by Sandi Mann
Once the data from the simple experiment has been collected, the researchers will perform a number of statistical tests to determine whether any differences found between the control and the experimental group are statistically significant (or whether they could have occurred by chance).
We could make the experiment more complex by adding another IV to the mix – or another DV. Another IV might be placing adverts for ice cream around the mall to see whether this affects ice cream consumption. Another DV might be a measure of which flavours or types of iced treat are preferred.
An experiment could also involve either independent or repeated measures. In the ice cream example, we are comparing different groups of people in each condition, experimental and controls – the two groups are independent of each other and thus this is an independent measure. We could do, however, a repeated measures experiment where we use only one group of participants; for example, we could, in theory, measure ice cream consumption before increasing temperature, during, then again once the temperature went back to normal.
Of course, this would make no sense as most people don’t eat that much ice cream! But imagine a different experiment, such as one to measure the effect of violent computer games on aggression in children; here, an independent measures experiment would have one group of children play the games and one group play non-violent games – and then compare the amount of aggression between the two groups. A repeated measures experiment would take a baseline measure of aggressive behaviour of one group of children, have them play violent computer games, then measure aggression again. Here the measure after the intervention would be compared with that before it.
Correlational research
Not all psychology research is conducted by carrying out experiments. An alternative is to use correlational methods to look for relationships between variables. Such relationships, or correlations, can be found to be either positive, negative or non-existent. Positive correlation means that as the amount of one variable increases, the other one does, too. For example, age of children is correlated with height (in that, on average, older children will be taller than younger ones, though there will always be exceptions). Negative correlation means that as one variable increases the other decreases; for example, as age of adults increases, elasticity of their skin decreases (sadly). If no correlation is found, this means that the two variables have no relationship with each other; for example, head circumference does not correlate and thus has no relationship with intelligence.
Spotlight: Correlation and causality
A word of warning: correlation does not imply causality! This means that just because two variables may correlate, we cannot say for sure that one causes the other. Increasing A may cause B to rise, or increasing B may cause A to rise. Or an entirely different variable, C, might be responsible for both A and B rising together.
Correlation can be quantified by using a mathematical measure called the correlation coefficient. Once calculated, a correlation coefficient will have a value ranging from −1 to +1. A value close to one indicates a strong positive correlation. A value close to −1 indicates a strong negative relationship.
Statistical tests of significance can work out how likely any correlation found is to have occurred by chance, and how likely it is a real reflection of a relationship between the two variables.
Spotlight: Experiments and correlational studies
Be alert to the difference between experiments and correlational studies.
An experiment isolates and manipulates the independent variable to observe its effect on the dependent variable (while attempting to control for any extraneous variables that could be responsible for the effect). Experiments can thus establish cause and effect.
A correlational study, on the other hand, identifies variables and looks for a relationship between them. It cannot predict cause and effect.
Observational methods
All psychological research involves observation to some extent, but observational studies are investigations where the researcher observes a situation and records what happens but does not manipulate an independent variable. There is no intervention and the researcher does not influence the proceedings in any way. These methods can be useful when it is not ethical to manipulate a variable or when there is a need to minimize demand characteristics (whereby participants feel the need to provide the researcher with what they think they are looking for in an experiment – we will look at this further below). There can be, however, the problem of observer bias, whereby the researcher inadvertently skews their recording according to the results they expect to find. Observational studies can also be quite invasive in terms of privacy and there is the whole issue of whether people react differently when they know they are being observed (again, we will look more at this later in this chapter). Sometimes this can be overcome by not informing the participants that they are being observed, perhaps by having the researcher join in with the activities they are observing.
There are generally two ways that observations can be carried out: using time sampling methods or situation sampling.
Time sampling involves observing participants at different time intervals. These time intervals can be chosen randomly or systematically. Using systematic time sampling allows the information obtained to only really apply to the one time period in which the observation took place (e.g. mornings). In contrast, the random time sampling would allow greater generalizability across all time periods.
Time sampling is not useful if the event pertaining to your research question occurs infrequently or unpredictably (e.g. buying a car or going on holiday), because you will often miss the event in the short time period of observation. In this scenario, event sampling is more useful. In this style of sampling, the researcher lets the event determine when the observations will take place.
Situation sampling involves the study of behaviour in many different locations, and under different circumstances and conditions. By sampling different situations, researchers reduce the chance that the results they obtain will be particular to a certain set of circumstances or conditions, and this significantly increases the generalizability of findings. Researchers may determine which subjects to observe by selecting them either systematically (every tenth student in a cafeteria, for example) or randomly, with the goal of obtaining a representative sample of all subjects (see the later section on the Hawthorn Effect).
Reliability and validity
Any psychology research must be both reliable and valid for it to be considered scientific. Reliability is sometimes used to mean ‘consistency’; a reliable research technique will yield the same results consistently if we were to repeat the measure time after time. Thus, if I were to measure intelligence, the technique I use should roughly yield the same results for the same people each time (of course, there are factors that could affect reliability, such as practice effects, but more on that later).
Validity refers to the extent to which a research technique actually measures the behaviour it claims to measure. Thus, I could use a tape measure to measure head circumference as a measure of intelligence; this would be reliable (in that the tape measure is likely to yield a very similar circumference of a head each time I was to repeat the measurement) but not valid as a measure of intelligence (but perfectly valid as a measure of head circumferences).
There are ways to test for reliability and validity in order to produce a ‘score’ that can indicate whether the technique has good or weak reliability and validity, but discussion of this in great detail is beyond the scope of this book. Suffice it to say, reliability tests look for consistency or correlation between measures, so one way would be to correlate scores of a test at time point 1 with scores on the same person at time point 2; a high correlation indicates high reliability. There are various ways to obtain these two sets of data with which to correlate.
For validity, researchers are also looking for correlations, but this time with something that will tell us that what we are meas
uring is what we intend to measure. Imagine we want to measure intelligence – one way to check that our measure is valid would be to see whether scores correlate with other existing measures of intelligence or other indicators of intelligence (such as exam grades). Again, scores nearer to 1 indicate the highest correlation and thus validity.
Internal and external validity
There are several different types of validity, and psychologists talk about factors that may threaten the validity of their research. These threats can affect internal validity or external validity. Internal validity refers to how much we can be sure that the findings are due to the variables under consideration rather than to extraneous variables, while external validity considers how much the results from psychological study can be generalized more widely.
Threats to internal validity: Psychology researchers Campbell and Stanley (1966) identify and discuss various types of extraneous variables that can, if not controlled, jeopardize an experiment’s internal validity:
• History These are the unique experiences subjects have between the various measurements done in an experiment that could influence outcomes. Thus, tests of aggression in children might be subject to history effects if some children access violent video games at home and others don’t.
• Maturation These are natural (rather than experimenter-imposed) changes that occur as a result of the normal passage of time. For example, the more time that passes in a study the more likely subjects are to become tired and bored, more or less motivated as a function of hunger or thirst, older and so on. Reading tests on children, for example, will always be subject to maturation effects.
• Testing The actual testing itself might change people; for example, taking an IQ test might improve people’s ability to do well at IQ tests so it might be the actual measure that leads to an apparent increase in IQ before and after an intervention, not the intervening experimental intervention.
• Selection If the comparison groups are different from one another at the beginning of the study, the results of the study are biased. For example, testing a depression intervention on two groups of volunteer patients might be influenced by the difference in the types of people who volunteer.
• Experimental mortality Subjects drop out of studies. If one comparison group experiences a higher level of subject withdrawal/mortality than other groups, then observed differences between groups become questionable.
Threats to external validity: There are a few major issues around external validity:
• Population Can we generalize from a relatively small sample (e.g. of psychology students or middle-aged shoppers, or people walking on a particular street at a particular time) to the population as a whole?
• Location Can the results obtained in a laboratory setting really tell us how people will behave in real life?
• History Are findings from older studies still valid today?
Demand characteristics and the Hawthorne Effect
There are a number of other important threats to the validity of research that every serious research psychologist must try to counter.
Demand characteristics occur when participants figure out the purpose of the experiment and unconsciously (or even consciously) change their behaviour to fit the aim. They might do this in order to ‘help’ the researcher or in order to make themselves look good (sometimes called the social desirability effect). Alternatively, the participants might try to skew their responses so they don’t fit what is being looked for, to try to ruin the study (some people are like that!). One way to avoid demand characteristics, then, is to deceive participants about the nature of the study, or to omit to tell them what you are hoping to find. However, there might be ethical considerations about their giving informed consent if they don’t know enough about the study (see the section on ‘The ethics of research’ later in this chapter).
Spotlight: A definition of demand characteristics
The concept of ‘demand characteristics’ originated in the work of Martin Orne in 1962. Orne more recently defined demand characteristics as ‘the totality of cues and mutual expectations which inhere in a social context … which serve to influence the behaviour and/or self-reported experience of the research receiver’ (Orne and Whitehouse 2000).
Double-blind studies are another way to counter demand characteristics; this is where neither the experimenter nor the participants know which experimental group they are in/dealing with so cannot consciously or unconsciously tailor their behaviour or responses (sometimes the experimenter themselves can be the source of demand characteristics as they imagine or look for results that will prove their theories).
In the late 1920s early 1930s a series of experiments took place at Western Electric’s factory at Hawthorne, a suburb of Chicago. The original purpose of the experiments was to study the effects of physical conditions on productivity. Two groups of workers in the Hawthorne factory were used as guinea pigs. One day the lighting in the work area for one group was improved dramatically while the other group’s lighting remained unchanged. The researchers were surprised to find that the productivity of the more highly illuminated workers increased much more than that of the control group. The employees’ working conditions were changed in other ways, too (their working hours, rest breaks and so on), and in all cases their productivity improved when a change was made. Indeed, their productivity even improved when the lights were dimmed again. In fact, even when the conditions were returned to how they were at the start, productivity increased.
It was concluded that it wasn’t actually the changes in physical conditions that were producing the increase in productivity. Rather, it was the attention that they were being given by the researchers that produced the improvements. They felt happy to be singled out and to feel that someone cared about their working conditions.
‘Careful studies of this … group showed marked increases in production which were related only to the special social position and social treatment they received. Thus, it was the ‘artificial’ social aspects of the experimental conditions set up for measurement which produced the increases in group productivity.’
J. R. P. French, on the Hawthorne Experiments, ‘Experiments in field settings’, Chapter 3 in L. Festinger and D. Katz, Research Methods in the Behavioral Sciences (New York: Holt, Rinehart & Winston)
What became known as the Hawthorne Effect stated, then, that individuals may change their behaviour as a result of the attention they are receiving from researchers rather than because of any manipulation of independent variables. In other words, we behave differently when we know we are being observed.
The ethics of research
‘Psychological investigators are potentially interested in all aspects of human behaviour and experience. However, for ethics reasons, some areas of human experience and behaviour may be beyond the reach of experiment, observation or other form of psychological intervention. Ethics guidelines are necessary to clarify the conditions under which psychological research can take place.’
British Psychological Society Code of Human Research Ethics, p. 4
Psychologists who conduct research must carry it out ethically. Today there are strict ethical codes in place from bodies such as the British Psychological Society and the American Psychological Association governing research. In addition, researchers wanting to conduct studies usually have to seek ethical approval from their own institution. These ethical guidelines are designed to protect participants in research and include adherence to the following principles:
• Principle of Positive Self-regard: this states that research participants should have as high an opinion of themselves at the end of the process as they did at the start and should not be harmed either physically or emotionally by the process. Part of this involves ensuring that they are fully debriefed at the end; participants must be given a general idea of what the researcher was investigating and why, and their part in the research should be explained. They must be told whether they have been deceived
and given reasons why. They must be asked whether they have any questions and those questions should be answered honestly and as fully as possible.
• Principle of Informed Consent: participants have the right to know what they are letting themselves in for. Participants must be given information relating to the purpose of the research, the procedures involved in the research, any foreseeable risks and discomforts involved, length of time the subject is expected to participate and whom to contact for answers to questions about the study. Deception for the purposes of the study should be balanced carefully against the need for informed consent.
• Principle of Competence: researchers should conduct high-quality research which meets high technical and professional standards.
• Principle of Confidentiality: information should be guarded so it cannot be used in a way that would be detrimental to the participant.
• Principle of Voluntary Participation: participants should be free to withdraw from the study if they choose to. This causes ethical issues sometimes when students are expected to take part in research in order to gain credits, but care should be taken to adhere to this principle as far as possible.
Many classic psychological experiments would never gain ethical clearance today, for example ‘Little Albert’ (see Chapter 14), Harlow’s monkey experiments (see Chapter 10), the Stanford Prison Experiment (Chapter 12) and the Milgram Experiment (Chapter 12). See the ‘Dig deeper’ section at the end of this chapter for more on this.