- •What is scientific method?
- •Defining the Problem
- •Reviewing the Literature
- •Formulating the Hypothesis
- •Figure 2-1 Causal Logic
- •Independent variable х
- •Collecting and Analyzing Data
- •Developing the Conclusion
- •Ideas for further research
- •M edia image
- •In Summary: Scientific Method
- •Participant Observation
- •Surveys
- •In preparing to conduct a survey, sociologists must exercise great care in the wording of questions (see Table 2-1). An effective survey question
- •Unobtrusive Measures
- •It is important to realize that research designs need not be viewed as mutually exclusive. As was
- •Ethics of research
- •Case Studies of Ethical Controversies
- •Current research
- •In order to study the lifestyle of homosexual males in tearooms, Humphreys acted as a participant observer by serving as a "lookout," warning patrons when police or other strangers ap-
- •Neutrality and Politics in Research
- •If possible, read your paper aloud. Doing so may be helpful in locating sections or phrases that do not make sense.
M edia image
16.2%
N = 767
source: Snow et al., 1986:414.
As this figure shows, there is a sharp contrast between the media image of the proportion of homeless Americans who are mentally ill and the findings of the study in Austin, Texas. According to media estimates, anywhere from 1/3 to 3/4 of the homeless are mentally ill. However, the Austin data suggest that only 16.2 percent of that city's homeless persons are mentally ill and only 10.8 percent have a history of institutionalization.
have reached a misleading conclusion concerning the effects of mothers' working outside the home. In terms of homelessness and mental illness, age and marital status may have an important influence on whether a person comes to the attention of an agency and is subsequently diagnosed as mentally ill. Socially isolated people, such as single persons and the elderly, may be less likely
to be dealt with in "private" by family or friends. However, when the Austin, Texas, sample was examined for such factors, comparisons between the mentally ill homeless, the nonimpaired homeless, and the state's hospitalized population were not found to be dramatically different in terms of age or marital status.
The findings of the Austin study can be expected to contribute to the questioning of the presumed relationship between homelessness and mental illness. Indeed, the researchers conclude that the typical homeless person is a "psy-chiatrically non-impaired individual trapped in a cycle of low-paying, dead-end jobs which fail to provide the financial wherewithal to get off and stay off the streets" (Snow et al., 1986:421). With this finding in mind, greater attention may be given to the structural problems of society that contribute to homelessness, including unemployment and the inadequate supply of low-cost housing. While Snow and his colleagues may have undertaken this research study as basic sociological inquiry, the implications are evident for applied sociology. If the homeless are not typically mentally dysfunctional—if they are merely trapped in economic conditions that lead to poverty and despair—then policymakers must begin to address the issue of homelessness in a very different manner.
An in-depth study of homeless families by sociologist Kay Young McChesney, director of the Homeless Families Project at the University of Southern California, generated similar findings. McChesney and her research team interviewed 87 mothers of children under the age of 18 who were temporarily living in five Los Angeles County shelters. "These women are not crazy," says McChesney; "they aren't substance abusers, either." Instead, she argues that the primary cause of contemporary homelessness is the desperate shortage of reasonably priced housing. "More low-cost housing must be provided," she concludes, "if we are to stem the rising tide of homelessness" (Society, 1987:4).
In Summary: Scientific Method
Let us briefly summarize the process of scientific method through a review of the example. The researchers defined a problem (the relationship between homelessness and mental illness). They re-
PART ONE ♦ THE SOCIOLOGICAL PERSPECTIVE
RESEARCH DESIGNS FOR COLLECTING DATA
An important aspect of sociological research is the decision as to how data should be collected. A research design is a detailed plan or method for obtaining data scientifically. Selection of a research design is a critical step for sociologists and requires creativity and ingenuity. This choice will directly influence both the cost of the project and the amount of time needed to collect the results of the research.
Sociologists regularly use experiments, partici-
pant observations, surveys, and unobtrusive techniques to generate data for their research.
Experiments
When sociologists wish to study a possible cause-and-effect relationship, they may conduct experiments. An experiment is an artificially created situation which allows the researcher to manipulate variables and introduce control variables.
In the classic method of conducting an experiment, two groups of people are selected and compared for similar characteristics such as age or education. The subjects are then assigned by researchers to one of two groups—the experimental or control group. The experimental group is exposed to an independent variable; the control group is not. Thus, if scientists were testing a new type of antibiotic drug, they would administer injections of that drug to an experimental group but not to a control group.
In response to rising concern about how to deal with the increased incidence of family violence, sociologists Lawrence Sherman and Richard Berk (1984) developed an experiment with the Minneapolis police department to evaluate the desirability of three different responses by the police. Officers called to the scene of a misdemeanor domestic assault were randomly instructed to
CHAPTER TWO ♦ METHODS OF SOCIOLOGICAL RESEARCH
take one of the following actions: (1) arrest the alleged offender, (2) make the alleged offender leave the residence, thereby separating him or her from the person who had been attacked or threatened, or (3) offer some form of advice, counseling, or mediation. The researchers took a number of precautions and verified that the police actually handled cases in a way called for by their experimental design. This sociological experiment did not have a control group, since the purpose was to determine which of the three alternative procedures would be most effective in deterring future domestic violence.
Sherman and Berk used two methods to establish which of the three responses by the police was most effective. They checked police records for six months to see if a suspect's name appeared again in a case of domestic violence and also interviewed the original victims by telephone over a six-month period to learn if there had been a repeat incident involving the same alleged offender. The clearest finding of this experimental study was that suspects who were arrested in domestic violence cases were less likely to be involved in such violence at a later date than suspects who were merely separated from the residence or offered counseling and mediation. Suspects who had been arrested and temporarily jailed (usually for less than a week) were less likely to appear on police records over the next six months; when interviewed, the victims of their
original violence were less likely to report repeat incidents. Overall, jailing of a suspect was associated with half as many follow-up cases of domestic violence as the other types of response were. Although there is need for further study of this subject, the use of an experiment led to dramatic results with practical applications.
Clearly, it is impossible for sociologists to observe the behavior of all police officers handling cases of domestic violence. Yet such experiments can still be valuable, because they place "com-monsense" generalizations in a more proper context. Conducting sociological research is more difficult, and therefore more costly, in the field than in a laboratory setting (often on a college campus). Consequently, researchers sometimes must rely on samples composed entirely of college students. Such participants may or may not be representative of the larger American public. There is an additional problem in using a laboratory setting: the responses of subjects in such settings may be different from people's responses in less-structured, real-life situations.