SURVEY RESEARCHreactions to questionnaires. Visuals sometimes may
have a larger impact than question wording changes.
Respondents tend to interpret the middle of a set of
responses as a typical or middle option, treat close-
ness in space on a questionnaire as indicating sim-
ilar meaning, view the top items in a vertical list as
being most desirable, and see differences in space
between answers or the use of different colors as
indicating more significant differences in meaning.
Also, respondents find that organizing response cat-
egories vertically is less confusing than if they are
organized horizontally.^34
Question format and questionnaire design may
influence the results we obtain. Rockwood, Sang-
ster, and Dillman (1997) asked college students how
many hours they studied per day. Some students got
a “low set” of five answer choices, ranging from
0.5 hour to more than 2.5 hours per day. Other stu-
dents received a “high set” of five answer choices,
ranging from less than 2.0 hours to more than
4.5 hours per day. Of students who received the
“low” set of choices, 28 percent said they studied
over 2.5 hours. Of students who got the “high” set of
choices, 69 percent studied over 2.5 hours. Appar-
ently, answer choices had influenced answers. The
researchers also compared survey format for the
same question and answer choices. They sent some
students mail questionnaires and interviewed others
by telephone. Answers changed with the survey for-
mat. Of students asked about studying with the
“low” set of five answer choices by mail question-
naire, 23 percent said they studied over 2.5 hours per
day. Of students interviewed by phone with the
“low” set of choices, 42 percent gave the answer of
2.5 hours per day. For students who received the
“high” set of five answer choices, answers by mail
questionnaire and phone interview were similar. In
the same study, the researchers asked students about
the number of hours they watched television with
similar “high” and “low” response category sets,
comparing mail questionnaires and telephone inter-
views. For the topic of television watching, ranges of
response categories or format did not affect answers.
This study shows us three things. Respondents
rely on the range of response categories in a question
for guidance; they answer more honestly with
more anonymous survey formats, such as a mail
questionnaire, compared to less anonymous formats,
such as interviews; and both response categories
and survey format shape answers about some top-
ics more than other topics.^35Wording Issues
We face two wording issues in creating question-
naires. The first, discussed earlier, is to use simple
vocabulary and grammar to minimize confusion.
The second issue involves the effects of specific
words or phrases. This is trickier because we do not
know in advance whether a word or phrase affects
responses.^36
A well-documented difference between forbid
and not allowillustrates the problem. Both terms
have the same meaning, but many more people are
willing to “not allow” something than to “forbid”
it. In general, less well-educated respondents are
influenced more by minor wording differences than
educated ones.
Certain words trigger an emotional reaction,
and we are just beginning to learn of them. For
example, Smith (1987) found large differences (e.g.,
twice as much support) in U.S. survey responses
depending on whether a question asked about
spending “to help the poor” or “for welfare.” He
suggested that for Americans, the word welfarehas
such strong negative connotations (lazy people,
wasteful and expensive programs, etc.) that it is best
to avoid it.
Possible wording effectsare illustrated by
what appears to be a noncontroversial question.
Peterson (1984) examined four ways to ask about
age: “How old are you?” “What is your age?” “In
what year were you born?” and “Are you... 18–24,
25–34,... ?” He checked responses against birth
certificate records and found that from 98.7 to
95.1 percent of respondents gave correct responses
depending on the form of question used. He also
found that the form of the question that had theWording effects Results in survey research when the
use of a specific term or word strongly influences how
some respondents answer a survey question.