This study would have to be experimental.  It would involve monitoring the number of hand washers either by having a "bathroom attendant" count the percent of those that leave without washing their hands or by filming the sinks. They would then add signs to see whether hand-washing increased. Causation could then be determined.

Men who are distracted by a beautiful woman are more likely to take financial risks.

This study would also have to be experimental.  But the limitation is that it is most likely that there were two separate conditions - the group with a beautiful woman and the group with no beautiful woman.  So, although causality may be inferred, one would have to be cautious because personality differences and experience in financial decision-making may be very different for the participants in the two conditions.

A study suggests attending religious services sharply cuts the risk of death

It would be rather difficult to set this up as an experiment. The study would have to be a correlational study.  It would look at different age groups - e.g. 45 - 55, 56 - 65, and 66 - 75.  For each group they would look at their attendance at religious services and then, perhaps two years later, see how many of each group is still alive.

Checking for understanding

Correlational studies