I. Exact (direct) replications

A. To see if original study victimized by chance

1. Type 1 errors:

a. Inevitable, and may be more likely than we think because null results are rarely submitted to journals

b. The "Sleeper effect", ESP and other examples

2. Type 2 errors

B. To prevent fudging:

1. Threat of replication discourages fraud

2. Examples of replication stopping fraud:

a. Operant conditioning of autonomic responses in paralyzed rats

b. Breunig's research on the effects of stimulants on mentally retarded patients

C. Generalize results

1. To present day

2. To your population

II. Systematic replications

A. For same reasons as doing an exact replication plus:

B. Improve external validity:

1. Other kinds of subjects (females rather than males, as 3X as many studies have all male subjects as have all female subjects), noncollege students, humans rather than rats

2. Other settings (field vs. lab)

3. Generalize to different levels of treatment variable

C. Construct validity:

1. Double blind

2. Placebo condition

3. Between rather than within design

4. Use multiple levels of treatment to make it more difficult for subject expectancies to account for results

5. Greater standardization to prevent experimenter bias

6. Increasing experimental realism

D. Power:

1. More participants

2. Participants who are more homogeneous

3. More standardized procedures

4. More reliable dependent variable

5. Stronger manipulation of treatment variable

6. Within-subjects design

III. Conceptual replications--because no operational definition is perfect

A. Using different treatment variable

B. Using different measure

IV. Summary: The value of replication

Lecture 4.2
Back to Chapter Main Menu