1
Research Methods For Studying
Family Communication and Relationships
The adept consumer of research on family processes must have an understanding of the different techniques that are employed in the design, collection, and analysis of family research studies and the data that they generate. Although a thorough treatment of these issues is beyond the scope of this book, in this appendix we provide a brief summary of some of the different research methodologies, study designs, and measurement techniques that are commonly used in research on families. More in-depth analyses of these issues can be found in Acock (1999), Copeland and White (1991), Markman and Notarius (1987), Miller (1986), Noller and Feeney (2004), and Socha (1999).
Research Methodologies
There are a variety of different methods that family researchers use to investigate family phenomena. Each of these methods has strengths and weaknesses. Understanding these different methods, what they can reveal, and what they cannot reveal, is useful for interpreting the results of various research studies. In this section we briefly review some of the more common research methodologies that appear in the family communication and relationships literature.
Surveys
The majority of what we know today about family relationships came from survey research. The common element of all survey research is that investigators ask research participants to provide information. This produces what is known as self-report data.Qualities of self-report data are discussed in more detail later in this appendix. Ordinarily survey research involves large numbers of participants. This is because surveys are fairly easy to administer to large samples even if spread out over diverse geographic regions.Survey researchers can use the mail, telephone calls, and internet questionnaires to gather information, making it easy to reach many people. For example, the National Survey of Families and Households (e.g., Bumpass, Martin, Sweet, 1991) involved interviews of over 13,000 households, producing one of the more intensively analyzed data sets in family science.
There are a number of different data collection methods that are used by survey researchers. Perhaps the most common is the use of self-administered questionnaires.These are paper and pencil measures that are given to respondents to complete and return to the researcher. Questionnaires often contain statements and closed-ended questions that respondents answer with various numerical scales. For example, the Family Assessment Device (Miller, Epstein, Bishop, Keitner, 1985) contains items to measure family communication, such as “People come right out and say things instead of hinting at them,” and “We are frank with each other.”Respondents indicate their answer by circling a number on a scale where 1 = strongly disagree, 2 = disagree, 3 = agree, 4 = strongly agree. This is known as a Likertscale. However, not all questionnaires contain closed-ended questions that are answered on Likert scales. Some might ask family members to respond to open-ended, essay-type questions such as “Describe your ideal family vacation.” Such questions produce qualitative data that could be analyzed or coded for various themes that appear in the answer.In some cases researcher-administered questionnaires are used instead of self-administered questionnaires.Researcher-administered questionnaires are typically read out loud to the participant and the researcher records the answer.This can be a very useful technique for studying certain populations such as children who cannot read very well or elderly people with poor eyesight.Sometimes survey research is conducted over the telephone, in which case researcher-administered questionnaires are employed.Self-administered questionnaires are useful when the questions are straightforward, not easily misunderstood by respondents, and when it is desirable to maintain the respondent’s privacy and anonymity.Researcher-administered questionnaires are useful when questions are complicated and might need to be clarified by the researcher and when the data are collected over the telephone or in face-to-face interviews.
Many interview studies could be classified as a type of survey research.Like surveys more generally, interview methods are diverse and range from unstructured to highly structured.The more structured the interview is, the more the interviewer knows exactly what will be asked during the interview.An interview schedule is a set of questions that will be asked by the interviewer.In highly structured interviews the schedule will literally be a questionnaire that the interviewer reads to the participant, recoding answers on defined scales.In more qualitative investigations the interview schedule might involve only a general outline of issues that are to be raised in the interview.In such cases the interviewer might make decisions about what topics to pursue based on the responses of the participant.Responses to less structured interviews are often recorded through note taking or audiotape.
Experiments
The hallmark features of a true experiment are manipulation of an independent variable by the researcher and random assignment of research participants to the different experimental conditions.In the prototypical experiment, only the independent variable is manipulated—all other variables are controlled or held constant.This way, if the different groups (e.g., experimental vs. control) differ at the end of the experiment, that difference can be attributed to the effect of the manipulated independent variable.For this reason researchers will often design and conduct experiments when they are interested in isolating the effect of some variable (independent variable) that is assumed to have a causal effect on some outcome (dependent variable).For example, Noller, Feeney, Peterson, and Atkin (2000) created a series of audiotapes that portrayed different styles of marital conflict.These included mutual negotiation, coercion, mother-demand/father-withdraw, and father-demand/mother-withdraw.In this experiment Noller and her colleagues manipulated the conflict style heard on the audiotape.The tapes were then played to father, mother, and child family triads.Results indicated that the mutual negotiation conflict was viewed more positively than the other types, and that the mother-demand conflict was rated as more typical than the father-demand.Through this experiment, the researchers were able to understand the qualities of family conflict that produce positive or negative reactions in family members.
As Noller herself points out, experiments have not been extensively used to study family communication and relationships (Noller Feeney, 2004).However, experiments still represent an important tool for family researchers.Experiments are much more valuable than surveys for demonstrating causeeffect relationships.This is because they are high on internal validity.The question of internal validity concerns how confidently the researcher can conclude that the dependent, or outcome, variable was affected by the independent, or manipulated, variable.Because of their control over extraneous variables, experiments are ordinarily strong in internal validity.However, this strength comes at a price.In order to control extraneous variables (e.g., time of day, room temperature, physical environment, noise, etc.) that could affect the dependent variable, experiments are often conducted in rather artificial laboratory environments that bear little resemblance to the “real world.”Consequently, it is sometimes difficult to generalize the results of laboratory experiments to more naturalistic and realistic environments.For this reason many experiments are low in external validity, which represents the extent to which the results of the investigation can be generalized to environments and contexts external to the laboratory.
Is there a happy medium between the high internal validity and low external validity inherent in most experiments?To better balance these two legitimate features of experiments, some researchers conduct quasi-experiments.Simply put, a quasi-experiment is not a true experiment because the researcher does not actively manipulate the independent variable or because the researcher could not randomly assign participants to conditions. There are naturalistic experiments going on all the time in society.For instance, Menees and Segrin (2000) were interested in the effects of various family stressors (e.g., death of a parent, parental alcoholism, parental divorce, etc.) on the social climate in families.For obvious reasons researchers could never randomly assign people to these different family stressors.Instead the authors selected people from the population who had been naturally exposed to these family stressors.The problem with this technique is that internal validity is in question.If the family environment in households with an alcoholic parent has more conflict than those that experienced the death of a parent, is that because of the nature of the two different stressors, or perhaps some other variables like education or income that also differ as a function of the stressors?It is sometimes impossible to answer these questions in a quasi-experiment.On the plus side, quasi-experiments are often much higher in external validity than laboratory experiments.After all, the people in quasi-experiments usually experienced the “manipulation” as a part of their everyday life and live with it in their natural environment.
Content Analysis
Content analysis is a technique for quantifying recorded communication or communication texts.So, for example, researchers might use content analysis to describe the prevalence of divorced characters on prime time television, the number of interracial families in situation comedies, or the frequency of extra-marital relationships portrayed on daytime soap operas.In the domain of family communication, content analysis is most often used to describe media depictions of family interactions and relationships.In Chapter 12 we discussed a content analysis by Robinson and Skill (2001), which showed that depictions of stepfamilies in prime time fictional television have been increasing over the years but still lag behind the actual prevalence of stepfamilies in the general population.
Content analysis is a useful tool for describing what appears on television, in magazines, romance novels, motion pictures, etc., but it does not tell us who consumes these messages or what effect they have on the viewer.For example, if a content analysis showed that more divorced people were portrayed on network television in the 1990s compared to the 1960s, it would be tempting to conclude that these media depictions are at least partly responsible for the more widespread acceptance of divorce in recent years.However, this inference goes far beyond what the data actually indicate.Similarly, if a content analysis showed that family situated programs (e.g., Roseanne, The Osbournes) had more intense conflict today than they did in the past, some might conclude that this teaches young people that extensive bickering and fighting is normative in the family.As reasonable as this conclusion would appear, the content analysis alone does not tell us that young people are actually viewing these programs.That would require additional data, perhaps from a survey of television viewing habits of young people.
One vital decision that must be made when conducting a content analysis is determining the unit of analysis.The unit of analysis is the basic unit or segment of the communication text that will be measured or classified.Consider for example a content analysis of family situated television programs (e.g., The Brady Bunch, Roseanne, The Osbournes).At a broad level, one could treat the entire series as a unit of analysis.For example, these shows could be classified as either “fiction” or “reality” based.At a more specific level, the unit of analysis could be the episode.Researchers could, for instance, classify each episode according to its dominant plot.Getting even more specific, individual scenes or characters could be the unit of analysis.More specific yet, a researcher could classify individual utterances of the characters on these programs.Perhaps there is more profanity used in family situated programs today than there was back in the days of Leave it to Beaver.In this case, a researcher could classify the individual utterances of each character in terms of whether or not they contain profanity.In any event, decisions about the appropriate unit of analysis are inextricably connected with the nature of the research question.If the research question concerns how often extended family members are portrayed as living in the same home, the entire series or individual episodes could be the unit of analysis.On the other hand, a research question about whether fathers are depicted as reprimanding children more often than mothers are would require a much more microscopic unit of analysis, perhaps down to the individual scene or individual utterances of each character.
A second vital element of content analysis is the coding scheme.A coding scheme is simply a classification system for describing the content of the communication text.For example, a content analysis of family roles in motion pictures might include the following categories: mother, father, husband, wife, son, daughter, brother, sister, grandfather, grandmother, aunt, uncle, stepmother, stepfather, cousin.Researchers could then classify each character into one of these categories to see what family roles are most and least commonly portrayed in the movies.Coding schemes used in content analysis must have categories that are mutually exclusive.This means that each unit to be coded can be classified into one and only one category.What if the coding scheme above was used to classify Kevin Spacey’s character in the film American Beauty?He was both a husband and a father.As straightforward as this coding scheme appears, it could be problematic when actually getting down to the business of coding real communication texts.A coding scheme must also be exhaustive.This means that there is a category for every unit to be coded.What if a character in a film was the brother-in-law of another character?The scheme presented above has no category called “brother-in-law,” so it is not truly exhaustive.Sometimes researchers will make exhaustive coding schemes by using an “other” category for all those units that cannot be fit into one of the existing categories.
Finally, there are some cases in family research where content analysis is used as a method of analyzing data that were gathered as part of an interview or survey research study.For example, Fiese and her colleagues interview families and record narratives or stories about their experiences (Fiese et al., 2001; see Chapter 3).These stories would then be transcribed and essentially content analyzed for narrative coherence (how well the story is constructed and organized), narrative interaction (how well the family works together to jointly construct the narrative), and relationship beliefs (the way that the family’s view of the social world is reflected in their story). In this way, content analysis is useful for summarizing and cataloging the qualitative data that are provided by research participants.
Ethnography
Ethnography is a type of qualitative field research that is aimed more at description than explanation.In particular, ethnographic studies examine various phenomena in their natural settings rather than in laboratories.Unlike experiments or surveys, in ethnography investigators will often get directly involved in the subject matter that they are studying and interact frequently with the research subjects.The idea is that the closer one’s contact is with the phenomenon under investigation the better able he or she is to offer a detailed and accurate description of that phenomenon.
Ethnographic studies tend to follow the grounded theory approach.Researchers who develop grounded theories start with careful and detailed observations and then develop more general theoretical explanations based on those descriptions.This is more of an inductive than deductive approach to theory development.
In ethnographic research there are a number of different relationships that might exist between the researcher and his or her research subjects.When the researcher acts as a complete observer he or she simply observes phenomena as they occur in their natural environment.In such cases the research subjects are unaware of the ethnographer’s observations and are therefore unlikely to alter their natural behavior.In a closer level of involvement with research subjects, ethnographic investigators sometimes assume the role of observer as participant.In such cases research participants know that they are being observed but the researcher tries to “fit in” to the situation or context without actually participating in it.When researchers act in the role of participant as observer they actively participate in the phenomenon under investigation.For example, if an ethnographic investigator attended a family reunion and ate hotdogs and talked with various family members, he or she would be acting as a participant observer.This perspective gives the researcher considerable insight into the phenomenon under investigation but this familiarity comes at a cost as the researcher may actually influence the phenomenon that is being studied.Finally, ethnographic researchers sometimes act as complete participant.In this case they actively participate in the phenomenon that is being studied but do not inform research participants of their dual role as both participant and researcher.This is comparable to being an undercover police officer.If the behavior being observed is public, this is not a problematic technique, but if it is private there are obvious ethical issues associated with the difficulty of securing informed consent from the potential research participants.