Understanding Experiments
➔ internal validityis the extent to which a causalrelationship existing between two or more
variables can be assumed
◆ IV changes the DV
➔ conditions of causality
◆ covariance: as one variable changes, so does the other
◆ temporal precedence: establish direction of causality
◆ independent variables: IV causes the DV to change(direction matters)
➔ threats to internal validity
◆ selection effects: when groups differ from one anotherin more ways than just the
treatment/IV of interest
◆ confound: a third variable systemically changes theIV/DV, leading to alternative
explanation for the results
◆ false experiment: design that seems like an experimentbut lacks defining features,
making it ultimately correlational
● one-group posttest only design
○ treatment or intervention is implemented within a group of participants
and the DV is measured after the treatment, no control/comparison
group
○ results are confounded because too many factors that could cause the
outcome, other than the treatment
● one group pretest-posttest only design
○ DV is measured before and after implementation of a treatment within a
single group of participants but no separate control/comparison group
○ confounded results because too many factors that could change between
administration of treatment/control and the two outcome measures
● non-equivalent comparison groups
○ group of participants that receive treatment are compared to group that
didn’t receive treatment, but the groups were not randomly assigned to
the conditions
○ confounded results because the groups themselves could differ in
various ways other than the IV
◆ history threat: other events may have happened betweenpretest and posttest that
would explain the outcome
◆ maturation threat: participant themselves may havechanged between pretest posttest
◆ testing threat: participant improves due to priorexposure during pretest
◆ instrumentation threat: outcome measure might changebetween pretest posttest
➔ to avoid these threats to internal validity, useexperimentaldesignswhich meets all 3
conditions
◆ covariance- manipulating conditions of IV allow forclearer comparison of outcomes
, ◆ temporal precedence- directionality of effect is established by manipulating the IV
before measuring the DV
◆ rule out alternative explanations- control for alternativeexplanations
● random assignment
● holding other factors constant
➔ treatment vs control conditions
◆ treatment condition: intervention meant to changepeople’s behaviours
● compared to the control condition
◆ no-treatment control condition: no treatment or interventionat all
◆ placebo control condition: simulated “treatment” thatlacks active elements of the IV
● participants think they’re getting a treatment
◆ treatment control condition: control condition receivesthe standard treatment or an
alternative treatment
➔ ‘active’ control conditions
◆ placebo effect- participants get better based onthe expectation of receiving the
treatment and it being effective
◆ reactivity- participants change b/c aware of beingwatched
◆ demand characteristics- participants may pick upon subtle cues of how they “should”
respond
◆ active control conditions- ensure each conditionhas equal expectations of receiving
treatment and receives equal attention from the researchers
➔ researcher bias
◆ observer bias- expectation influence interpretationsof the observations
◆ experimenter expectancy effect- expectations influencebehaviour towards participants
◆ solutionto researcher bias would be to use a double-maskeddesign
● neither the researcher nor the participant are aware of the who is in which
experimental design
➔ methods forrandom assignment
◆ simple random assignment
● random method of assigning participants to conditions, everyone has equal
chance of being in each group
● advantage: control for selection effects and unknown confounds
● disadvantage: can fail if sample sizes are small; could result in uneven sample
sizes across conditions
◆ block randomization
● chunking participants off into the conditions
○ 1,2,3,1,2,3,1,2,3….
● advantage: ensure equal sample size
● disadvantage: fail if sample size is small; must ensure recurring blocks are
random and unbiased
◆ matched-group design
➔ internal validityis the extent to which a causalrelationship existing between two or more
variables can be assumed
◆ IV changes the DV
➔ conditions of causality
◆ covariance: as one variable changes, so does the other
◆ temporal precedence: establish direction of causality
◆ independent variables: IV causes the DV to change(direction matters)
➔ threats to internal validity
◆ selection effects: when groups differ from one anotherin more ways than just the
treatment/IV of interest
◆ confound: a third variable systemically changes theIV/DV, leading to alternative
explanation for the results
◆ false experiment: design that seems like an experimentbut lacks defining features,
making it ultimately correlational
● one-group posttest only design
○ treatment or intervention is implemented within a group of participants
and the DV is measured after the treatment, no control/comparison
group
○ results are confounded because too many factors that could cause the
outcome, other than the treatment
● one group pretest-posttest only design
○ DV is measured before and after implementation of a treatment within a
single group of participants but no separate control/comparison group
○ confounded results because too many factors that could change between
administration of treatment/control and the two outcome measures
● non-equivalent comparison groups
○ group of participants that receive treatment are compared to group that
didn’t receive treatment, but the groups were not randomly assigned to
the conditions
○ confounded results because the groups themselves could differ in
various ways other than the IV
◆ history threat: other events may have happened betweenpretest and posttest that
would explain the outcome
◆ maturation threat: participant themselves may havechanged between pretest posttest
◆ testing threat: participant improves due to priorexposure during pretest
◆ instrumentation threat: outcome measure might changebetween pretest posttest
➔ to avoid these threats to internal validity, useexperimentaldesignswhich meets all 3
conditions
◆ covariance- manipulating conditions of IV allow forclearer comparison of outcomes
, ◆ temporal precedence- directionality of effect is established by manipulating the IV
before measuring the DV
◆ rule out alternative explanations- control for alternativeexplanations
● random assignment
● holding other factors constant
➔ treatment vs control conditions
◆ treatment condition: intervention meant to changepeople’s behaviours
● compared to the control condition
◆ no-treatment control condition: no treatment or interventionat all
◆ placebo control condition: simulated “treatment” thatlacks active elements of the IV
● participants think they’re getting a treatment
◆ treatment control condition: control condition receivesthe standard treatment or an
alternative treatment
➔ ‘active’ control conditions
◆ placebo effect- participants get better based onthe expectation of receiving the
treatment and it being effective
◆ reactivity- participants change b/c aware of beingwatched
◆ demand characteristics- participants may pick upon subtle cues of how they “should”
respond
◆ active control conditions- ensure each conditionhas equal expectations of receiving
treatment and receives equal attention from the researchers
➔ researcher bias
◆ observer bias- expectation influence interpretationsof the observations
◆ experimenter expectancy effect- expectations influencebehaviour towards participants
◆ solutionto researcher bias would be to use a double-maskeddesign
● neither the researcher nor the participant are aware of the who is in which
experimental design
➔ methods forrandom assignment
◆ simple random assignment
● random method of assigning participants to conditions, everyone has equal
chance of being in each group
● advantage: control for selection effects and unknown confounds
● disadvantage: can fail if sample sizes are small; could result in uneven sample
sizes across conditions
◆ block randomization
● chunking participants off into the conditions
○ 1,2,3,1,2,3,1,2,3….
● advantage: ensure equal sample size
● disadvantage: fail if sample size is small; must ensure recurring blocks are
random and unbiased
◆ matched-group design
