14 Research Methods 4: Testing Procedures

Sudarshana Sen

1. Objective

The objective of the module is to enlighten the students regarding the use of experimental method in social science research. This method enables the student to understand the ways to find a cause of an event.

2. Introduction

Experimental research is based on testing of hypothesis which states the relationship between a dependent and an independent variable. Experiment is a well known method in social science for establishing causality. It enables the researcher to measure the value of the dependent variable, introduce the independent variable which the researcher suspects to be the cause and to observe whether any change occurs in the dependent variable for this. Experiments are highly controlled method of attempting to demonstrate the existence of a causal relationship between one or more independent variables and one or more dependent variables.

The four elements of an experimental research design are: manipulation, control, random assignment and random selection. When the researcher purposefully brings in changes in the environment in which the experiment is conducted to observe the changes in the variable s/he is concerned about, it is called ‘manipulation’. ‘Control’ is exercised to prevent the influences of extraneous variables. Random assignment means that if there are groups/treatments in the experiment, participants are assigned to these groups/treatments randomly (like the flip of a coin) no matter who the participant is. This also means that a participant has an equal chance of getting into all of the groups or treatments in an experiment. This process helps to ensure that the groups or treatments are similar at the beginning of the study so that there is more confidence that the manipulation (group or treatment) “caused” the outcome. Random selection refers to how the researcher draws the sample of people for the study from a population.

Experimental research is very useful in explanatory researches. In a sense, experiment can be defined as observation under controlled situations. The module explores the issues, possibilities and ways of conducting experimental research.

3. Learning Outcome

The student will understand that the experiment is a means of collecting data to allow the researcher for making inferences about the testability of a hypothesis. A hypothesis asserts relationship between variables under certain conditions. Since a variable has at least two values (for example, gender), the experimenter must establish which value has to be manipulated (independent variable) in order to note its effect on the dependent variable. The student will come to know that experiment, as a method of data collection and analysis, is a way to determine what changes in the independent variable cause the observed changes in the dependent variable.

4: What is an Experimental Study?

John Stuart Mill (1930) formulated the logical proofs which form the basis of experimental designs. His analysis provides two methods, commonly called method of agreement and method of difference, which function as the basis for designing experimental studies. Putting it simply, the method of agreement states that in two situations A and B if we find the presence of X in both result in producing Y, then it can be inferred that there is a causal relationship between X and Y. How do we derive this? Suppose X is present in situation A and Y is also present, and in situation B the same happens, and corroboratively whenever there is an absence of X, Y is also absent, we can infer that there is a relationship between X and Y and that the relationship is of a causal nature. It means that X  is the cause of Y and that whenever X occurs, Y follows and not vice versa. This method has the following advantages. First, such logic helps rule out the impact of various factors that may occur together. Second, it helps us to recognize that two or more factors may occur together. It also allows us to observe that in concrete situations X always occur prior to Y. So X can be treated as the probable cause of Y.

In case of the method of difference, it is formulated that in two situations A and B, if X is present and Y occurs, while in the other situation B, the absence of X does lead to absence of Y, then it can be inferred that X is the cause of Y. The difference between the two methods is that in the first we agree that if X is present and Y is also present, presence of X in another situation entails the presence of Y. But here we look for the absence of X in one situation which leads us to assume that it will also lead to the absence of Y in another situation.

4.1: An Experimental Study

Based on the logic of the two methods, the method of agreement and method of difference, there are various types of experimental research designs. It is stated that if there are two or more cases, such as A, B, C, and in one of them X can be observed, while in the other where X is not present, Y does not occur, then it can be asserted that there is a causal relationship between X and Y. There is one problem that such a design does not answer. There may be a possibility that other factors are significant than the one classified as X in our example. But it depends on the hypothesis formulation whether or not the researcher intends to cure the weakness and look for the other factors as well.

The experimental logic in its pure form can be executed in a laboratory setting. In order to conduct an experimental study, the researcher starts with a hypothesis. In order to test this hypothesis the experimenter must form two groups, the control group and the experimental group. Each group is matched in every possible way. They will be put into identical environments which are controlled in some important respects by the experimenter. The experimenter will then deliver certain amount of stimulus to the control group. If there is variation in the experimental group and this variation can be linked with the variations in the stimulus, then the experimenter can assume that there is a causal link between the stimulus and the result since there is no other difference between the two groups. The variable that is controlled by the experimenter is called the ‘independent variable’ (X) and anything that varies as a result of this is called ‘dependent variable’ (Y). The purpose of the experiment is to create a standardized situation for the researcher to study in which all variables are under the control of the experimenter and in which the results of manipulated variables can be studied and measured. If some correlation is found between variables, the researcher has to show that this is a causal relationship rather than a coincidence (McNeill et al. 2010).

It is a common practice to search for a single cause of a single effect (event). But scientific thinking tells us that there are necessary and sufficient conditions that also influence an event. Scientific explanations differ from common-sense in way of explaining that there is a single cause of an event. Scientific explanations take into account not only conditions in which one identified stimulus can be preset, but also bother about contributory conditions, contingent conditions and alternative conditions – all of which are expected to be operating upon the effect. A ‘necessary condition’ is one which must occur to be the cause of an event. A ‘sufficient condition’ is one which is always followed by the effect. A condition may be both necessary and sufficient for the occurrence of a phenomenon. A ‘contributory condition’ is one that increases the probability that a given phenomenon will occur, but does not make it certain. This is because it is one of the many collective factors which determine the effect. A ‘contingent condition’ is one under which a given variable is a contributory cause of a given phenomenon. An ‘alternative condition’ is that which may make occurrence of a phenomenon more likely (Selltiz et al. 1959: 81-85). These help us understand that no phenomenon can occur due to a single cause, or putting it in a different way, it can be said that there are multiplicity of causes behind the occurrence of any phenomenon. Experimental research designs and its types depend on this logic.

5: Classical Experimental Research Design

To execute a classical experimental design, a researcher has to take two series of observations and situations. In the two situations A and B, there may be elements E, F, X in one and E, F, non X in the other. But both may not display same results as in situation A, we may find the presence of Y, whereas in situation B, we may not find Y. The situation in which we have stated X to be present and have observed Y to be present too, is called the ‘control series’ and the other situation in which there is an element called non X, is called the ‘experimental series’. It is the experimental series that the researcher wants to test, i.e. whether or not X and Y are related and if there is a causal relationship between the two. The results of both are measured to see if the hypothesis has correctly stated/proposed a causal relationship between the two X and Y.

In implementing an experimental design, one has to first identify a problem for research and then state a hypothesis where the researcher wants to find out something. To do this in real life, the researcher has to select a population appropriate for the hypothesis to be tested and divide the sample to be studied into two equally representative groups. There should not be significant difference between the two groups except the experimental variable (Y, which the researcher wants to test). Both the groups will be given a similar test, say, for example, the researcher wants to know the attitude of people towards nuclear electric power stations in neighbourhood. From the results of both the groups, the researcher will match the response of the test and arrive at a conclusion about factor that may signify the difference. The factor (X) identified will be the cause of the difference (Y).

The ways in which the two tests are evaluated are called i) precision matching, ii) matching by frequency distribution, and iii) by randomization. In precision matching, every element in the control series is represented with the same quality in each experimental series. To find out individual(s) in the control series with same (matched) qualities in experimental series is a difficult endeavour. In matching by frequency distribution, the researcher seeks to have equal representation of relevant factors. In randomization, the researcher starts with the assumption that half of those having a particular characteristic will fall in each group even though the researcher may discover only later that a certain variable is significant.

There are some problems in the classical experimental design. First, there are problems in recognizing and controlling the variables. ‘Control’ here in practice means that the values of the variables are or may be deliberately minimized to achieve desired results. Second, the causal relationship may not be clearly stated. For example, if the researcher finds that X is the cause of Y, there still remains the possibility that a) Y is the cause of X, b) both X and Y are caused by some other variable, c) X may be the cause of Y, but only in the presence of some unknown variable, d) X does not cause Y or simply it is accidental in the two chosen situations. The third problem is the element of time. Any social variable requires time to affect social behaviour. If tests are carried out immediately after the stimulus (experimental variable), the change is less likely to be observed. The fourth problem is that the classical design is stated in a very simplistic manner. It is stated in a quantitative form. Through the statistical technique with the use of probability theory, it may be proved that Y is the cause of X. With the use of better techniques of measurement like concomitant variation (X and Y occur together in the way predicted by the hypothesis that X is a contributory condition of Y), for example, this has become easier. So the classical design though paved the road for experimentation in social science, there are various other types that have been formulated at present. Goode and Hatt (2006) have described deviations from classical design as ‘Before-After’ and ‘After Only’ study designs whereas Kothari (1991) has devised experimental designs into two broad types: Formal experimental designs and Informal experimental designs. Taking both into consideration and the overlaps between the two experimental designs in the next section.

Self-Check Exercise 1

1. What is Causality?

Causality is searching and establishing a cause for an event.

2. What is an Experimental Design?

An experimental research design is a controlled method of establishing causality.

3. How can one prove causality between two variables?

In order to establish causality, a researcher has to hypothetically propose a relationship between two variables (independent and dependent), and prove through experiment that the stated causality is true or false.

6: Deviations from the Classical Experimental Research designs

Before-After Experimental Designs

In such a design, a single group is selected and the dependent variable Y is measured before the introduction of the experimental variable (often called treatment). The effect of the introduction of the treatment on the dependent variable Y is measured after the introduction of the stimulus (treatment). The effect of the treatment is level of phenomenon after the treatment (Y) minus the level of phenomenon before the introduction of the treatment (X).

After-Only Experimental Designs

In this design, two groups are selected and the treatment or the stimulus is introduced into the experimental group only. The effect of the dependent variable is measured at the two groups, but at the same time. The impact of the treatment (stimulus or experimental variable) is assessed by subtracting the value of the dependent variable (Y) in the control group from its value in the experimental group. Here the assumption is that the two groups, control and experimental, are identical with respect to their behaviour towards the experimental variable.

Before and After Experimental Design /Successional Experimental Design

In this design, two groups are selected and the dependent variable is measured in both the groups – control and experimental – for an identical time-period before the treatment (stimulus) has been introduced. The treatment (stimulus) is introduced only in the experimental group and the dependent variable is measured in both the groups for an identical time-periods. The effect of the treatment is measured by subtracting the change in the dependent variable in the experimental group/situation.

Completely Randomized Experimental Design (CRD)

The characteristic of this design is that the elements are randomly subjected to treatment or vice-versa. It provides maximum degrees of freedom to the error. Such designs are used when the experimental variations are homogeneous or technically speaking when all the variations due to  principles: replication and randomization. In a randomized design, a sample is selected randomly from the population and randomly identified as control and experimental groups. The two groups are given two different treatments (stimulus) and each group is tested before and after the treatment is introduced. The results are compared and the findings help to either accept or reject the hypothesis. A replication design tries to control differential effects of the extraneous independent variables and it helps to randomize any individual differences among those conducting the treatment. The sample is taken randomly from the population and equal number of elements is put in each group so that the size of the group is not likely to affect the results of the study.

Randomized Block Design (RBD)

In this case, the elements are first divided into groups, known as blocks, where elements in each block are given same number of treatments and each element in one block is randomly assigned a treatment. At the block level, the effect of extraneous variable is held to be fixed so that the contribution to the total variability is measurable. The purpose of such design or randomization is to take care of the extraneous variable and its effect on the experiment.

Latin Squares Design (LSD)

This type of design is usually used to test effectiveness of seeds, fertilizers etc, i.e. this is more used in agricultural studies. Here the field is divided into blocks and each block is again divided into many parts, but the experimental variable is applied in such a way that each of the stimuli is used in every block. The design presents as many independent variables as there are subjects, but the variables are presented in a different and unique order for each subject. That is, a table presenting the results of LSD will have four experimental conditions (independent variable) once in each row and once in a column. Each of the four subjects is exposed to all four independent variables, but the orders in which they are presented are different for each subject and each independent variable appears in one position only once.

Factorial Design (FD)

This design is used to see the effect of two factors on the dependent variable. Here, the extraneous variable is called the ‘control variable’ and the independent variable (which is to be manipulated) is called the ‘experimental variable’. The sample is divided into four cells/groups. Each of the four combinations provides one experimental condition. Elements are randomly assigned to each treatment in the same manner. In a more complex factorial designs, three or more independent variables are considered simultaneously. From this design, it is possible to determine the interaction between each possible pair of variables (the three variables are one experimental variable and two control variables). Factorial designs provide equivalent accuracy as with experiments with one factor with much less labour. Moreover, comparisons over such effects are easily determined. The disadvantage of this design is that presentation of more than one independent variable requires a larger number of experimental groups.

Quasi-Experimental Designs

a). Ex-post-Facto Research Designs

This design also referred to as the post-test-only design where the experimenter introduces the test stimulus in the experimental group or, in a most popular version of this design, the test stimulus is beyond the control of the researcher as s/he comes to the scene after the test stimulus had already occurred. This research proceeds from the past to the present and does not contain any orientation for the future. The researcher can control only one variable which has already been recorded (in the past).  The aim is to measure a characteristic by comparing two groups similar on all grounds except that characteristic. The researcher will have no chance to pre-test because s/he will not know when or where the characteristic so important has occurred and at what length. The procedure is to secure a list of elements to be included in the experiment having similar experiences, matched with relevant factors such as age, skill, race, nationality etc. The next step is to locate these elements on two lists matching all relevant characteristics. The test of the hypothesis is to determine relative proportions of the two groups present at a subsequent date in another category with a different experience. It is difficult to secure a sample which has clear differentiations from a group of an earlier period. Many people may have had both kinds of experience, but at varied lengths of time.

1. b) One Group Post Test Only Design

This experiment lacks a control group and the experimental group lacks a pre-test. It has only an experimental stimulus and a post-test. Researchers use this design when they want to measure many variables at the post-test, and if they are familiar with the field situation they might know what the results will be without a test stimulus (control group). Nevertheless, this is considered a weak design without a control group and it is not sufficient for establishing causality.

Matched Study

The most important way to make an experimental group and a control group is to match both pairs on identical subjects such as sex or age. Unless the important characteristics of both are matched, the control group and the experimental group cannot be levelled as equal. The disadvantage of the method is that if any element in the experimental group does not have matching partners, it cannot be included unless the researcher (experimenter) has a large pool of prospective subjects to complete the experiment with control and experimental groups. Moreover, if the relevant characteristics of the experiment increase, it becomes more difficult to apply the test because it is a tremendous task on the experimenter to match the subjects and the characteristics. Only the characteristics that are correlated with the independent and the dependent variables will affect the experiment so the experimenter needs to control those variables. Another difficulty with this is – if the groups are correlated with each other and can be matched, then the values of the dependent variable will roughly be the same for each group and therefore the pre-test values will also be equal. Since such matching on a large scale is difficult, there may be more than one variable which cannot be matched or there may be a number of characteristics that the experimenter does not match because s/he does not realize that they are relevant. Selltiz (1959: 104) has suggested randomization in place of matching as an alternative to the problem faced in such situations. Randomization can be done first by deciding the group to which each member of the pair may be assigned and in this way the uncontrolled factor without any pattern will cancel each other out. In random assignments, it is unlikely that the two groups will be substantially different in value on a particular variable and thus the opportunity for bias can be minimized.

A true experimental design can accommodate a number of groups and independent variables at different levels. With all these possibilities, it is not surprising that questions arise as to which one to use and when. A ‘two-group design’ is used when the independent variable is not repeatable and order affects are likely to happen. A ‘before-and-after design’ is used when it appears unlikely that pretesting will influence the treatment and differences between subjects will mask treatment effects. The ‘factorial design’ is used when there is a concern about the interactions between the independent variables. When the independent variables have a range of intensity levels or values and it is desirable to know the relationship between the independent and dependent variables, it is best to use a ‘parametric design’. When it is not possible or ethical to control or manipulate the independent variable, or when large samples are required for descriptive or explanatory research, other study techniques are needed, it is more appropriate to consider one of the special designs like case study or longitudinal studies.

7. Classic Example of Laboratory Experiment

Let us here refer to the famous Hawthorne experiment. It was believed until 1920s that the only way to raise productivity was to labour hard. However, some studies of physiological fatigue done prior to this period began to ward off this belief. These studies showed that fatigue, a distinct possibility in physical labour, could affect worker’s output. In 1926 F. J. Roethlisberger and William J. Dickson carried out experiments at Hawthorne Works Division of the Western Electric Company in Illinois. Four different illumination studies were conducted to understand how different levels of lighting affected worker’s efficiency. In the first experiment, the illumination was raised in stages. In the second, a control group that experienced no change in lighting was added to the experimental groups. The third effort was similar to the second effort with the exception that all windows were covered in order to exclude variation in daylight. The final experiment was the same as the previous one but only with two workers. In this way, four experiments with different purposes were held, one being on illumination and its effect, second on relay assembly test room, where it was examined if monotony of work affected output or not. In the third, supervisors were trained to improve working conditions to see the relationship with supervisors and a channel for exhausting worker’s complaints can be a way out to combat effects of fatigue in raising the output level. The fourth experiment was conducted to see a sociological picture of the actual events occurring in a factory situation and how informal, spontaneous social organization is taking place within the company structure. The researchers believed that in order to deal with the rise in output, the industrial organization should be treated as a social system.

Self-Check Exercise-2

1. What is a Randomized Block Design?

A Randomized Block Design is a deviation from the classical experimental research design. It is based on the same logic of a classical design though the method is slightly different. Here, the elements are first divided into groups, known as blocks, where elements in each block are given same number of treatments and each element in one block is randomly assigned a treatment. At the block level, the effect of extraneous variable is held to be fixed so that the contribution to the total variability is measurable. The purpose of such design or randomization is to take care of the extraneous variable and its effect on the experiment.

2. What is a Factorial Design?

A factorial design is a deviation from the classical design where the logic of experiment remains the same even though the method of testing the relationship between variables is different. The sample is divided into four cells/groups. Each of the four combinations provides one experimental condition. Elements are randomly assigned to each treatment in the same manner. In a more complex factorial designs, three or more independent variables are considered simultaneously. From this design, it is possible to determine the interaction between each possible pair of variables (the three variables are one experimental variable and two control variables).

3. Give an example of a classical Laboratory design.

In 1926, F. J. Roethlisberger and William J. Dickson carried out experiments at Hawthorne Works Division of the Western Electric Company in Illinois.

8. How to infer on the evidence collected?

The basic outline of an experimental design is simple. An experimental group is exposed to assumed causal variable while a control group is not. The two groups are then compared in terms of the assumed affect (Selltiz 1959: 94). The research design based on this simple line of logic depends on three types of evidence to be collected:

1. Evidence of concomitant variation: the investigator knows which subject has been exposed to the assumed causal variable. S/he measures all subjects in terms of the assumed dependent variable and then determines whether the dependent variable occurs more frequently among the subjects who have been exposed to the independent variable than those who have not.
2. Evidence that dependent variable did not occur before the causal variable: it is secured in two ways: one by showing that before exposure to the independent variable, both the control and the experimental groups did not differ in terms of the dependent variable; and second by measuring the position on the dependent variable before exposure to the independent variable.

3. Evidence on ruling out the other factors: this can be secured in several ways. Among the major ones are: identifying the factors that have occurred in the past, contemporaneous variables other than those exposed to the experimental variables, developmental changes that are taking place etc.

The experimental design is framed in such a way that inferences of a causal relationship between the independent and the dependent variables can be legitimately drawn. However, certain important factors are to be reckoned with care. These are: the method of selecting experimental and control groups, the point of time when dependent variable is measured, pattern of control groups used, number of possible causal variables systematically included etc.

9. Field Experiments

The laboratory method appeals to those who believe that human behaviour can be studied and explained in the same way as the natural phenomena are studied. This positivist perspective also assumes that society is a part of the nature, so the methodology of the natural science can be adopted in the study of social phenomena. But laboratory experiments cannot be conducted in field settings, i.e. in real world because such ‘experiment’ tends to relax the controls to produce naturalistic conditions. Field experiments appeal to interpretivists because they focus on how the real world is interpreted by people. Field experiments have given us insights about processes such as labelling, stereotyping and discrimination. They are also popular because they are seen as easier to generalize the findings and relating them to the real world. Filed experiments are less vulnerable to the experimenters-effect. Field experiments also incur ethical problems because they involve manipulation of behaviour and deception (Moores 1998: 4, McNeill 2010).

9.1: Classic Example of Field Experiment

In 1961, Muzafer Sherif, O. J. Harvey and Carolyn W. Sherif studied inter-group conflicts and cooperation. The researchers believed that intra-group and inter-group relations had to be analysed within the framework of on-going interactions among members involved. The subjects of the experiments were 24 twelve year old boys with similar educational levels from middle-class families. In order to prevent the influences of extraneous variables, the experimental site was selected in isolation, a Boy Scout camp in large wooded area in Sans Bois Mountains of Oklahama. The boys were first housed in a large bunkhouse for a few days and were asked informally to identify their best friends after staying together. After being identified, the members of these groups were separated into two sub-groups without strong internal ties. Through a series of common and interdependent activities, each sub-group developed a strong sense of unity and identity. In stage two, which was designed to see the effects of inter-group relations, it was thought to bring the two groups in friction against each other. In stage three, attempts were made to see intra-group relations by introducing cooperation in order to obtain super-ordinate goals for both groups but which neither group could achieve without the other group’s help. To make sure that their behavioural trends were being studied, a participant observer was introduced as camp counsellor. The findings from the experiment suggested that the two extreme values of competitive goals and of cooperation between groups to achieve goals were only possible through cooperation.

10. Issues in Conducting Experimental Research

All experiments are subject to the threats of validity. They are expressed in terms of experimenter’s internal validity and threats to external validity. We can consider an experiment valid if we are able to determine accurately the causal effect of the independent variable on the dependent variable. In order to do this, we first need to do a valid pre-test and post-test measure of the dependent variable in order to determine the overall amount of change in scores on the dependent variable.

Internal validity refers to the confidence the researcher has about the causal inference from a study that accurately explains whether one variable is a cause of another. If the three conditions of causality, namely, i) cause precedes the effect, ii) there is empirical correlation between them, and iii) relationship between them is not found to be the result of the effects of some third variable on each of the two initially observed, are observed it is said that the causal inference has internal validity. Internal validity is concerned about those things that may affect whether true measurement has been obtained using the measuring instrument. The factors that may lead to problems of internal validity are:

1. History (events that occur in society between the first and the second measurement which could explain the change in the dependent variable),
2. Maturation (other processes that may be influential by the passage of time between the two tests),
3. Mortality (happened when some of the experimental or control group leave the experiment thus affecting the two group’s comparability),
4. Instrumentation (any variation in the test whether between the two groups or over the two tests),
5. Testing (the possibility that the test itself may explain the change in the dependent variable). In case of subjects scoring very high or very low, the threat of statistical regression may arise any time.

There are also threats of external validity in experiments. The most common threat is the knowledge of participation in a study. Such knowledge is likely to influence the behaviour of the research participants (Henn et al. 2006: 123). This is known as the problem of reactivity. Reliability in experiments is also of concern to the researcher. If the causal effect of other factors, such as test stimulus, ensures at least in the experimental group that the post-test score will not be as same as the pre-test, then this can prevent the reliability of a measure (Bailey 1994: 238). It is imperative that the instrument used in the pre-test and post-test be rigorously tested prior to the experiment in order to ensure the truth of reliability assumption.

11. Advantages of Experiments

The scope to establish cause-effect relationship is the obvious advantage of an experimental design. No other method in social science matches with an experimental design precisely for this. In a survey method, for instance, the usual way to establish cause-effect relationship is cross-sectional rather than the longitudinal method used in experiments. An observational study is also longitudinal, but generally the researcher cannot control the effects of extraneous variables that have an impact on the relationship between the independent and dependent variables. A documentary study is also longitudinal, but there is also very little scope to control the extraneous variables. If the researcher wishes to study causality, experimentation is the obvious choice. The more s researcher can establish control, the more likely s/he will take up experimental design. The experimental design gives an opportunity to do a longitudinal study. In order to study changes over time, this type of research design is superior to any other form. If experiments are carried over short period of time, they give an opportunity to study change than in a cross-sectional study.

12. Disadvantages of Experiments

The main problem of experimentation in social research is its artificial setting. Artificial setting has tremendous impacts on the natural behavioural pattern of a group. Obviously, in social science research, one cannot ‘create’ an environment. For example, creation of a situation of violence and observation of such behaviour within a family or among strangers is a difficult endeavour. Also important is the fact that the effect of the situation may be unethical as this may create uncontrolled aggression and uncomfortable effects on the subjects of study.

The experimental situation is also a reactive method. Hence, it is less widely used in social science where better control is impossible. Often the social scientist in an experimental study faces a dilemma: placing subjects in a closed artificial setting in order to observe the effect of changes in behaviour and also the fact that an artificial environment may be a hindrance to observing actual behaviour.

It should also be noted that experimental studies are generally conducted on limited samples. The larger the sample becomes, the difficult it becomes to generalize on it and also to take control over extraneous variables.

13. Ethical Issues in Experimental study

Experimental studies raise ethical dilemmas concerning the manner in which researchers treat people. Some researchers consider it inappropriate to treat humans like animals, manipulating their behaviours in a laboratory-like setting. Others question the pragmatism of including some and excluding others in the two groups under consideration. Another problem is that in experiments the researcher is able to establish exactly what it is that they will be looking for as an outcome in their research. Experimentation in natural sciences does not face this problem. For example, a physicist may know exactly the time taken for an object to fall depending upon the weight of the object. S/he knows exactly and can predict accurately the time that will be taken for the object to fall on ground. But in a social world, nothing can be laid down in so clear-cut terms.

Self-Check Exercise 3

1. Write one advantage of using an experimental research.

The opportunity to establish causality is the obvious advantage of an experimental design.

2. Write one disadvantage in using experimental research.

The main problem in an experiment is the artificial setting. The artificial setting alters social behaviour to a large extent.

3. What is the ethical dilemma that usually arises in an experimental research?

The ethical dilemma that usually plagues an experimental research is the inappropriateness of treating humans as animals.

14. Summary

The purpose of experimental research is to establish causal relationships between variables. As the social phenomenon is multi-layered and complex, control over extraneous variables is very important in order to establish causality. The control can be established in an experimental research. The logic of experiments, therefore, is to study and establish causality. The classic experimental study with the different deviations employs the same logic, but the method in which the logic is employed may be different. The experimental research is conducted in artificial settings in closed environments. These are called laboratory experiments, but the experimental study can also be conducted in field situations. There has been examples of using both laboratory and field studies in experimental research tradition. The obvious advantage of experimental study is establishing causality and the main disadvantage of this method is its artificial setting. The ethical issue raised in an experimental study is concerned with the inappropriate treatment of the respondents and deliberately altering of the situation for them forcing them to act in a predetermined way.