Research methods- Discussion Post- QUALITATIVE/QUANTITATIVE

Rose2015

social-research-a-practical-introduction.pdf

Home >Psychology homework help >Research methods- Discussion Post- QUALITATIVE/QUANTITATIVE

: Experimental Research – in the Laboratory

and Beyond

In: Social Research: A Practical Introduction

By: Bruce Curtis & Cate Curtis

Pub. Date: 2017

Access Date: July 29, 2020

Publishing Company: SAGE Publications, Inc.

City: 55 City Road

Print ISBN: 9781847874757

Online ISBN: 9781526435415

DOI: https://dx.doi.org/10.4135/9781526435415

Print pages: 146-169

This PDF has been generated from SAGE Research Methods. Please note that the pagination of the

online version will vary from the pagination of the print book.

https://dx.doi.org/10.4135/9781526435415

: Experimental Research – in the Laboratory and Beyond

Key Words

experimental research

experiment

Lost Letter

Milgram

obedience

behavioural study

ethics

Summary

An outline of experimental research in psychology, and the social sciences generally, as an

approach with which causal relationships can be effectively measured, including: (i) a brief

history of experimental research;(ii) the key themes; and (iii) the use of a classic example

to provide an overview of the key concepts relevant to experimental research.

Experimental research has been a core method in psychology, including social psychology. In this chapter,

we argue that while other approaches have become increasingly popular among social psychologists,

experimental research has much to offer all social scientists. In particular, it is the only approach with which

cause-and-effect relationships can be confidently measured. We'll begin with a brief history of the approach

before working through our key themes, making use of a well-known classic experiment to provide examples

of key concepts.

Doing Data Collection and Analysis

In the social sciences, experimental research has been largely confined to the discipline of psychology. When

other social scientists do conduct experiments, the basis is often derived from psychology. For example, the

‘Lost Letter Technique’ (Milgram et al., 1965), which has been widely replicated, including by sociologists

(e.g., Ahmed, 2010; Burwell, 1987; Tykocinski and Bareket-Bojmel, 2009), is based on the work of the

extremely influential psychologist, Stanley Milgram.

SAGE

SAGE Research Methods

Page 2 of 25 : Experimental Research – in the Laboratory and Beyond

The Lost Letter Technique

This experiment involves leaving stamped and addressed envelopes in public spaces and

tallying how many are picked up and posted. The purpose is to examine prejudice and

altruism. In Milgram, Mann and Harter's original experiment (1965), the envelopes were

addressed to four bogus recipients: either Medical Research Associates, Friends of the

Communist Party, Friends of the Nazi Party, or an individual, Mr Walter Carnap. A total of

400 envelopes were dispersed across various locations (100 for each of the ‘recipients’).

As expected, frequency of return of the envelopes was related to social desirability of the

recipient (Medical Research Associates: 72%; Mr Carnap: 71%; 25% each for Friends of

the Communist Party and Friends of the Nazi Party). That is, assumptions made about the

supposed recipients of the letters ‘caused’ unwitting participants to decide either to post

the letter, or not.

Variations conducted by other researchers have included envelopes or misdirected email

messages to individuals who appear to belong to a particular ethnic group (for example,

Bushman and Bonacci, 2004; Tykocinski and Bareket-Bojmel, 2009) or religious group

(Kremer et al., 1986), and in another case, students and non-students (Burwell, 1987). In

the latter case, additional variables included male and female names, different types of

envelope, and type of location.

Note that the Lost Letter Technique is an example of an experiment being conducted in

the field, rather than in a laboratory.

So why are experiments more popular among psychologists?

• Psychology initially developed from the ‘hard sciences’ – chemistry, biology and so on – in which the

experiment is by far the preferred, if not the only, approach to empirical research.

• Psychology has traditionally taken the individual as the unit of analysis: experiments are relatively

easy to conduct on individuals. In contrast, many of the social science disciplines take groups or

societies as the unit of analysis, and it is far more difficult (though not impossible) to involve a large

group in a robust experiment.

• Other disciplines have other well-established research approaches, such as unobtrusive research

(see Chapter 8), secondary analysis (see Chapter 10), or (supposedly) more naturalistic approaches

such as interviews and focus groups (see Chapters 2 and 5, respectively). It is worth noting that

some of these social scientists – and including some social psychologists – have been highly critical

of experimental research, as will be discussed in more detail later in this chapter.

SAGE

SAGE Research Methods

Page 3 of 25 : Experimental Research – in the Laboratory and Beyond

http://methods.sagepub.com/book/social-research-a-practical-introduction/i1017.xml

http://methods.sagepub.com/book/social-research-a-practical-introduction/i1221.xml

http://methods.sagepub.com/book/social-research-a-practical-introduction/i297.xml

http://methods.sagepub.com/book/social-research-a-practical-introduction/i686.xml

Practice Point 1

Experimental research dominates psychology, but also has potential for use across the

social sciences. It avoids some of the pitfalls of other forms of primary research through

allowing the researcher a high degree of control, lessens reliance on the perceptions

and honest disclosure of participants and allows conclusions about cause-and-effect

relationships to be drawn.

Then why include experimental research in a book aimed at social science students in general, if most social

science disciplines have tended not to use experiments? Because the experiment:

• can be, and has been, used to good effect across the social sciences, despite some limitations

• is arguably the best way of determining cause-and-effect relationships

• may avoid some of the pitfalls of other research approaches, such as participants saying what they

think the researcher wants to hear, or the researcher misinterpreting social actions.

Practice Point 2

An experiment involves the manipulation and measurement of a variable (e.g., obeying

orders, completing a task, demonstrating empathy). It is a variable-centric approach.

Origins of the Approach

The first social psychological experiment is widely attributed to Norman Triplett (1898) – at least in texts

published in the English-speaking world. Triplett appears to have been interested in cycle races, and this was

the impetus for this early experiment. He had observed that cyclists go faster when being paced or riding with

others, than when riding alone, even when racing against the clock alone. He hypothesised that the presence

of others increases competitiveness and nervous energy. To test this hypothesis – the phenomenon which

social psychologists now refer to as ‘social facilitation’ – Triplett had children wind in fishing reels as quickly

as possible, either in pairs, or alone. Each child performed this task alone three times, and in a pair three

times. Although there was some variation in the results, overall the children wound in the fishing reel faster

when doing so as one of a pair, confirming Triplett's hypothesis (it's unclear why Triplett went from cycling to

winding-in fishing reels).

There are, however, other contenders for the crown of ‘first person to conduct a social psychological

experiment’. The claim is contested in part because of difficulty in defining exactly what is a social experiment.

For example, Binet and Henri (1894) published the results of an experiment on suggestibility four years earlier

than Triplett's publication (1898) and Charcot published on suggestibility even earlier (1881), but his work is

SAGE

SAGE Research Methods

Page 4 of 25 : Experimental Research – in the Laboratory and Beyond

often considered to be of a clinical nature, rather than social. Another issue is that years may elapse between

conducting an experiment and the results being published. Ringelmann conducted experiments on what is

now called ‘social loafing’ between 1882 and 1887 but his work was not published until some years afterward

(Ringelmann, 1913). (Many of you will have observed social loafing yourselves – the phenomenon in which

members of a group get less work done than if they had worked individually. For example, five students

working on a group assignment that is to result in a 5,000–word paper will typically put in less hours each

than if each of the five is to write a 1,000–word paper.)

Regardless of who conducted the first social psychology experiment, there are several classic examples that

are well known throughout the social sciences and beyond. Perhaps the most famous of all is Milgram's

shocking study of obedience (Milgram, 1963).

Milgram's ‘Behavioral Study of Obedience’ (1963)

Milgram's work stemmed from a desire to understand the mechanisms by which humans

obey the commands of others, in particular in regards to the obedience to commands

that led to the death of millions in Nazi death camps during the Second World War.

The participants in the original study were men aged 20–50, of various income levels.

The experiment consisted of ordering participants to administer increasingly severe

punishment to another person in the context of a learning experiment, the punishment

apparently consisting of an electric shock. Each time the learner answered a question

incorrectly he (all the participants and the ‘learner’ were men) was to be punished. In fact,

the ‘shock generator’ was non-functional and the ‘learner’ was an actor. Of course, the

participants were not aware of this and believed they were actually administering electric

shocks, as marked on the machine from ‘Slight’ to ‘Danger: Severe Shock’, supposedly

ranging up to 450 volts. The ‘learner’ was strapped into a chair, ostensibly to prevent

movement while being shocked, and electrode paste was applied to the learner's skin to

‘avoid burns’. The ‘learner’ and research participant were then put into adjacent rooms so

that they were unable to see each other.

All of the participants continued administering shocks up to the 300 volt ‘Intense Shock’

level, although in many cases they questioned the need to continue and were instructed

to do so by the experimenter, despite earlier complaints and requests from the ‘learner’

to stop. At the 300- and 315-volt levels the learner was heard to pound on the wall and

thereafter failed to respond to questions. Participants were told to treat a non-response as

an incorrect answer – in other words, to continue ‘shocking’. Thirty-five of the 40 original

participants continued beyond the 300–volt level, and 26 (65%) continued to the maximum

shock level, although many participants exhibited signs of extreme stress. This point is

SAGE

SAGE Research Methods

Page 5 of 25 : Experimental Research – in the Laboratory and Beyond

worth emphasising – participants who continued did not appear to lack empathy, but this

empathy was overridden by the need to be obedient.

We will return to this classic experiment throughout this chapter, to illustrate key terms,

discuss controversies and, finally, in our discussion of ethics.

Hypothesis-testing

Deductive reasoning works from the general to the specific. It is top-down, starting with a theory for which

confirmation is sought (see Chapter 1) via development of a hypothesis to be tested. It may be associated

with objectivity and seeking patterns of causation.

Milgram's theory could be conceived as ‘… the individual who is commanded by a legitimate authority

ordinarily obeys. Obedience comes easily and often’ (Milgram, 1963: 372).

Conceptual Concern 1

The development of an experiment begins with a theory to be tested. It is a deductive

approach (inductive approaches typically end with a theory).

When using experimental research, a theory is followed by a hypothesis, which is an explicit, testable

prediction about the conditions under which an event or action will occur. A hypothesis takes the form ‘If a,

then b’, thus in situation a, b will occur. For example, a hypothesis about Milgram's research may have been ‘If

40 American men are placed in the experimental situation described above, less than 5% will obey commands

to continue to the end of the shock series’. Hypothesis-testing is discussed in more detail in Chapter 1.

Once we have our hypothesis, the next step is to move from an abstract concept to develop operational

definitions and variables. An operational definition states specifically how the variable will be measured,

manipulated, or otherwise studied. One of the operational definitions that Milgram used (1963: 374) was that

a participant who refuses to continue with the experiment before administering the thirtieth shock is deemed

a defiant subject, while any participant who complies fully with the experimenter's commands is termed an

obedient subject.

Variables in Experiments

Experimenters make use of several types of variables:

Independent variables: The factors experimenters manipulate to see if they affect the dependent variable.

SAGE

SAGE Research Methods

Page 6 of 25 : Experimental Research – in the Laboratory and Beyond

http://methods.sagepub.com/book/social-research-a-practical-introduction/i167.xml

Dependent variables: The factors experimenters measure to see if they are affected by the independent

variable.

Subject variables: Variables that characterise pre-existing differences among study participants.

The independent variables include the experimental setting itself, for example, the formality of a laboratory

setting in a university with a man in a laboratory coat overseeing the experiment and urging the participant to

continue, versus a less formal setting in an office block, with the experimenter dressed casually and leaving

the room as soon as the experiment begins. There are many other potential independent variables, some of

which have been used in further studies by Milgram and others, such as having the ‘learner’ in the same room

as the participant and shock machine during the experiment (when the ‘learner’ was in the same room, the

compliance rate dropped from 65% to 40% (Milgram, 1974)).

Conceptual Concern 2

Experimental research is a variable-centric approach. Following the development of the

hypothesis, the key concern is the development of appropriate variables and ways of

measuring these.

Milgram stated that the primary dependent variable was the maximum shock a participant administers.

This was measured by noting the shock levels, from 0 (refusal to administer any shock at all) to 30

(willing to administer the maximum shock), and measuring the duration of shocks with timers. In addition,

experiments were filmed and/or audio-taped, photographs were taken through one-way mirrors and notes of

any unusual behaviours were kept. Finally, after the experiment, participants were interviewed using open-

ended questions, projective measures and attitude scales (see Chapter 5 on focus groups and Chapter 2 on

in-depth interviews for further discussion of these).

Subject variables include the gender of participants (in later experiments females were just as likely to

administer the maximum shock as males), or education levels or social class.

Conceptual Concern 3

Experimental research is initially characterised by fixed framing. Once the key components

of the experiment are decided upon – the hypothesis to be tested and the means for doing

so – it is very difficult, if not impossible, to change them. In order to conduct statistical

analysis, the conditions of the experiment must be the same in each case. However, it is

common for variables to be altered in subsequent experiments, incorporating some fluidity.

SAGE

SAGE Research Methods

Page 7 of 25 : Experimental Research – in the Laboratory and Beyond

http://methods.sagepub.com/book/social-research-a-practical-introduction/i686.xml

http://methods.sagepub.com/book/social-research-a-practical-introduction/i297.xml

Consider Milgram's later variations conducted in an office building, with the ‘learner’ in

the same room as the participant, with the experimenter in a different location, etc. These

variations provide important additional insights.

Given the fixed framing of this approach, the validity of the variables is imperative. Validity refers to the extent

to which the variables measure what they are intended to measure. With some other approaches this is quite

straightforward. For example, when designing a survey the researcher makes sure the questions (variables)

are easily understood and that the available answers (values) capture all possibilities accurately. This may

be tested by discussing the questions in a focus group (see Chapter 5) or a pilot survey (see Chapter 6).

However, in an experiment the question of validity may be more complex.

Validity and Reliability in Data Collection

Validity and reliability refer to the robustness of the research – how likely it is that experiment ‘works’

and produces results that are accurate and meaningful beyond the experiment itself. They are often

conceptualised as:

• internal validity, and

• reliability (or external validity).

The key concern of internal validity is ‘How certain can we be that the independent variable caused the effects

obtained on the dependent variable?’ Thus, is the experiment itself valid? That is, do changes in the setting

in which electric shocks are administered (e.g., from university laboratory to anonymous office) really cause

changes in obedience, or is there some other reason changes occur?

Construct validity is a type of internal validity and is used to evaluate the manipulation (change) and

measurement of variables. It refers to the extent to which the manipulations in an experiment really

manipulate the conceptual variables (constructs) they were designed to manipulate, and the measures used

in a study really measure the conceptual variables (such as obedience) that they were designed to measure.

For example, is the strength of the electric shocks (the dependent variable) that someone is prepared to

administer really an indication of how obedient that person is? Control groups may be useful for ruling out

alternative explanations for results. In a variation on his original experiment, Milgram included a control group

who were not given continuing commands. Once participants were given the initial instructions, they were left

to decide at what point to cease giving shocks, in contrast to the experimental (non-control) groups who were

continually encouraged to keep shocking. Less than 5% of control group members gave the maximum shock

level, indicating that the ongoing commands (a manipulation of the independent variable) were an important

aspect of continued obedience.

Reliability (or external validity) refers to the degree to which the findings can be generalised to other people

and to other (external) situations, as opposed to behaviour that occurs in an experimental setting only. It also

SAGE

SAGE Research Methods

Page 8 of 25 : Experimental Research – in the Laboratory and Beyond

http://methods.sagepub.com/book/social-research-a-practical-introduction/i686.xml

http://methods.sagepub.com/book/social-research-a-practical-introduction/i801.xml

refers to the extent to which the results can be reliably produced at another time, or by another researcher

(assuming all the conditions/variables are the same). Some key considerations are:

• Is the sample representative of the population as a whole?

• Is the setting in which the research is conducted likely to affect participants' behaviour?

Reliability may be measured by repeating the experiment with another sample of participants, to see if

the same findings are achieved. If the samples are representative, you would expect to get the same

results, providing the setting and other experimental conditions are the same. (Briefly, we say that a sample

is representative when it largely reflects the composition of the population from which it is drawn – the

same proportions of men and women, of various age groups, ethnicities, etc.) This type of repetition of an

experiment is known as replication. We know that Milgram's original experiment was not representative of

the population in so far as all the participants were men, therefore it was not possible to assume that women

would behave in the same way.

Although the setting of Milgram's experiment was a laboratory, as opposed to a ‘real world’ setting, we can

assume that the experiment had a relatively high level of realism in some respects; that is, participants really

believed they were administering electric shocks. However, it is worth noting again that the results varied

according to the setting. We cannot conclude that the 65% of participants who administered the maximum

level of shock in the laboratory would also do so in other settings – the original experiment only tells us

what people would do in a laboratory. Although participants were less likely to administer the maximum

level of shock in later experiments conducted in more neutral settings, this is likely to have been due to

the deliberate manipulation of the independent variable: the more casual atmosphere, and the experimenter

therefore seeming to have less authority. It seems plausible that the experiment appeared real regardless of

whether it was conducted in the laboratory or an office building. Experimental realism is achieved most readily

in field experiments, as summarised in Table 7.1:

Table 7.1 Laboratory and Field Experiments Compared

Laboratory experiments Field experiments

Conducted in settings designed for the purpose.

• Advantages: The environment can be

controlled; the participants can be carefully

studied.

• Disadvantage: Participants may behave

differently from how they normally would.

• Example: Milgram's ‘Behavioural study of

obedience’.

Conducted in real-world settings.

• Advantages: Realism; participants may feel more

comfortable; participants are more likely to act naturally.

• Disadvantage: Experimenter has less control.

• Example: Milgram's ‘Lost Letter Technique’.

Experimental realism is also discussed below, in relation to the use of technology – video games – in a study

on prejudice.

SAGE

SAGE Research Methods

Page 9 of 25 : Experimental Research – in the Laboratory and Beyond

Cases in Experiments

Our discussion of data collection has focused on issues around variables. This is natural, as the experiment

is a variable-centred approach. However, we have also discussed cases. In experimental research, the cases

are the participants.

Conceptual Concern 4

As noted, experimental research is a variable-centric approach. Here the cases are

participants in the experiment. The number of cases is relatively few compared to other

variable-centric approaches (such as the survey). Like the survey, experimental research

is primarily quantitative, with data analysed using statistics. However, there is often

untapped potential for incorporating qualitative material, for example in debriefing the

participants.

Figure 7.1 shows us the data from Milgram's original experiment, as it might look entered into a computer

spreadsheet. There are only two variables showing: ‘Participant Number’ and ‘Shock’. The rows represent the

cases – the 40 people who took part in the experiment. The information within each of the cells in the column

headed ‘Shock’ are the values, i.e., the possible ‘answers’ to the variable (question) ‘What is the maximum

level of shock a person is prepared to give in this experiment?’ Because the cases have been entered in

order according to maximum shock level, we can see that 26 people (65% of the sample) gave a maximum

shock of 450 volts, one person refused to continue past 375 volts, and so on.

Compared to deciding on the variables, and appropriate ways of measuring them, determining the cases

is relatively easy. The key question is whether it is desirable for the cases/participants to be representative

of the population in general. Often the answer will be yes – you will want to be able to generalise your

findings to the population at large. In this situation, your next concern is recruiting a representative sample.

For example, how will you make your research known to people? Are there any incentives you can offer

them for participating? Sometimes, though, you may wish to conduct your experiment with a particular group

of people. For example, a repeat of Milgram's experiment with members of the armed forces may yield

interesting results, especially given that this experiment came about as a result of Milgram's interest in the

obedience of Nazi soldiers. Even in a situation such as this, however, you will probably want to be confident

that your sample of soldiers represents the population of soldiers from which it is drawn, for example in terms

of rank, gender, age and time served.

Figure 7.1 Word Data Matrix Showing the Variable ‘Shock’ by 40 Cases

Participant Number Shock (volts

1 300

SAGE

SAGE Research Methods

Page 10 of 25 : Experimental Research – in the Laboratory and Beyond

2 300

3 300

4 300

5 300

6 315

7 315

8 315

9 315

10 330

11 330

12 345

13 360

14 375

15 450

16 450

17 450

18 450

19 450

20 450

21 450

22 450

23 450

24 450

25 450

26 450

27 450

28 450

29 450

30 450

SAGE

SAGE Research Methods

Page 11 of 25 : Experimental Research – in the Laboratory and Beyond

31 450

32 450

33 450

34 450

35 450

36 450

37 450

38 450

39 450

40 450

Practice Point 3

Generally speaking, it is better to have a large sample if you intend to undertake statistical

analysis, as is usually the norm with experiments. You will want to be able to make some

claims that your sample represents the population from which it is drawn, and that your

results didn't occur by chance. If you only have five people in your sample, it is possible

that they could all be unusual in some way, and you wouldn't have achieved the same

results with another sample. If you have a sample of 50, it is likely that any individual

idiosyncrasies will be averaged out, and is even more likely with a larger sample.

A final point to consider before we move on to discuss data analysis is that psychologists have often

recruited participants from first-year psychology classes. This is convenient because psychology students are

reasonably likely to be interested in psychology experiments, and they are often offered an incentive such as

a credit of marks. What are some situations in which it may be desirable to recruit participants solely from this

group? What are some situations in which it would be problematic?

Analysing Experiments

Experimental data analysis is usually analysed using software such as SPSS. A thorough discussion of

statistical analysis is beyond the scope of this text. However, we will introduce you to some core concepts

regarding the analysis of experimental data. These basics include statistical significance, normal distribution

and standard deviation.

Statistical Significance

SAGE

SAGE Research Methods

Page 12 of 25 : Experimental Research – in the Laboratory and Beyond

In order to be able to understand research findings that you are likely to encounter in the course of your

studies, there are some key terms you should be familiar with. Chief among these is ‘statistical significance’.

This term refers to the likelihood that the results could merely have occurred by chance or coincidence. If

a result has occurred five or fewer times in 100 possible outcomes, then it is considered to be ‘statistically

significant’, i.e., it is unlikely to be a coincidence. This is often expressed in academic journal articles as a ‘p

value’: p <0.05 (the ‘p’ stands for probability, and 0.05 is another way of saying five in a hundred or 5%). The

more zeros after the decimal point, the more significant the result: 0.0005 is much more significant than 0.05.

However, a p value does not necessarily include a 5. In fact, although 0.05 (or 5%) is the general cut-off point

for significance, you are more likely to see p values of <0.01 (or 1%, which is more significant than 5%) and

<0.001 (which is the same as one in a thousand or 0.1%).

Normal Distribution (also Called the Bell Curve) and Standard Deviation

The normal distribution tells us how we would expect scores to be distributed, according to the mean

(average) score for the sample. A standard deviation is used to demonstrate how much a score deviates

(is different) from the mean. For example, imagine that the marks for your final examination have just been

posted, showing that the mean was 56.3%, and that you scored 82%. It seems likely that you did well in

comparison to your classmates, but you can't be sure unless you know what is within the normal range –

in other words, how much your score has deviated from the norm. It could be that lots of people got scores

above 80% (although this would also mean that a lot of people got very low scores since the mean is 56.3%).

Figure 7.2 illustrates how these issues may be represented graphically. To the right of the graph is some

basic information: the mean, the standard deviation, and the number of participants. You can see that your

score was more than one standard deviation from the mean (56.3 + 16.9 = 73), so was unusually high. In

fact, you are one of only three people out of the 42 who sat the exam to score above 80% (the vertical y axis

is labelled ‘Frequency’ and indicates how frequent each score was – i.e., how many students scored within

each range of marks – while the horizontal x axis shows the marks, with each bar representing a range of 10

marks, from 0–10, and so on). The line forming an upside-down U is the normal curve. We can see that the

marks attained reflect the normal curve fairly closely. A few more people scored in the 50–60% and 0–10%

ranges than the mean and normal curve would suggest, while the 30–40% and 40–50% are perhaps a little

low, but only slightly.

SAGE

SAGE Research Methods

Page 13 of 25 : Experimental Research – in the Laboratory and Beyond

Figure 7.2 Histogram Illustrating Standard Deviation and Distribution – Exam Results

Table 7.2 Frequency Table – Milgram's Experiment

Volts Frequency Percentage Valid percentage Cumulative percentage

300 volts 5 12.5 12.5 12.5

315 volts 4 10.0 10.0 22.5

330 volts 2 5.0 5.0 27.5

345 volts 1 2.5 2.5 30.0

360 volts 1 2.5 2.5 32.5

375 volts 1 2.5 2.5 35.0

450 volts 26 65.0 65.0 100.0

Total 40 100.0 100.0

Let's look again at our core example for this chapter. A simple calculation tells us that the mean shock level

SAGE

SAGE Research Methods

Page 14 of 25 : Experimental Research – in the Laboratory and Beyond

administered by the 40 participants in Milgram's original experiment was 405 (in the ‘Danger: Severe Shock’

range). Table 7.2 gives us some basic information on the frequency of administration of maximum shock

levels, while the graph in Figure 7.3 gives us some additional information, including the normal curve and the

standard deviation. Both have been exported from SPSS.

The x axis (the one that goes across) represents the level of volts. The y axis (the one that goes up)

represents the frequency at which participants stopped giving shocks – or, to put it another way, the number

of people/cases who stopped at each volt level. So by reading across to the first bar on the left we can see

that nobody stopped administering shocks before they reached 300 volts, but five refused to go on past that

point. (We can also see this in the first line of Table 7.2.) Then four people stopped at 315 volts, etc. If we look

across to the last bar on the right, we can see that 26 people stopped at the 450–volt level (the maximum).

The gently curving line across the graph is the normal curve; it tells us what we could expect to happen if

people conformed to the norm that is predicted by the mean – where we would have expected the tops of

the bars to reach. It looks quite different from the normal curve in the graph of exam results (Figure 7.2), in

which the normal curve is quite close to the tops of the bars. Given that the mean in Milgram's study is 405

volts, we would expect the highest bar to be located at the 400 point on the x axis, the next highest bars

to be immediately to the left and right at 380, and so on, with the lowest bars being at either end. Instead,

something interesting has happened. The actual height of the bars – the numbers of participants and the

maximum shocks they were prepared to give – does not conform to the normal curve at all. The highest bar is

at the extreme right, not the middle of the curve. This suggests that something very important has happened

in this experiment.

It is also interesting to note that no one stopped between 375 volts, which had the additional notation on the

machine of ‘Danger: Severe Shock’, and the maximum – people continued through 390, 405, 420 and 435

volts. Unfortunately, Milgram gives no explanation for this, but we will return to it later in the chapter.

SAGE

SAGE Research Methods

Page 15 of 25 : Experimental Research – in the Laboratory and Beyond

Figure 7.3 Histogram Illustrating Standard Deviation and Distribution – Milgram's Experiment

Some Issues in Research

The use of experimental research in the social sciences has been highly criticised. In this section we discuss

the key concerns: ethics and epistemology. We also consider the impact of the availability of resources.

Ethical Issues

Researchers have a moral and legal responsibility to abide by ethical principles, based on moral values.

Chief among these is the minimisation of harm, such that if there is a risk of distress, procedures must be

in place to prevent this occurring wherever possible, and if it should occur, it should be dealt with as quickly

and effectively as possible. To further minimise the risk of harm, ‘informed consent’ is considered paramount.

Every participant should be clear about what the research will involve before agreeing to take part, and

should be able to refuse to participate, or to stop participating at any time, without negative consequences.

SAGE

SAGE Research Methods

Page 16 of 25 : Experimental Research – in the Laboratory and Beyond

These strictures may seem quite sensible and straightforward at first glance, but consider how they apply

to Milgram's experiment, in which deception was an essential ingredient and participants were very strongly

encouraged to continue whenever they showed reluctance, with repeated exhortations to complete the

‘learning task’: ‘It is absolutely essential that you continue’ and ‘You have no other choice, you must continue’.

Indeed, Milgram used the phrase ‘bring the subject [participant] into line’ when discussing unwillingness

to continue (1963: 374). As you can imagine, the use of deception such as this has caused particular

concern in social psychology. Virtually every study now has to be evaluated for its ethical appropriateness by

experienced researchers before the study can be conducted. Debriefing is generally considered a necessity

whenever a deception has occurred as part of an experiment.

Debriefing refers to the process of discussing the research with the participants after completion of an

experiment (or interview, etc.). It allows the participant to ask questions about the procedure, and for the

researcher to explain, which is particularly important where deception has taken place. Milgram describes

‘dehoaxing’, or procedures to ensure that the participant would ‘leave the laboratory in a state of well-being.

A friendly reconciliation was arranged between the subject and the victim, and an effort was made to reduce

any tensions that arose as a result of the experiment’ (1963: 374). No doubt participants were greatly relieved

that they had not, in fact, inflicted pain and suffering on anyone. However, it seems reasonable to assume

that the experiment had ongoing consequences for at least some of the participants. How would you feel if

you had behaved in a way that seemed to be causing someone great pain, even endangering their life, and

you had continued to do so, despite their protests (even if you later found it had been a hoax)?

As mentioned above, moral values inform ethical practice. But do moral values affect science in areas other

than ethical issues? Should values affect scientific inquiry? Traditionally, (social) psychologists strived for an

objective, value-free approach to research. It was felt that one's personal values could impinge on research

in a negative manner, resulting in unconsciously leading participants to behave in ways that justify the

researcher's values or beliefs, or the misinterpretation of findings. However, can science be totally unbiased

and objective? The experiment is based on a positivist approach, and the positivist would likely say that

science can, and should, be value-free and objective. However, we can see that researchers' values are often

inextricably intertwined with their research. Milgram's research was based on his desire to contribute to our

understanding of some of the most reprehensible behaviour to ever occur. This is clearly based on a moral

value. Further, social research is often undertaken with a goal of ‘making the world a better place’. A major

figure in the social sciences of the 1930s to 1950s, Kurt Lewin, argued that research should be applied to

important, practical issues (Lewin, 1951). This in itself is a value-laden statement.

Despite a desire for scientific neutrality, our personal ethics and values often do contribute to the research

process, through the research topics we find interesting to the application of findings in the wider community,

whether through informing social policy or social service and health programmes. Many current researchers

would argue that it is better to recognise personal positions and consider how they may contribute to our

research.

SAGE

SAGE Research Methods

Page 17 of 25 : Experimental Research – in the Laboratory and Beyond

Conceptual Concern 5

The experiment is firmly rooted in the positivist tradition. Developing from the hard

sciences, experimenters attempt to apply scientific approaches to the understanding of

society. Theories are formulated and tested with reference to measurable, quantifiable

results.

While it appears to be unusual for experimenters also to use other research approaches, there is certainly the

potential to do so. Again, Milgram provides an example. As mentioned above, he gathered various types of

additional information, including notes of participants' behaviour, tape-recordings and photographs. Although

the use made of these is quite limited, some illuminating observations are reported. For example:

I observed a mature and initially poised businessman enter the laboratory smiling and confident.

Within 20 minutes he was reduced to a twitching, stuttering wreck, who was rapidly approaching a

point of collapse. … At one point he pushed his fist into his forehead and muttered: ‘Oh God, let's

stop it.’ And yet he continued to respond to every word of the experimenter, and obeyed to the end.

(Milgram, 1963: 377)

Milgram also quotes some subjects who did not continue to the maximum shock level (remembering that

everyone began at 15 volts and continued to 300 volts, with a minimum of 15 shocks administered):

‘I think he's trying to communicate, he's knocking. … Well it's not fair to shock the guy … these are

terrific volts. I don't think this is very humane. … Oh, I can't go on with this; no, this isn't right. It's a

hell of an experiment. The guy is suffering in there. No I don't want to go on. This is crazy.’ (Milgram,

1963: 376)

Yet think how interesting it would be if we had information on how participants felt after the experiment, what

their thoughts and feelings were, how they explained their obedience to themselves, both before and after

they were told they had been deceived. Further interviewing may also have helped us to understand why all

those who quit the experiment before the end did so between the 300–volt level (the top of the ‘Intense Shock’

range, at which point five people quit) and the 375–volt level. No one quit the experiment between 375 volts

and 450 volts (there were four shock levels in between, at which participants could have refused to go on, yet

no one did refuse). What was the reasoning of the individuals behind this distinct pattern? Similarly, we have

no information on the longer-term impact of taking part. Issues such as these led to experimental research

coming under increasing criticism from the 1950s to the 1970s.

The Crisis in Social Psychology: A Crisis in Positivist Epistemology?

The 1960s and early 1970s can be thought of as a time of questioning and consciousness-raising in western

society in general, engendering both optimism and expansion, and crisis and debate. This extended to social

SAGE

SAGE Research Methods

Page 18 of 25 : Experimental Research – in the Laboratory and Beyond

psychology too, in particular the appropriateness of the dominant research method, the experiment.

Hepburn (2003) conceptualises the crisis in social psychology as consisting of three areas:

• Critique of method.

• Critique of individualism.

• Theoretical critique.

Together, they can be understood as a critique of the positivist standpoint on which experimental research

was based.

The experimental/laboratory approach had been dominant in social psychology for decades, but by the 1950s

it already had critics. These criticisms increased in the 1960s, particularly around the inadequacy of laboratory

methods to understand cultural and historical significance and motivations. In addition, the experiment was

criticised as mechanistic: the study of cause and effect treats people as machines without capacity for plans

and options, while the artificiality of the laboratory meant (it was argued) that findings cannot be extrapolated

to real life (this is why experimental realism is important). In some cases, there are only a certain number

of possible responses. It is assumed that the researcher knows all the possible options and therefore the

possibilities for alternatives are negated. Kenneth Ring stated that social psychological research treated

important social issues as fun-and-games experiments (Ring, 1967).

In regards to individualism, concern was raised that social psychology was focused on social aspects of the

individual, ignoring how the social environment impacts the individual. This is a one-sided view of human

behaviour. For example, Lewin (1951) developed a set of fundamental principles for social psychology:

(social) behaviour depends to a large extent on how we perceive and interpret the world around us; it is

a function of the interaction between the person and the environment; social psychological theories should

be applied to important, practical issues. His concluding statement may be seen in part as an indictment of

the constraints of the prevailing research method: ‘No research without action, no action without research’

(Lewin, 1951: 193), while his emphasis on the interaction between the person and his or her social world

points to the need to consider the social context as much as the individual. In Milgram's obedience experiment

we can learn about a specific aspect of social behaviour in a specific setting – obedience in a laboratory.

However, the broader social and historical context is ignored and the ability to use the findings for social

action is limited. For example, we learn nothing about the impact of upbringing, or ethnicity, on participants'

behaviour, or how the participants felt about their behaviour, either during or after the experiment. Nor do we

learn anything about the 35% of people who did NOT continue administering ‘electric shocks’ to the maximum

level. When the unit of focus in research is the individual, it is likely that solutions to social issues will be

focused on changing individuals, such as behaviour, personality, or moral reasoning, rather than considering

social structures and their impacts, such as inequality, income, (un)employment and social influences.

Traditional psychology places little value on theory, beyond a rather narrow focus on the specific topic. This

lack of focus on wider applications of research was seen increasingly in the 1960s and 1970s as the result of

SAGE

SAGE Research Methods

Page 19 of 25 : Experimental Research – in the Laboratory and Beyond

an inappropriate emphasis on positivist epistemology. (We discuss epistemology in more depth in a number

of chapters, particularly in Chapter 1.) The crisis in social psychology was evidence of a broader challenge

to positivism and the positivist emphasis on observation and measurement, and the potential for scientists to

form objective/unbiased understandings of the social world. In contrast, a number of approaches drawn from

social realism and social constructivist epistemologies were advanced by a disparate group of critics. Thus

phenomenology, ethnomethodology, structuralism, Marxism, feminism, and other critical theories offered new

conceptions which claimed to be more appropriate for understanding social psychological behaviour and

society and for improving them.

A major European social psychological work published at this time was Serge Moscovici's The context of

social psychology (1972). Moscovici criticises the individualist and capitalist assumptions of traditional social

psychology, arguing that they reinforce the political status quo by treating it as universal and inevitable. In

contrast, Moscovici argued from a social realist stance that:

• Society both individualises and socialises: it teaches us how to behave.

• Behaviours only acquire their meaning according to the responses received and our interpretations

of these responses, and this is regulated by our common history of norms and rules.

• Therefore, the study of social psychology must include the broader social context.

The ‘crisis’ strengthened psychology as a discipline, and furthered the rise of social psychology, particularly

through the adoption of pluralism: that is, the acceptance of many approaches to research in addition to

the laboratory experiment and the development of international and multicultural perspectives (Kassin et al.,

2008). However, positivism has proven to be more robust in psychology than its sister disciplines in the social

sciences. A cursory glance at almost any introductory psychology or even social psychology textbook will

reveal that experimental research still dominates.

Practice Point 4

The resources required to conduct an experiment can vary enormously, from almost

nothing but the experimenter's time, to extremely expensive equipment and the time of a

technician and a multi-person research team.

Resourcing

Many experiments require little in the way of resources, beyond any single researcher's time and some office

space. Consider the recent variations of the Lost Letter Technique, using email. In these experiments all

that would be required is time and a computer with an email connection and statistical analysis software.

In an experiment one of the authors participated in while an undergraduate student, all that was required

was an office and a large bag of condoms (don't be alarmed – the experiment was on embarrassment and

SAGE

SAGE Research Methods

Page 20 of 25 : Experimental Research – in the Laboratory and Beyond

http://methods.sagepub.com/book/social-research-a-practical-introduction/i167.xml

nothing untoward happened). In contrast, Milgram's original obedience experiment required rather more: a

confederate (the ‘learner’), three rooms beside each other (the room in which the ‘learner’ sits, the room with

the shock machine and, on the other side of one-way glass, an observation room), the experimenter himself,

at least one observer, the ‘shock machine’ and associated equipment, and recording equipment. As new

technology becomes available, the potential for more complex and expensive experiments increases.

Putting the Approach in Context

While the availability of new technologies has certainly allowed experimenters to undertake innovative and

complex work, it has also standardised the approach in some respects, particularly in regard to data analysis.

In this section we consider the impact of new technology as both an aid to the development of experiments

and as the impetus behind new methodological orthodoxies.

Practice Point 5

Experimenters have a history of adapting experiments and taking advantage of new

technology as it becomes available. Milgram's ‘shock machine’ is an example of the use

of the available technology in the 1950s being utilised, while his Lost Letter Technique has

been adapted by others using electronic mail. Another example follows.

In an experiment on prejudice, participants played a video game in which a person popped up unpredictably,

holding either a gun or another object (Correll et al., 2002). Sometimes the individual was white, sometimes

African American. Participants were to ‘shoot’ the individual if he was holding a gun, or to press a different

button if he was holding another object – and in either situation to respond as quickly as possible. Participants

made mistakes equally often when the target was white – shooting at an unarmed target and not shooting the

armed target. However, they were significantly more likely to shoot the unarmed African American target, and

significantly less likely to fail to shoot the armed African American target.

In similar experiments that didn't involve video games (such as showing photographs: Judd, Blair and

Chapleau (2004); Payne (2001) a tendency to associate weapon use with African Americans was clearly

displayed. Note that while these experiments do appear to indicate a tendency towards stereotyping African

Americans as more likely to be dangerous, there is an important aspect of a ‘good’ experiment missing –

realism. This lack of realism raises issues of generalisability, as discussed above. While participants were

willing to ‘shoot people’ while playing, they were fully aware that they weren't really harming anyone and it is

highly unlikely that they would so willingly shoot people in real life. While these experiments appear to tell us

something about the frequency of prejudicial thinking, we cannot conclude anything about racist actions (in

addition, they tell us nothing about the participants' underlying beliefs or attitudes, so are not a useful source

of information for counteracting prejudice). On the other hand, experimental realism was an integral part of

SAGE

SAGE Research Methods

Page 21 of 25 : Experimental Research – in the Laboratory and Beyond

Milgram's electric shock experiment, but carried with it some major ethical concerns. Here are some points to

ponder regarding Correll, Park, Judd and Wittenbrink's (2002) work:

• How might these experiments be adapted to heighten experimental realism – that is, to encourage

the sense of a ‘real life’ situation rather than a game?

• What would be the ethical concerns with doing so?

• How might these ethical concerns be addressed?

We can see that the availability of computers and other electronic equipment has allowed experimenters to

both update old experiments and develop new ones. The use of video games in experiments probably also

adds an additional element of fun, and may well assist in the recruitment of participants through making the

experience seem attractive.

An undeniable advantage of new technology is the availability of software that allows us to quickly and

accurately undertake statistical analysis. Once data is entered, tables and graphs showing results can be

produced in a matter of seconds – work which once would have taken hours or days. Note that the results

shown in this chapter, including frequencies, norms and standard deviations, are merely the ‘tip of the iceberg’

in terms of the analysis it is possible to do. For example, if an experiment uses a control group and an

experimental group, it would be usual to conduct an ‘analysis of variance’ to determine if any differences

between the two groups are significant (are unlikely to have occurred by chance) or one may perform cross-

tabulation analysis using chi-square to see if there were differences according to subject variables, such as

gender. Such analyses are now considered standard, even necessary.

However, it is interesting to note that until fairly recently many of these analyses were not undertaken. The

paper reporting the results of Milgram's Lost Letter Technique includes the percentages of letters returned

only, broken down by the four addresses on the envelopes. We are left to draw our own conclusions about

the fact that 25% of the letters addressed to ‘Friends of the Communist Party’, and to ‘Friends of the Nazi

Party’ were posted, while 72% of the letters posted to ‘Medical Research Associates’ were posted. We are not

told if these numbers are in any way significant. Similarly, Milgram tells us only the numbers of participants

who stopped administering shocks at each level – he does not even give us percentages. He does tell us

how many participants are in each of the three age groups, and in each of the three occupational groupings,

but he does not tell us if there are any differences between the groups in terms of shocks administered. Of

course, given that the sample is relatively small, if resulting differences are small it could simply be due to

coincidence, BUT differences between the groups could be large (and therefore probably significant) – we

simply do not know. We suspect that if a researcher were to submit a manuscript for publication that contained

so little analysis now, it would be rejected. We suggest that the possibility of undertaking complex analysis

quickly and easily has rendered it a necessity.

Conclusion

SAGE

SAGE Research Methods

Page 22 of 25 : Experimental Research – in the Laboratory and Beyond

Experimental research has a somewhat checkered history in the social sciences – arguably it is over-utilised

in social psychology and under-utilised in the other disciplines. Certainly, it has much to offer in terms of

discovering cause-and-effect relationships and uncovering behaviour that may be difficult to investigate using

other means. After all, if most people were asked ‘If I put you in a room with an electric shock machine …

(etc.) … would you give the “learner” a 450 volt shock that could seriously harm him or her?’, we suspect they

would say ‘No!’, yet we know that in some circumstances that is precisely what the majority of people will do.

The experiment is a variable-centric approach and fixed (once the experimental protocol has been decided

upon and experiments begun, it is problematic to change it), so careful planning is essential. Experiments,

like all research, can vary in their efficacy. In order to conduct a successful experiment, we must have

clearly defined variables – it is a variable-centric, fixed approach, based on the positivist perspective, so it

is imperative that variables do measure the thing we want them to measure – and a high level of realism,

especially if deception is involved.

A well-designed experiment can tell us much about individual behaviour in specific circumstances. In addition,

the high level of control the experimenter has and the straightforward analysis process means that it is a

straightforward approach for emergent researchers.

However, experiments alone tell us little about participants' underlying thoughts and motivations, and nothing

at all about the social context that lead to the development of those thoughts and motivations. Indeed, some

of these experiments remind us of Ring's comment about trivialising important issues to ‘fun and games’

experiments. We may learn that people continue to hold racial prejudices, but experiments such as these,

unless supplemented with other approaches, may be of limited practical use. Of course, one may argue

that knowing that people hold prejudices (for example) is important for its own sake. Nevertheless, some

researchers will want to apply their findings in practical ways – to respond to Lewin's call for ‘action’. To do

so, the application of other approaches, perhaps incorporated in a debriefing session, are required.