SOCIAL PSYCH. CHAPTER 2 SUMMARY
CHAPTER 2 - The Methods of Social Psychology
Study by Cohen & Nisbett: letter sent to Northern & Southern USA for job
application where the individual explained a murder he committed and why
before being refused for it. The study was to measure how cooperative
potential employers were w/ the client. Rated how note was sympathetic,
encouraging and if it mentioned an appreciation for the candidature.
Distinct patterns in replies. Employers from South complied more w/ request,
were warmer and much more sympathetic.
WHY DO SOCIAL PSYCHOLOGISTS DO RESEARCH?
Seems like the world is a reasonably predictable place: most of our sit. are
similar to each other & our observations /b how ppl behave in those sit. are
accurate enough to allow us to navigate through the world w/ some
confidence in our predictions.
However, many sit. ca contain surprises.
Our opinion /b why we behave the way we do can be mistaken. Many
influencing factors are hidden from us - unconscious as seen in previous
Hindsight bias: ppl’s tendency to be overconfident /b whether they could
have predicted the outcome.
HOW DO SOCIAL PSYCHOLOGISTS TEST IDEAS?
Even if you can’t do a study to test smtg, thinking through how you would
test it might give you a new hypothesis that you haven’t considered before.
Hypothesis: a prediction /b what will happen under particular circumstances.
Example of hypotheses in social psych.: if person A likes person B and person
B dislikes person C, person A will either dislike C or start disliking B.
Hypotheses are tested in studies which test predictions /b what will happen
in particular concrete contxt.
Theory: a body of related propositions intended to describe some aspect of
the world. Usu have support from empirical data & have often made
predictions that are surprising except in the light of the theory. They lead to
understanding of natural phenomena or to imp. real-life conseq. for instance.
More general than hypothesis which are more general than findings from the
studies that test them.
Looking at a phenomenon in some reasonably systematic way w/ a view to
understanding what is going on & coming up w/ a hypothesis /b why things
are happening as they are.
Charles Darwin - 1st observer - looked at finches in Galapagos Islands which
led to his theory of evolution by natural selection. SOCIAL PSYCH. CHAPTER 2 SUMMARY
Participant observation involves observing a phenomenon at close range. I.e.
live w/ a group of ppl for a certain period of time while just noting what they
do & coming up w/ guesses.
Cultural anthropologist Shirley Brice Heath used this method to study
preparation for schooling by middle-class and working-class families. She
lived w/ the families - observing and taking part in their daily activities. She
found difference b/w the 2 groups.
Results show that middle-class families read to their children lots, included
them in dinner-table conversations, used the printed words to guide their
behaviors (recipes, game rules), and taught them how to categorize objects,
answer “why” questions, and how to evaluate/make judgements.
Working-class families didn’t do these things as much and although their
children were reasonably well prepared for school, their lack of preparation
showed up in later grades when faced w/ + complex tasks involving
categorization & evaluation.
Social psychologists Roger Baker & Herbert Wright studied how children in a
Midwestern town interacted w/ their surroundings. They followed children in
their daily activities. Study showed lots /b the way young ppl interact w/ their
environments - the opportunities & constraints that came w/ the
environment - and the factors that molded their characters.
Observe soc. sit. in a semiformal way taking notes & interviewing
participants. Also usu design additional research to verify the impressions.
Observations are often misleading & have to be tested w/ other methods.
Look at evidence found in archives of various kinds.
Nisbett & colleagues studied FBI reports of homicides and found that
homicides were + common in the South than North. Also analyzed various
types of of murders & discovered that in the South, most common kind
involved some type of insult & cases of man’s gf/wife leaving him for
another. Other kinds of homicide actually less common in the South than the
These observations led them to a research program /b whether Southerners
respond more agressively to insults or whether the higher rate of insult-
related homicides was due to factors such as hot temperature or lenient
One of the most common types of research.
Using either interview or written questionnaires.
Can be /b any kind of group of ppl - from small communities to large national
population. SOCIAL PSYCH. CHAPTER 2 SUMMARY
Need a sample of ppl in the survey to be random to represent the population
accurately. The only way to obtain a random sample is to give everyone an
equal chance to be chosen.
A convenience sample is obtained from available subgroups in the population
- i.e. contracting ppl as they enter a library. It is not random and might be
biased since it can include too many ppl from a kind and not enough of the
Info based on biased samples can be worse than no info at all.
For example, the survey by Literary Digest based on more than a million
respondents. Erroneously predicted that the Republican Alf Landon would
defeat Franklin Delano Roosevelt in 1936. Landon carried only 2 states. Why
biased? Because survey was drawn from telephone directories & automobile
registrations. In these yrs, wealthy ppl were + likely to own phones & cars
than poorer ppl and they were more likely to be Republicans.
Magazine surveys are like fictional results. Ppl who take time to respond to
these polls are likely to be different from those who do not respond &
therefore unlikely to represent the population as a whole. Cosmo readers
who lost weight are more likely to answer a survey /b weight loss then the
ones who didn’t lose any. Random criteria not met.
The # of ppl needed in a sample to get a reasonably accurate count on smtg
is usu independent from the size of the population. For instance, a sample
of /b 1,200 ppl from the entire adult population of USA is enough to estimate
the population value to a degree of accuracy of +/- 3% and to be 95%
confident that the true value is within the range.
Cohen & Nisbett used surveys to see attitudes towards violence. Guessed
that Southerners committed more homicides bc they might be more
accepting of violence. However, when looked at published national surveys
on attitudes towards violence, found few regional diff. For ex., Southerners
were no more likely than Northerners to agree w/ “eye for an eye” as a
justification. They were + likely to agree to “when a person harms you, you
should turn the other cheek and forgive him”. But they still found that
Southerners were + likely to favor violence towards insults & when ppl get
on other’s gf/wife. More likely to approve of violence when threats to
home/family/thinking. Also + approving of violence in socializing children.
Turned to anthropologists & historians to try to explain. South might be a
“cult. of honor” since the North was settled + by farmers from England,
Germany & Holland as South was + settled by herding ppls from the edges of
Britain - Scottish, Irish, etc. They tend to be more tough guys throught the
world. Protection is very imp. to them.
Correlational Research SOCIAL PSYCH. CHAPTER 2 SUMMARY
Correlational research: does not involve random assignment to diff. sit., or
conditions, and that psychologists conduct just to see whether there is a
relationship b/w the variables.
Experimental research: in soc. psych., research that randomly assigns ppl to
diff. conditions, or sit., and that enables researchers to make strong
inferences /b how these diff. conditions affect ppl’s behavior.
Corr. research is an imp. way to begin line of inquiry but needs further
exploration. Does var. 1 causally influence var. 2 or is it the other way
around? (meaning reverse causation: when var. 1 does not cause var. 2, yet
the opposite may be the case).
Does some 3rd var. influence both? (both fluctuations in the var. 1 & 2 are
not caused by one another but by some 3rd).
We are never certain of the causality.
For example: TIME Magazine published a story /b love & sex being good for
physical and mental health. Quoted statistics /b married ppl being happier
than unmarried ppl. However, not clear /b the variables. Happier ppl may be
+ appealing to others & + likely to get married for that reason, so happiness
may cause marriage rather than marriage happiness - example of reversed
Or good physical & mental health leads to greater likelyhood of marriage as
well as greater likely of being happy. In this case, the causal factor would be
a 3rd var.
Cannot prove a causal relship bc of self-selection - a problem that arises
when the participant, rather than the investigator, selects his or hers lvl on
each var., bringing w/ this value unknown other properties that make causal
interpretation of a relshp diff.
I.e. in the TIME Magazine article, they were not aware of the other
properties /b ppl other than their marital status & physical & mental health -
claimed by themselves. Opposed to investigator-selection.
Investigator can only look at the degree of relshp b/w var. Strength of relshp
can range from 0 to 1 - zero being none and 1 being perfect. By convention,
corr. of 0.2 indicates a slight relshp, 0.4 a moderately strong relshp & 0.6 + a
very strong relshp. Showed in scatterplots. Can be + pr - relshp.
p.50 various explanations for some specific corr.
i.e. 0.2 - overweight & cardiovascular illness. Might cause it but is not the
only possible cause
0.5 - Height & weight. Are related but not caused by each other.
0.8 - math scores on the scholastic aptitude test & SAT. Still no causation,
external factors explain it.
TRAIN ON P.51 BOX 2.2 SOCIAL PSYCH. CHAPTER 2 SUMMARY
Corr. research can be very helpful in alerting investigators /b various
possibilities for valid causal hypotheses /b nature of the world. Can also be
the best option when an experimental study could be diff. or unethical.
However, don’t tell us /b direction of causality and if some 3rd var. is driving
the ass. b/w the var.
Still, clever analysis of corr. data can be persuasive /b the meaning of the
I.e.: “Ppl who watch local news showing murders, fires & other newsworthy
mayhem see + danger in the world than ppl who do not” - Obvious expl. =
ppl feel + at risk. But it could also be that ppl who are anxious watch more to
justify themselves. Or 3rd var. such as elderly ppl watching w/ + anxiety ?
But elderly ppl do not watch necessarely as much tv as younger ppl or may
not be more anxious. All the possibilities could be assessed by research.
Ruling out one hyp. doesn’t prove the relshp is causal but as other
alternatives are tested & rejected, the nature of the relshp b/w the 2 var. can
“Ppl who watched a great deal of violent TV at 8 yrs old are + likely to be
incarcerated later in life”. - Obvious that watching violent tv can make ppl +
violent. On the other hand, also obvious that ppl w/ violent tendencies would
be + likely to watch violent TV when younger & + likely to engage in criminal
behavior when older. Such longitudinal study (a study conducted over a long
period of time w/ same population which is periodically assessed regarding a
particular behavior) rules out the opposite direction of causality: nothing that
happens to someone at 30 can affect anything the person did when younger.
Clever analysis of data can make some interpretations much less plausible
The best way to be sure about causality. Requires ind. & dep. var.
Independent var.: in experimental research, the var. that is manipulated. It is
hypothesized to be the cause of a particular outcome.
Dependent var.: in experimental research, the var. that is measured - as
opposed to manipulated - it is hypothesized to be affected by the
manipulation of the ind. var.
The experimenter determines which var. is dependent and independent and
what lvls for for that var. will be. Dependent var. can be measured in many
ways: verbal reports, behavior, physiological measures or neural measures.
Advantage of exp. research: ability to expose participants to diff. lvls of the
ind. var. by random assignment - assigning participants in exp. research to
diff. groups randomly such that they are as like to be assigned to one
condition or another. There should be no differences across experimental
groups that way. It also rules out t