There are many different methods of research and we’ll be covering some of the more common

ones. There is no method that is superior to another and they all have advantages and
disadvantages. The method chosen depends upon practical and ethical considerations and the
type of question you’re asking.

INTRODUCTION TO EXPERIMENTAL RESEARCH –

CHAPTER 5
• In an experiment the researcher manipulates certain variables and holds others constant
(ie controls them)
• Advantage: you can infer a cause and effect relationship
• Disadvantage: because experiments tend to be conducted in artificial environments (lab)
results might not be generalizable.
• It can involve basic or applied and be conducted in the field or lab

INDEPENDENT VARIABLE (I.V.):

- the variable that’s manipulated or changed to see if it has an effect on behavior.
-in other words, the situation that the experimenter creates
-there must be a minimum of 2 levels (the groups or conditions that are being compared)
Note: the terms “level of the IV”, “group” and “condition” can be used interchangeably.

DEPENDENT VARIABLE (D.V.):

-the behavior that’s measured to see if the IV has an effect
• Think of the IV as the “cause” and the DV as the “effect”
• The DV is dependent upon the IV
• There will typically be more than one IV and DV in an experiment
Example: Males aged 20, 25 and 30yrs were given a memory test consisting of 20 words to be
recalled to see if memory got worse with age.
I.V. = age only (3 levels)
Sex and list length are not IVs because they were not manipulated (they were held constant). If
females were also included and if different list lengths were compared, then they would be IVs.
DV= memory score

1
EXTRANEOUS VARIABLE:
-variables that are of no interest to the experimenter, but that could affect behavior. As long as
they are controlled for or held constant (all groups have the same exposure to the variable), it’s
not a problem.
Study time Grade
0 hrs 60%
5 hrs 74%
10 hrs 62%

CONFOUNDING VARIABLE:
-an extraneous variable that’s not controlled (held constant) and that changes along with the IV
(ie, the extraneous variable is not equal across groups)
- a confounding variable is something that “interferes” with your study making it impossible to
determine if it’s the IV or the confounding variable that’s the cause for the DV (only the IV is
supposed to be the “cause”).
Study time Grade
normal sleep 0 hrs 60%
normal sleep 5 hrs 74%
no sleep 10 hrs 62%

Ex. Suppose you’re testing the effects of study time on grade. Sleep could be something that
affects grade (it’s an extraneous variable) but it’s not a problem as long as the average amount of
sleep that subjects (S) got across ALL study time groups is constant. However, if everyone in
the 10 hr study group got no sleep while the 0 hr and 5 hr study groups got a normal amount of
sleep, sleep now becomes a confounding variable because it changes with the IV. In other words,
sleep isn’t held constant across levels of study time so you don’t know whether it’s the study
time or the sleep that’s the cause for the grade.
You can resolve this problem by ensuring that all 3 groups get roughly the same amount of sleep.

The INDEPENDENT VARIABLE can be either a MANIPULATED VARIABLE or a

SUBJECT VARIABLE
MANIPULATED VARIABLE: this is when the experimenter actually creates the situation that
the S encounters and Ss can theoretically be randomly assigned to the levels of the IV.(example
– alcohol dosage, how long you’re told to study for a test etc)
SUBJECT VARIABLE: refers to already-existing characteristics of subjects. You can’t
manipulate a subject variable and you can’t randomly assign Ss to these levels of the IV. Rather,
Ss must be selected. (examples- age, IQ, personality traits, gender, psych disorder). You can’t
manipulate someone’s age or IQ for example.

2
When a subject variable is present, you have a quasi-experiment (not a true experiment).
Drawing Conclusions:
If you have a manipulated IV can you infer cause and effect (as long as there are no flaws in the
study) but with a subject IV you cannot infer causation. All you can say is that the groups
performed differently.
A given variable could be a subject or a manipulated variable depending upon how it’s used in
the study. Example – people’s natural level of anxiety would be a subject variable, but
manipulating a person’s anxiety level by telling them that a test is easy or hard, would act as a
manipulated variable.

VARIETIES of MANIPULATED INDEPENDENT VARIABLES

All manipulated variables can be classified into one of three general types:
Task Variables - subjects are given different types of tasks to perform
Ex, one group might have to memorize nouns, whereas another memorizes verbs.
Instructional Variables - subjects are given a different set of instructions on how they should
perform the task.
Ex, one group of subjects might be told to memorize a list of words by simply repeating the
words (rote memorization) whereas another group is told to create a mental picture of the words.
Situational Variables - subjects encounter different environmental circumstances.
Ex, subjects complete the task in either a lab or in a classroom.

CONTROL GROUP
-the group that gets the “0” level of the IV or the group that doesn’t get the “treatment”
-this group is used as the basis for comparison
EXPERIMENTAL GROUP
-the group that gets the “treatment” and whose performance is compared to the control group.
The control and experimental groups should be similar to one another in every way except for
the IV. That way, if differences exist between the groups, you know that the difference must
caused by the IV and not by some other uncontrolled variable – ie, a confounding variable.
Ex. – In comparing the effects of 0, 2, and 4 oz of alcohol on memory, the 0oz group is the
control group and the 2 and 4 oz groups are experimental groups.
It’s not always possible to have a true control group – ie you can’t have a “0” amount of age or
gender or intelligence etc
4 TYPES of VALIDITY in an EXPERIMENT:
Statistical conclusion validity – stats are properly used
Construct validity – how good the measures are for the IV and DV; how representative are they
of the construct (a construct is something we assume exists even if we can’t directly see it).

3
External validity – the degree to which results generalize to other populations, times,
environments. Ecological validity refers to whether the study has relevance to real-life situations.
Internal validity – the degree to which the study has no flaws, and thus, we can infer causation

Things that can threaten internal validity:

-if you have a confounding variable
-if you don’t have a true control group
-if the IV is a subject variable
-when a S is measured repeatedly resulting in a practice effect (to be discussed later)
-if you have a Pre-Post Study
PRE-POST STUDIES (Studies Extending Over Time):
-tends to be used in studies that extend over a period of time (like when you want to determine
the effectiveness of a therapy or treatment).
-if you wanted to determine the effectiveness of a therapy, you could randomly assign patients to
a control group (no therapy) or the experimental group (receives the therapy) and then compare
the two groups.
Alternatively, you could first give people a “pre-test” (before therapy), they would then undergo
the therapy over a given time period, and then give the same people a “post-test” (after therapy).
You would then compare their pre-test score to their post-test score to see if the therapy was
effective (a control group is not typically used).
Problem: factors other than the therapy could have influenced post-test scores and so, results
could be due to a confounding variable rather than due to the treatment.
Ex. IF a control condition was used to determine if treatment successfully reduced anxiety:
Pre-test Treatment Post-test
80 65 (treatment is effective)

Pre-test ---------- Post-test (control condition)

80 80 (you would expect no difference)

What factors could threaten internal validity in pre-post studies (ie, could act as a
confounding variable)?
1) History -an event outside of the study produces an effect on post-test scores.
2) Maturation – post-test scores could be due to a subject’s natural maturation that occurred
between pre-post measures, and not due to the treatment per se.
3) Regression to the Mean – When Ss are selected because they score extremely high or low on
some characteristic, their scores tend to change in the direction of the mean when they’re re-
tested. Ex – a very high pre-test score is likely to become lower in the post-test, and a very low
pre-test score is likely to become higher in the post-test (independent of actual treatment effects).

4
4) Testing and Instrumentation – post-test scores could be influenced by the practice gained
by first doing the pre-test. Instrumentation could be a problem if the measurement tool used for
the pre-test is slightly different from the one used for the post-test (a different version was used,
for example).

THREATS to INTERNAL VALIDITY – SUBJECT PROBLEMS

A) Subject-Selection Effects – can occur when subjects cannot be randomly assigned to groups.
For ex, if you wanted to determine if lecture courses lead to higher grades than on-line courses,
students have already enrolled in these courses so you aren’t randomly assigning them.
Consequently, if grades are higher in lecture courses, you don’t know whether it’s due to the
course-type or whether it’s due to the type of student who chooses a lecture course.
B) Attrition – refers to students “dropping out” of the study before it’s completed. Those
dropping out may not be equivalent to those finishing it, so you might be left with a “specific
type of person” in your study (confounding variable).