11 - 11
one or more of these other factors rather than to the type of TV program presented by the researcher.
Unfortunately, we cannot choose among these alternative interpretations. On the other hand, if this study
had used random assignment of participants, alternative interpretations would have been eliminated and
strong, confident conclusions could have been drawn.
Other examples of quasi-experiments are cross-sectional research designs. Developmental
psychologists are often interested in how the behaviors of individuals may change across portions of the
life span or across the entire life span. We could ask the question “Do people’s responses to violence on
TV change as they develop from young children to school-age children to teenagers?” An approach using
a cross-sectional design might involve testing children of different specific age groups and then making
comparisons across the age groups. Once again, notice that the researcher can manipulate the
independent variable (level of TV violence) but does not use random assignment to form the age groups.
The different age groups are preexisting groups.
When quasi-experiments are conducted using an experimental group and a comparison group, there
is always concern about the equality of the two (or more) groups because random assignment is absent.
Any differences between the two groups could be due to the treatment that only the experimental group
received or to differences between participants in the two groups (perhaps brighter, healthier, or more
motivated in one group than the other). Researchers conducting quasi-experiments may try to minimize
participant differences by selecting participants as similar as possible for the two groups, but this is very
difficult. There is no way to be sure that the composition of participants in the two groups will be equal,
but a random assignment procedure comes closest to achieving this goal.
As we have just discussed, quasi-experiments involve manipulation of an independent variable but
lack random assignment. Other studies may lack both manipulation of the independent variable and
random assignment. Common examples include many typical undergraduate studies that involve
comparisons of gender, race, socioeconomic status, fraternity/sorority status, and other preexisting
characteristics. Because these are not true experiments, strong cause–effect conclusions cannot be made.
These types of studies will be discussed further in Chapter 15.
Case Analysis
One important area of behavioral research involves attempts to reduce the occurrence of risky sexual
behaviors in teenagers. A recent study showed that a high school program taught by teachers was more
effective in the long term than a similar program taught by peers (Fisher, Fisher, & Bryan, 2002). You
decide to determine whether the same principle applies to drug education. Using three high schools, you
implement a teacher-based program in one school, a peer-based program in a second school, and no
program in a third school. For the dependent variable, you ask students to anonymously report the number
of times they have used drugs in the previous month. You measure level of drug use at the beginning of
the school year and at the end of the school year and calculate a “change” score by subtracting one from