Social Psychology I 18/09/2013 2:51:00 PM
With experiments, you can determine causality; with correlational
studies, you cannot
Correlational research is valuable – used for prediction (what’s
going to happen neck) but will not know why or how
Does x cause y? (x->y)
There are different ways to explain causality (x->y, y->x, or z->xy)
Manipulate independent variable (IV): cause
o Take control and manipulate variable (change levels,
o You want to know whether alcohol has an effect on memory
for social psychological experiments
o Ex. reading textbook, drinking wine or grape juice – also how
much? One cup? Four?
Evaluate effect on dependent variable (DV): effect
o Effect on memory – what do you remember studying?
o Ex. People who drank four glasses of wine remembered less?
Need random assignment!
o Put people in different groups – those eating chocolate, those
forced to chain smoke, etc.
o Everyone has equal possibility of being in either group
o Different than random sample! – random sample is when
everyone has equal possibility of being part of the experiment
o Random assignment is a “great equalizer”
Not based on personality aspects or height//weight/sex
Individual differences average out!
What does it mean if X causes Y?
If someone tells us smoking causes lung cancer, what does it
mean? Does it mean EVERYONE who has ever smoked gets lung
cancer? Does it mean people who don’t smoke cannot get lung
Causal influences on Dr. Robinson’s Happiness o Chocolate, class participation, watching first dates, high class
average, getting paid, other (variety)
If you give her chocolate she will be happy, but if she
hasn’t slept then her happiness won’t improve as much
Control sleep, test in experimental way
Valid/reliable/consistent influence on what will occur
o Goal is to account for as much variation in happiness as we
possibly can – identify more and more variables
o There are always going to be unidentified variables (error) –
things we haven’t assessed or predicted
o IV coin flip
Heads first group, tails second group then assess an
Manipulation I = heads, Manipulation II = tails;
outcome = DV
Ex. Hypothesis -> similarity causes liking
o Reading info on others
o Liking = desire to spend time with other person (operational
o Importance of control groups for conclusions
Comparison to experimental group
Tells us what the data actually means
manipulation I (similar other)
manipulation II (dissimilar other)
o What would you conclude if you found this?
Similar group – liking score 9/10
Dissimilar group – liking score 4/10
o What if there was a control group?
Neutral control – liking score 8/10
Control group = similarity group; baseline! Doesn’t
Shows us that dissimilarity causes people to reduce
their liking, NOT that similarity causes liking o Not all experiments have control group!
We especially see this in drug trials – pharm company
developing new drug, compare administration of their
drug vs. no drug, but NOT to other drugs already on the
We are able to interpret data in various ways; were
there other ways to manipulate the IV?
Issues with Experiments
Importance of control groups for conclusions
o Comparison to experimental group
Ethical, practical considerations, or both
o Just because we want to do the experiment, doesn’t mean we
o Practicality – depending on methods – can take a LONG time
– expensive – no evidence…do you want to invest in the
experiment? Gather correlational data first!!!
o Ethicality – what are we doing to people?
o Mundane vs. psychological
how close to real life is the situation that you’re
Hard to recreate since people are always looking
for cameras, adjusting their behaviour, etc.
about the experience you’re going through; is that
an approximation of what is real?
What most experiments have – stress is real in
Internal vs. external validity
o the more similar to real-life, the more difficult it is to attribute
the effect to just ONE cause – balancing act
Internal validity In the lab, small space – control everything else –
the more you can control everything BUT the
variable you’re manipulating
The more internal validity, the less external
How much can you really generalize? Can you
extend the causal variety to other situations? If
you take the experiment out of the lab, will the
results be the same?
Experimenter and subject bias
o Hans the counting horse – experimenter thought the horse
was smart enough to count, BUT horse was readin