Experimental research designs
- Aresearch strategy that attempts to establish the existence of a cause-and-effect
relationship between two variables by manipulating one variable while measuring the
second variable and controlling all other variables
• Rubin (1970) – Relation between love and gazing
• Davis et al. (2001) – Ternus task and dyslexia
• Rosenthal and Lawson (1964) – “Dull and bright” rats
• Gauthier et al. (1999) – Prosopagnosia study
- Science is a method of knowledge acquisition that:
• Addresses empirically solvable problems;
• Uses systematic empiricism;
• Allows its findings to be publicly verified.
- Experimental research designs are the pinnacle of systematic empiricism
Designing an experimental study
- Research question. Does time spent studying lead to better academic performance?
- Does time spent studying cause better academic performance?
• What is “time spent studying”?
• What is “academic performance”?
Operationalized constructs and hypothesis
- Academic performance. Ability to recall 50 definitions in Gravetter and Forzano (2009)
• External validity
• Case study
• Sample - Time spent studying.
• 2 hours
• 4 hours
- Hypothesis. Participants who study for 4 hours will recall more definitions on a test than
participants who study for 2 hours.
“But it’s not real life” revisited
- Experimental psychologists know that it is better to have “limited” causal explanations,
than generalizable, but vague explanations.
Increases Internal validity
Decreases External validity
PSYC 2001 – Introduction to Research Methods 1
- The variable manipulated by the researcher. The variable controlled to ascertain cause
- The thing you’re going to control, vary, manipulate to see its cause on the dependent
The levels of the independent variable
- The number of treatment conditions to which participants are exposed.
Dependent variable - The variable that is observed for changes in order to assess the effects of manipulating the
Extraneous and potential confounding variables
- How can we guarantee that other variables are not causing changes in the grades?
- Aprocedure in which a random process is used to assign participants to treatment
- Must randomly assign participants to conditions
- Idea is that there will be all types of conditions within each group, they are no longer
- The assignment of individuals to groups so that a specific variable is balanced or matched
across the groups.
- It is possible to mathematically remove the contribution of a variable using anAnalysis of
- You enter an IQ score for everyone, and remove the IQ... leave the difference in study
- How can we guarantee that environmental variables are not causing changes in the
Control - We make the testing conditions across participants as similar as possible.
- If 4 hours of study does lead to better results than 2 hours, then we can say that study
time does cause better performance (independently of other variables).
- This will be true independently of other variables
Four important questions
- How do experimental designs deal with the directionality problem?
• One is free to vary, but the other is controlled
• Directionality problem is when you don’t know what caused what. By having
independent and dependent variables, you are only controlling one and you can see
how the changes affect the second variable.
- How do experimental designs deal with the third-variable problem?
• Third variable is when you’re not sure if independent variable is actually changing
the dependent variable, and that it is being caused by a third variable. this design
eliminates this problem because of all the control placed on experiment eliminates as
much as possible that there is a third variable causing the changes
- Do experimental designs guarantee that the most significant independent variable(s) have
been selected for study?
• No, all we can guarantee is that it does seem to have a causal effect if it has a
- Do experimental designs guarantee that all threats to internal validity have been
• No, no experimental design can be 100% factual. There could always be something
affecting things somehow
• No because it is impossible to run a perfect experiment
Two basic experimental designs - For many experiments, the researcher has a choice. Will participants be tested in only one
condition or will they be tested in all conditions?
- ½ go in 1 group, the other ½ goes in the other
- This is called a between-group design *****????
- Every participant gets every level of the independent variable
Two basic experimental designs
- Between-subjects or independent-measures experimental design: An experimental
design using separate, independent groups of individuals for each treatment condition
- Independent Groups design: every participant gets a single level of a single variable
- T-value: statistic you’re testing to see if its significant or not
- P-value: the smaller the better
• Looking at the probability that the results are due to chance
Advantages and inconveniences of between-subjects designs
• No practice effects
If you are doing 1 level of independent variable... you are doing it once
• May limit fatigue or boredom effects (because you are only doing 1 condition)
• Limited power (requires a larger number of participants than other designs)
• If you only have 2 sources of errors ***** it makes the test less .... *****
• More expensive – more labor intensive
Using more than two levels of an independent variable - Theoretical motivation leads researchers to create designs with more than two levels of a
given independent variable (1, 2, or 3 hours)
Using more than two levels of an independent variable
1 hour 2 hours 3 hours
Hours of study
- Do we use t-tests to compare 1 with 2, 1 with 3, and 2 with 3?
- To avoid Type I error (claiming that there is a significant difference when there isn’t),
Analyses of Variance (ANOVAs) are used.
- When you run inferential statistics, you increase