Mortality is the only plausible threat to internal validity for TEs.
All OGPTPT threats except for mortality can’t be a threat to TEs because they can’t explain
differences among groups.
With internal validity, the issue is not whether the process has occurred. It’s about whether the
process could have led to results obtained.
In TE: 1.The conditions are being run In parallel, at the same time. 2. Participants are run
individually or in small or mixed groups (not all the time, though). 3. Ps in the conditions
experience the same things except for the things being manipulated.
1. Creates condition that is equal before the manipulation.
2. Equal in terms of individual differences (related or not related to the research).
All Static Group design threats are implausible:
1. Selection: All pre-existing characteristics were ruled out by Random Assignment.
2. Ambiguity: The IV was manipulated which secures that the IV comes before the DV.
3. Local History: Plausible in Intact groups; implausible in mixed groups and individuals.
1. Can examine pretest scores to make sure RA works at least for that variable.
2. Makes statistical test more sensitive – more likely to show effects
1. Sensitizing effect: Cues to the Ps what the Research is all about.
2. Anchoring effect: I will answer the same in the post test.
With POST TEST Design LOGIC:
1. Researchers could assume that participants started out as equal (because of Random
2. If they differe later on, it must be due to manipulation.
ADVANTAGES OF TRUE EXPERIMENTS:
1. Best design for addressing causality because they rule out most threats to internal validity.
2. It has the power to detect effects because manipulation can control research situations,
reduces research errors and can increase research sensitivity.
LIMITATIONS OF TRUE EXPERIMENTS:
1. Can’t rule out Mortality, Diffusion or imitations of treatments (when the Ps with chocolates
shared their chocolates to others), and Resentful demoralization (was it the chocolate that
really makes the Ps happy or was it the resentment from the ones who does not receive
2. Can’t explain some important social problems experimentally (can’t bring poverty to the lab)
3. External validity issues (TE in general raises more external validity issues than correlational
studies in the field) 4. Construct Validity Issues – Confounds: some additional variables that varies systematically
with the IV except from the DV (i.e., internal validity threats)
- Did we manipulate what we really wanted to? And only what we really wanted to?
- Is it what we think it is about the manipulation that had the effect?
Distinction between Internal validity and Construct validity:
- Internal Validity question: Did the operationalization of the IV lead changes in the
measure of the DV?
- Yet we still have questions about: what was abt the predictor (operationalization)
variable that has an effect? (issues of construct validity)
- True experiments ruled out almost all threats to internal validity but does not rule out
all threats to construct validity
THREE REQUIREMENTS FOR CAUSALITY IN TRUE EXPERIMENTS:
2. Temporal order
- to determine which comes first, must measure DV in time 1 too; cross lagged panel
- When only one direction is plausible; when the measures are given more than one time
– these can show temporal order but not causality
- Present in true experiments because researchers manipulated the IV