Methods and History of Gender Research 33answer the phone, or have given up land lines
for cell phones.
Although random selection is impor-
tant for the validity of correlational research,
it is difficult to achieve and is rarely em-
ployed. Often, we want to make inferences
that generalize to the entire population, or at
least the population of our country. It would
be difficult to place 250 million names in a
hat. Instead, we approximate and randomly
select from a community we believe is fairly
representative of the population. The impor-
tant point to keep in mind is that we should
generalize our findings only to the popula-
tion that the findings represent. This is par-
ticularly important in the study of gender
because the vast majority of research has been
conducted on White middle-class Americans,
and findings may not generalize to people of
other races, other classes, or other cultures.
You are probably wondering how a
research participant pool at a university fits
into the random selection process. The an-
swer is, not very well. Do you have a research
participant pool at your institution in which
you are asked to participate in experiments
for credit? Or, are there postings that request
volunteers to participate in research? In ei-
ther case, you are choosing to participate in
a particular experiment; that is, you were not
randomly selected from the entire popula-
tion of college students. Worse yet, the kinds
of people who choose to participate in a cer-
tain experiment may not be representative of
the entire population of students. We must
keep this research limitation in mind when
generalizing from the results of our studies.Experimental Study
A second research method is theexperimen-
tal method. In an experiment, the investi-
gator manipulates one variable, called thenegative correlation. As shown in Figure 2.2,
a negative correlation is not weaker than a
positive correlation; it simply reflects a dif-
ference in the direction of the relation.
Correlational studies are often con-
ducted with surveys or by making observa-
tions of behavior. It is important how you
choose the people to complete your survey or
to be the subject of observation; they need to
be representative of the population to whom
you wish to generalize your findings. I once
had a student in my class conduct an observa-
tional study to see if sex is related to touching.
She conducted the study on the bus and con-
cluded that touching is rare. This study suf-
fered from aselection bias; people on the bus
are not a representative sample, especially dur-
ing the crowded morning commute to work.
To ensure a representative sample, the re-
searcher shouldrandomly selectorrandomly
samplethe participants from the population
of interest. Random selection ensures that each
member of the population has an equal chance
of being a participant in the study. You could
randomly select a sample by putting the names
of all the people in the population in a hat and
drawing out a sample of names. That would be
cumbersome. It would be more feasible to as-
sign every member of the population an iden-
tification number and randomly select a set of
numbers. Imagine you want a representative
sample of 100 adults in your community. If
every phone number in your community be-
gins with the same first three digits, you could
have a computer generate a series of four-digit
random sequences and call those phone num-
bers with those sequences. Would this proce-
dure result in a random sample? Close—but
the sample would be biased in one way: You
would not be representing the people in your
area who do not have telephones. This kind
of research is more difficult to conduct today
because so many people have caller ID, fail toM02_HELG0185_04_SE_C02.indd 33 6/21/11 12:19 PM