Modified: 2007-02-04
Control of extraneous variables by different research designs.
(?? means the procedures must be studied to determine if extraneous variable is present)
Designs Extraneous Variables
|
Selection |
Differential Attrition |
Diffusion of Treatments |
Testing |
Instrument Change |
History |
Maturation |
Regression |
Quasi-experimental |
Not |
Not |
Not |
Cont |
?? |
?? |
?? |
Not |
Random Assignment |
Cont |
Not |
Not |
Cont |
Cont |
Not |
Cont |
Cont |
One-Group Two-Treatment |
Cont |
Cont |
Not |
Not |
?? |
Not |
Not |
Cont |
One-Group Pretest-posttest |
Cont |
Cont |
Cont |
Not |
Not |
Not |
Not |
Not |
Quasi-experimental Two-Group |
Cont |
Not |
Not |
Cont |
?? |
?? |
?? |
?? |
Random Two-Group |
Cont |
Not |
Not |
Cont |
Cont |
?? |
Cont |
Cont |
Beyond naturalistic and participant observation lies a method nearer to the experimental designs discussed in earlier chapters. Field experiments are similar to controlled experiments except that the researcher does not create the treatment conditions. Instead, researchers discover and use those treatments in the environment. The example below will help explain how field experiments are conducted.
Field Experiments
If you lived in Munich a few years ago, you knew that the airport was closing and a new one was being built. When the new one opened, the old one closed. This situation provided an opportunity to study the effects of aircraft noise. People living around the old airport experienced the cessation of aircraft noise, while those living near the new airport experienced its onset. This was prospective research because researchers took advantage of an ongoing situation and collected data while the changes happened.
A body of previous research on the effects of aircraft noise on physiology and behavior exists, but most of that information was collected cross-sectionally (a between subjects design, see Chapter 9) not prospectively. Consequently, the conclusions of the previous studies could have been invalidated by extraneous variables such as history or maturation. A prospective study provides clearer evidence of the effects of airplane noise because researchers eliminate history or maturation as extraneous variables by using the same children before and after noise exposure. The building of the new airport and the closing of the old airport provided an opportunity to collect data using a repeated-measures experimental design. However, the researchers could not randomly assign the children to live near either airport. So this research used a non-random assignment design, not a random assignment design. (see Chapter 7 for further discussion of both of these designs).
In any free and open society it would be impossible to assign people at random to live near or far away from an airport. So, the opening of the new airport and the closing of the old one is about as close as anyone can hope to approach random assignment outside of the laboratory. Additionally, the opportunity to measure participants at the old airport while the aircraft noise is still there and then to measure them again later once the noise is gone, and to do the opposite with participants at the new airport is an extremely rare event.
Hygge, Evans, and Bullinger (2002) took advantage the situation described above. The opening of the new Munich (Germany) International airport and the simultaneous closing of the old airport gave them the opportunity to collect data on the effects of aircraft noise on cognitive performance of 8-12 year old children. The data collection took almost three years, and started six months before the old airport closed. After the new airport opened, data were collected again. The researchers were interested in determining if aircraft noise had effects on reading, memory, attention, and speech perception.
The researchers found that reading scores and long-term memory were strongly affected in the children living near each airport. Furthermore, the direction of the effect was opposite depending on whether the children lived near the old or the new airport. Children who lived near the noise of the old airport improved in both areas when the airport closed. Children who lived in the now noisy areas near the new airport declined in both areas after the airport opened. This study supported the conclusions of older, cross-sectional data and also added something new. For the first time, researchers demonstrated that the effects of aircraft noise on reading and long-term memory were reversible. The design of the field experiment allowed the researchers to control for two extraneous variables, history and maturation.
Definition of Field Experiments
Field experiments are observational research techniques too, but they typically involve more preparation than naturalistic observation or participant observation.
Field experiments usually take advantage of some naturally occurring changes or events in the environment. Researchers who are aware of those changes may study their effects. Here are two more examples of field experiments.
Examples of Field Experiments
Scheduled events often provide researchers with the opportunity to use the event as a treatment. In one such instance, Knaus, Pinkleton, Austin, and Weintraub, (2000) assessed the effect on college students of viewing the NAMES Project Foundation’s AIDS Memorial Quilt (AIDS Quilt). The AIDS Quilt is a traveling exhibit designed to encourage information seeking about AIDS and to promote safe sex practices. In this field experiment, the researchers used the scheduled arrival of the AIDS Quilt as the treatment. Using a modified Solomon four-group design (see Chapter 9), they randomly assigned groups of students either to view the AIDS Quilt when it came to campus or to participate in one of two control interventions. The results were that students who viewed the AIDS Quilt were significantly more likely to seek information about AIDS, to discuss HIV/AIDS with their sexual partners, and to use a condom. The researchers used the AIDS Quilt event on their campus as an experimental treatment and were able to statistically assess the Quilt’s impact on the dependent variables that they measured.
Patients in a hospital can provide another opportunity for field experiments. Rotton and Shats (1996) analyzed the effects of humor on hospital patients after surgery. They used a 2 x 2x 2 factorial design (see Chapter 9) with both between and within groups and with an appended control group. In the between portion, the researchers manipulated type of movie (comedy vs. action-adventure), choice (high-picking from a list of 20 movies vs. low-picking from a list of 4 movies), and expectancy (positive expectancies about movies and healing vs. neutral expectancies about movies and healing). For the within portion of the design, all patients were measured on both the first and second day after surgery. The dependent variables were a variety of self-reports, use of major and minor analgesics (pain medicines), and social interaction (whether they had watched the movie alone or with others). The results were that humor, when combined with positive expectations, reduced requests for minor analgesics. However, movies increased requests for major analgesics for patients in the low choice conditions.
The major difference between field experiments and controlled experiments is that the researcher is not actually setting up the independent variables in field experiments. Instead, researchers conducting field experiments look for situations in which a naturally occurring change is about to take place. All three examples in this section illustrate this data collection strategy. In the airport study, a major source of noise was moving from one part of Munich to another. The arrival on campus of the AIDS Quilt was the change in that experiment. Surgical patients’ pain and its relief was the setting for the last example. Although field experiments can be as simple or as complex as other experiments, researchers have less control over the choice and type of independent variable. In field experiments, psychologists are piggybacking their research onto the changes others are already making or plan to make in the natural environment.
Conducting Field Experiments
Participants In field experiments, the participants can be animal or human. The selection and recruiting of participants in field experiments is often more limited than in other types of research. Only children living near one of the Munich airports, students at the campus where the AIDS Quilt was displayed, and surgical patients at a hospital are potential participants for the studies described above. Researchers could not build a new airport, send the AIDS Quilt to another college, nor refer patients for surgery. Researchers could and did take advantage of planned events to conduct worthwhile field experiments. Once the participant assignment limitations of field experiments are acknowledged, recruiting participants is similar to other experiments.
Apparatus The apparatus for conducting field experiments is no different from that used to conduct other experiments.
Procedure Procedures for field experiments are very likely to be more constrained by external factors as compared to other experiments. Those external factors may dictate procedural accommodations. In the airport study, noise levels were studied, but the experimenters could not manipulate those noise levels directly. Instead, they indirectly controlled noise by choosing neighborhoods closer or more distant from the flight patterns. In the AIDS Quilt study, the researchers could not schedule the arrival of the Quilt, but they were able to control (using random assignment) the groups who viewed the Quilt or who participated in the control conditions. In the study of humor among surgical patients, there was much the researchers could not control. There were some variables that they did control, however, including the type of movie, and choice of the movie. Thus, once the procedural limitations of a particular field experiment are understood, there are many variables still available for the researcher to control.
Results In field experiments with quantitative or qualitative data, the analysis is conducted in the same manner as in other experiments.