Measuring Interpersonal Communication

Measuring Interpersonal Communication John P. Caughlin Erin Basinger University of Illinois at Urbana-Champaign Caughlin, J. P., & Basinger, E. D. (2016). Measuring interpersonal communication. In C. R. Berger & M. E. Roloff (General Eds.) & S. R. Wilson, J. P. Dillard, J. P. Caughlin, & D. H. Solomon (Assoc. Eds.), International encyclopedia of interpersonal communication (pp. 1-13). New York, NY: Wiley. doi: 10.1002/9781118540190.wbeic0229 Final Draft: This version may differ from published version due to copy editing. Abstract Measurement is a core activity and challenge in all social sciences, and the complex nature of interpersonal communication makes it particularly challenging to measure. This entry provides a discussion of basic concepts in measurement and an introduction to the myriad considerations to be made in measuring constructs in interpersonal communication. It includes a review of the advantages and disadvantages of the most common techniques for measuring interpersonal communication, including observational assessments, self-reports, diaries and logs, and physiological assessments. In addition to the perennial challenges of measuring in interpersonal communication scholarship, this entry notes some of the promises and difficulties implicit in the rise of new communication technologies in interpersonal interaction. Measuring concepts is essential to the systematic study of interpersonal communication. Although sound measurement is a foundation of good social science, it can be extremely complex. Communication scholars are typically interested in abstract concepts such as affinity-seeking, attributions, uncertainty, and many other constructs explicated in this encyclopedia. Although communication theory and research focus on abstract constructs, measurement of such concepts requires the use of empirical variables that provide some indication about the amount or levels of the constructs of interest. What researchers study directly are the variables that are meant to represent constructs, rather than the constructs themselves (Capella, 1991). Finding empirical variables to assess concepts is far from simple. Because “abstract constructs do not have one-to-one correspondence with empirical indicants…abstract concepts can only be approximated by empirical indicants” (Zeller & Carmines, 1980, p. 3). The notion that measurement is an approximation of constructs is important to keep in mind when considering research on communication. Scholars sometimes claim that their measures are objective, and this can be true in the sense that some variables are based on observable and discrete information that anyone could see (or be trained to see). For instance, one behavior that has been coded in interpersonal conflict is eye rolling, and with high quality video, coders can discern a precise and objective number of times people roll their eyes during a particular segment. Yet, readers should remember that having an objective behavior to count does not change the fact that the empirical indicant is still an approximation (or, more likely, one indicator among several that collectively serve as an approximation). Few communication scholars would be interested in counting eye rolls as an end in itself; instead, examining how often people roll their eyes nearly always would reflect an attempt to assess some broader theoretical construct. Pointing out that the scholarly interest is usually in some broader construct, rather than in the empirical variables themselves, is important because it highlights the fact that even when communication behaviors can be counted objectively, the meaning of those behaviors is not necessarily clear or straightforward. For instance, the meaning of behaviors during communication is greatly influenced by context; thus, even if one determines that a particular behavior in a particular situation provides useful information about a specific concept, that same behavior could mean something else in another context. Consider eye rolling behavior, for example. If it occurs in response to a relational complaint that one has heard many times (e.g., “you never do your fair share around the house”), rolling one’s eyes is probably an indicator of contempt or exasperation. Yet if the same behavior is in response to a complaint about a third party (e.g., “can you believe your mother said that to me”), the same eye roll probably would indicate solidarity with the partner. Assessing measures Because sound measurement is so crucial to systematic studies of communication, determining whether measures are appropriate for the purposes at hand is one of the fundamental aspects of the research process. Given that measures usually are merely indications or approximations of constructs, no measure is perfect. Nevertheless, some measures are better than others for particular purposes, and there are a number of basic concepts and considerations that help scholars decide whether a measure is suitable. One key aspect of measurement is conceptualization, which involves thinking about a potential construct and making it clear and precise enough that the concept can be assessed. This involves not only defining what the concept is but also distinguishing it from other potentially related constructs so that any resulting measure can be a good assessment of that construct alone (and not confounded with other constructs). Conceptualization is an inherently interpretive process that is often implicit, but it can be made more explicit and systematic if it is based on a clear theoretical perspective. Operationalization is the process of moving from conceptualization to variables that one can measure. It involves specifying qualities or characteristics that can actually be assessed and determining what will comprise the actual measure of a construct. Although any operationalization will be an approximation of the construct, the goal is to design a measure that captures as much of the breadth of the concept as possible without including indicators that are closely associated with other constructs. Once a measure is developed and used in research, it can be evaluated in terms of reliability and validity. Technically, reliability refers to the probability that if a measure is a concept is taken repeatedly, it will yield the same or a similar value. The desirability of reliability is obvious; if an assessment can yield many wildly different values for the same construct when there is no reason to believe the underlying construct has changed, it would be unwise to trust the results of the measurement. If a thermometer produced temperatures ranging from 96 to 104 degrees Fahrenheit on the same patient in repeated measurements of body temperature, one likely conclusion is that thermometer is broken (that is, no longer reliable). It is theoretically possible, of course, that a human’s temperature could cycle rapidly through extreme temperatures, but if a new thermometer showed steady readings, it would be safe to conclude that the first thermometer is unreliable and not to be trusted. Of course, the case of a thermometer varying by as much as 8 degrees is an extreme one; social scientific measures generally do not have perfect reliability. Some degree of unreliability must be tolerated. The extent to which a measure can show unreliability before it should not be trusted is a judgment call that is informed by some rules of thumb in the research community, but there is no definitively correct answer to the question of exactly how much unreliability is too much for a particular measure in a particular study. Although the notion of reliability involves getting similar results in repeated measurements, reliability is usually inferred in research by comparing values on related variables that were measured at roughly the same time. In a multi-item scale, for example, if all the items are highly correlated with each other or if subsets of the items correspond well with other subsets of items, one can infer that the scale has reliability. If a measure is based on people rating or coding communication behaviors, then the scores from multiple people should correspond highly to indicate reliability in the assessment. Even if a measure is sufficiently reliable, it is not necessarily a good instrument. Reliability is about consistency, and a measure can be consistently wrong. That is, even a measure with high reliability can lack validity. Validity concerns the correspondence between how a construct is conceptualized and how it is operationalized (Levine, 2011). The operationalizations of a construct are incomplete or imperfect reflections of the underlying construct, but the connections between construct and operationalization are based on inferences of varying degrees and soundness (Cappella, 1991). Because operationalization involves a translation of a concept into a measurable indication of the construct, validity is not a binary notion. Measures are not either valid or invalid; instead, they vary in the extent to which they are valid. Evaluating the validity of a measure involves gathering evidence that can allow one to make inferences about validity. For example, one can infer that a measure is sufficiently valid if it is correlated in the expected ways with measures of other constructs or if it is correlated with other measures of the same construct that are thought to be valid enough for particular research purposes (Singleton & Straits, 2005). In addition to comparing a measure to other measures for indications of validity, researchers also attempt to design measures that seek to minimize various factors that are known to threaten validity. One well-known threat to some measures is the set of biases that influence how people recall communication phenomena. Such threats can be minimized (although not necessarily eliminated) by taking steps such as asking people about recent events rather than ones that happened long ago and by phrasing questions in ways that reflect how people perceive phenomena rather than asking them to engage in a unfamiliar mental calculus (Huston & Robins, 1982). A number of textbooks and scholarly articles provide thorough lists and descriptions of common threats to validity and ways to minimize those threats (e.g., Frey, Botan, & Kreps, 2000; Huston & Robins, 1982). Common measurement techniques in interpersonal communication research In response to the fact that measures of communication constructs are imperfect translations of the underlying constructs of interest, researchers have developed a number of different measurement techniques. Each technique has strengths and weaknesses and is suited better for some purposes than others (Feeney & Noller, 2012). Because no technique is perfect and all have different strengths, the ideal is for research to utilize multiple measurement techniques for assessing interpersonal communication constructs (Caughlin & Basinger, 2014; Feeney & Noller 2012). This need not occur in all or even most specific studies, but there is value to using diverse measurement techniques to address important research questions. If findings based on multiple measurement techniques coalesce around similar conclusions, the combined evidence is more compelling than would be possible with strict replications using limited techniques. If, on the other hand, findings are consistently different depending on techniques, it can highlight previously unrecognized questions and insights. It is possible, for example, that one type of measure would reflect one aspect of a theoretical construct whereas another type of measure would be indicative of a distinct facet of the same construct. In such an instance, the systematic examination of a construct with multiple techniques could lead researchers to learn that a concept is more complex or multifaceted than previously understood. Observations One common technique for assessing interpersonal communication is using observations. To obtain measures based on observations, researchers must be able to observe a communication segment of interest, and they must assess what they observe in a way that allows them to derive variables from those observations. Although these two steps can happen simultaneously (e.g., by having judges watch ongoing conversations and rate them on some characteristics), it is more common for researchers to create an audio or video record of the interaction, allowing for a careful assessment of the interactions later. Although there is a wide range of possibilities for observing interaction, observational research frequently involves having participants in a study report to a laboratory and asking them to engage in some kind of interaction. The tasks people are given are intended to stimulate communication that is relevant to the research questions of interest. Scholars interested in relational conflict, for example, probably would find it exceedingly inefficient to simply wait for couples in a laboratory to spontaneously have conflicts; thus, conflict researchers provide instructions that result in a conversation that simulates or reflects conflicts that couples might have in their everyday lives. For instance, countless investigations have been based on having couples identify issues of conflict in their relationship and then discussing those issues in a laboratory. There are a number of advantages of observations in a laboratory. Researchers can control the environment and ensure that the participants produce a segment of interaction that is relevant to the research agenda. It also allows researchers to record the interactions in a way that allows for extremely precise or detailed analysis later. Because of these features, laboratory observations are often considered the “gold standard” for data about interpersonal communication. In spite of the important advantages of observing interpersonal communication in a laboratory setting, such methods also have some drawbacks. Because the usual procedure involves stimulating a conversation, it is not clear how much the conversations are representative of interpersonal communication that occurs spontaneously outside the laboratory. Moreover, in laboratory settings individuals know that they are being observed, which can influence their behavior in a number of ways, such as heightening their desire to manage a positive impression. Reis (1994) has even argued that the typical laboratory setting is so distinct from natural experiences that behavior enacted in such settings is best thought of as exemplary rather than typical. Although not everyone in a laboratory setting engages in the optimal behavior, Reis noted that people do try to optimize their behavior when they know they are being observed, even if they are not always successful at making their behavior seem ideal. The potential issues with laboratory observations are important because so many observational studies occur in such settings, but it is important to recognize that some observations can be made in ways that minimize some of the problems of laboratories. For example, Sillars (1991) reported that in direct comparisons of video recorded laboratory observations to audio recordings done at home, the audio recordings at home were much more “spontaneous and realistic” (p. 206). Moreover, even scholars best known for tightly controlled studies of couples’ conflicts in laboratories have recognized the limits of such studies and have augmented laboratory research with studies in which people were observed in apartment settings for long periods of time (e.g., 24 hours), which diminishes the effects of observation (e.g., Hawkins, Carrere, & Gottman, 2002). Another promising development in observational research on interpersonal communication involves the rise of new communication technologies that people are using for their everyday interaction. Many newer modes of communication are recorded by default. Communication by text, via social media, or by email is often conducted in a manner in which the entire interaction is documented by default. If people are willing to provide researchers access to these records, it can be a rich source of data that was generated entirely apart from the influences that researchers’ observations often have. Of course, this rich new source of observational data has its own drawbacks; for instance, if generated naturally, it may or may not evince the phenomenon of interest to the researchers. Researchers can, of course, give participants tasks that would stimulate them to engage in the types of interactions under investigation, but as soon as they are directed to interact via technologies in particular ways, the interactions become influenced by the research design, just as is the case with traditional laboratory studies. In sum, interaction data can be collected in a variety of ways, ranging in how naturalistic it is and how focused it is on the type of interaction that is of interest to the researchers. Because there are many research questions and purposes, there is no single best way to gather observational data. Scholars may find it useful to use different approaches at different times, even within the same research program. Regardless of how the interaction data are collected, gaining access to such data is only the first step in observational research. Next, the researchers must find some way to code or rate the data so that variables can be drawn from the conversations. As with any measurement issue, the researcher is usually interested in constructs that might be evident in the data rather than an overt behavior in and of itself. For example, a researcher studying conflict might be interested in the construct of avoidance. The notion of avoidance is not something that can be directly observed, but there are behaviors that scholars have inferred reflect avoidance during conflict. For instance, if a person changes the subject during a conflict discussion, it often indicates that this person is trying to avoid the topic. This is a reasonable inference, but notice that it is still an inference and not a direct measure of avoidance. Because the meaning and importance of communication behaviors are contextual, one can make mistakes assuming that there is a direct one-to-one correspondence between communication constructs and behavioral indicants of those constructs. For instance, Sillars (1986) noted that in some conversations about potential conflict topics, there are circumstances in which the conflict has already been resolved or is simply not an issue for certain couples. In such cases, changing the subject simply reflects a lack of need to talk about that issue, and it would be a mistake to call that conflict avoidance. It is important to be cognizant of the interpretive aspect of observational assessment of interpersonal communication. Because observational records of data allow for objective counting of the indicants, some scholars assume that observational measures are also objective. If a researcher is truly only interested in the directly observable manifestation of communication (e.g., how many times a person changes the subject, how many times a person rolls his or her eyes), then the measure can fairly be called objective. Yet, typically researchers wish to study underlying constructs (e.g., conflict avoidance, expression of contempt) that are not directly observable and therefore must be inferred from indicants. Once such inferences are required, the measure is no longer entirely objective, and it is important to scrutinize the extent to which the inferences are sound. A careful reader of observational studies will notice that sometimes the inferences researchers (or other readers) make about the meaning of behaviors are actually quite debatable. Moreover, because the meaning of communication behaviors is contextual, there is always some risk in making inferences about what those behaviors mean from an outside perspective, particularly when the interpersonal communication is between longtime relational partners who have had time to develop idiosyncratic patterns of interacting and a history of discussing issues that imbues the discussion of current topics with meaning developed over previous encounters. In short, for a variety of reasons, one should not confuse collecting an objective sample of interaction data with being able to create objective measures of interpersonal communication. Retrospective self-reports. Retrospective self-reports are another common method for gathering data about interpersonal communication. In this technique, research participants are asked to reflect on communication events or experiences that have already occurred and then provide information about those occurrences. Retrospective self-reports can be used to gather data about a wide variety of phenomena, including reports about the frequency or quality of one’s own or another person’s behavior, the thoughts or experiences one had during communication episodes, and evaluations of communication or the people involved in a conversation. Retrospective reports are also quite flexible in terms of timeframe. Researchers have asked participants to report on events that happened recently as well as ones that happened long ago; for instance, a number of studies have asked adults to report on the communication patterns that they experienced in their family of origin when they were growing up. Researchers also have asked about very specific episodes (e.g., a memorable conversation) as well as broad patterns of communication that occur over longer periods of time (e.g., family communication patterns). Retrospective reports, in short, can be used to assess a very wide range of variables related to interpersonal communication. The flexibility and range allowed by retrospective self-reports is a key advantage of this technique. There are many important communication phenomena that are difficult to observe, and many more that are altered so much by being observed that it is not clear what the observations actually indicate. Research using retrospective self-reports in the two hours following sexual activity, for example, has shown that women who report experiencing orgasm report disclosing more about themselves afterwards than do women who do not report orgasm (Denes, 2012). Although the same research questions could conceivably be examined with observational methods, there would be numerous practical obstacles; for example, the number of people willing to participate in a study that observed sexual activity as it unfolds (especially one that included assessments of orgasm other than self-reports ones) may be so limited that one would wonder about the generalizability of the results. Even if one did find a representative sample of participants willing to be observed, it is likely that the sexual activity would need to be planned in such a way that it would introduce concerns about whether the activity was representative of sexual activity. Even apart from the sexual activity, there are reasons to believe that the self-disclosures would be altered if participants knew that their disclosures were being recorded. Yet it is possible to gather self-report data about such phenomena, especially if anonymity or confidentiality is assured. According to Feeney and Noller (2012), “the limitations of self-report questionnaires are well known (in fact, they have been more widely acknowledged than the limitations of other methodologies such as observation)” (p. 30). Most often, scholars focus on problems that can lead to inaccurate or biased responses, such as those caused by the inability to remember what happened or the desire to make a good impression. There are also known and systematic errors in judgment that lead people frequently to be more negative in their reports of others’ behaviors than they are with reports of their own behaviors. Such limitations are inherent to all self-report measures. Yet just because self-report measures are inherently limited does not mean that they are all equally problematic. There are certain threats to validity that can be minimized—or amplified—when using self-report measures. The ability to report accurately on communication behavior can be diminished in a number of circumstances, such as when participants are asked to report on events from long ago, when the behaviors they are asked to report on or the metrics used for the items are vague and unspecific, or when answering the question requires mental calculations that deviate from how participants spontaneously think about such events. Consider, for instance, the difference between the following two questions: “Did you send a text message to your partner since you came to the lab five minutes ago?” and “Over the past year, how often did you communicate with your partner each week on average?” The former question asks about a very specific behavior, a short and recent timeframe, and does not require any calculations. Unless they have some reason to overtly lie to the researchers, most participants could answer the former question accurately. The latter question, in contrast, requires participants to think back over the past year, reflect on an average across any waxing and waning in communication, and figure out how to define what the word often means (e.g., if they see each other twice daily but send hundreds of texts back and forth all day, how many times have they communicated?). It is unlikely that this question would be a particularly accurate reflection of how many times the participants communicated no matter how the researchers were defining what counted as communicating. Because retrospective self-reports vary considerably in how biased they are, it is important to examine the specific measures and goals of the research before deciding whether a retrospective report is appropriate. For instance, there are times when researchers believe that people’s perceptions of their communication behaviors are more important than an accurate count of such behaviors. Moreover, if the communication phenomenon is one that cannot be easily inferred from observable behaviors, a self-report can be more accurate than an observational measure. If one is studying the keeping of secrets, for example, it may be possible to infer that some observable behaviors indicate that someone is keeping a lot of secrets, but those behaviors undoubtedly would be open to multiple interpretations and would be influenced by context. In contrast, an individual asked to report on secret keeping needs to make far less challenging inferences to be able to provide data on the subject. To put this another way, one strength of many retrospective report measures is that the inferential leap between indicants (such as items on a scale) and a construct can be much smaller than is the case with observational methods. Thus, retrospective measures can address one of observational measures’ biggest weaknesses, just as observational measures address the most obvious limitation of retrospective measures: bias. Because they have complementary strengths and weaknesses, retrospective self-reports and observations often can be used productively together to provide different viewpoints on important research questions. Diaries and logs Diaries and logs are self-report measures, but they are usually treated as a separate type of measure because they involve getting reports of the same experiences or behaviors soon after they occur (or even during), and the reports are gathered repeatedly over time. Diary and log data tend to refer to relatively short periods of time that are presumably accessible to participants. For example, rather than asking about behaviors over a two week period in general, a diary approach could entail having participants report on what happened on a particular day and then repeat the diary each day for two weeks. Because participants are being asked to consider a fairly short and recent time period, some of the biases associated with retrospective self-reports are minimized. Not all of such biases are reduced or eliminated, however. If the questions are about stigmatized behaviors, for instance, there is no reason to believe that people would be less likely to obfuscate in diary reports than they are with retrospective ones. By aggregating across days, researchers can create summary variables that are conceptually equivalent to retrospective reports of frequency but are probably more accurate because the recall time is minimized and because the researchers do the calculations rather than asking participants to make estimates such as the average number of times per day that an event occurred. Moreover, collecting data on the same experience repeatedly allows for an examination of questions that are difficult to assess with retrospective reports. For example, diary studies have shown that the patterns of marital interaction vary according to the day of the week, but it is not clear that such temporal rhythms would be apparent to participants, even if the researchers could pose clear retrospective questions about such temporal rhythms. Physiological measures Perhaps the fastest growing category of assessment technique in the study of interpersonal communication is physiological measurement. Because there is a separate entry on such measures, they are covered only briefly here to contextualize them within other types of measures. There is a wide variety of possible physiological measures, ranging from cardiovascular characteristics, hormones, brain activity, and so forth. At least one reason for the rapid rise in the use of physiological measures is the technologies to assess physiology have improved greatly in terms of lower cost and intrusiveness. One of the biggest advantages of physiological measures in general is they can provide data about communicators that would be very difficult for participants to report on or to observe without biological markers. For instance, cortisol levels and other hormones can indicate physiological stress, even when participants report not experiencing stress and there are no clearly discernable visible indicators of stress. Because they provide information about individuals who are communicating that cannot be discerned by other means, physiological measures are likely to continue becoming more prominent over time. Yet like any type of measure, it is important to keep in mind what such measures can and cannot do. Writing specifically about fMRI data, Aue, Lavelle, and Cacioppo (2009) argued that scholars often are too extreme in their views of physiological data, with some dismissing it entirely and others believing it will reveal all the mysteries of psychological experiences. In reality, physiological measures are very much like observations of behavioral data. The information that is gathered is different, but it usually is gathered directly by researchers, rather than via reports from participants. The link between the assessments and broader theoretical constructs is one that must be inferred by the researchers. Thus, although the indicators in physiological data are different than with traditional observational data about communication, some of the strengths and weaknesses are analogous. Physiological assessments can be made without respondent biases, and this lack of bias in the assessment itself creates a sense of objectivity to the research. Yet what the assessments actually mean must be inferred, and in many instances the inferences are unclear, or at least not as unequivocal as implied by some researchers (Aue et al., 2009). A particular physiological response, for instance, may be an indicator of a particular experience, but that same response often can indicate other experiences as well. That is, the connection between physiological data and what it actually means to people and their communication is not simplistic, which makes the interpretation of such data complex. Researchers can address some of these challenges by carefully examining their inferences to make sure they are as sound as possible; for instance, inferences can be improved by examining multiple physiological indicators together and by recognizing that the connection between physiology and psychology is not “an invariant relationship” (Aue et al., 2009, p. 13). Key challenges to measuring interpersonal communication The discussion to this point has focused on various measurement techniques that are common in the study of interpersonal communication. Given that each has strengths and weaknesses and that the strengths often complement each other, one obvious implication is that multiple types of measures should be used to address important questions in interpersonal communication research. Beyond the consideration of classes of measures, however, there are some larger conceptual issues that are important to consider. Implications of new communication technologies It is now widely recognized that interpersonal communication happens frequently via mediated channels. Whether texting, talking on mobile phones, or using social media, it is undeniable that some people frequently use technologies as they interact with people close to them. For instance, people now utilize communication technologies in a variety of ways during their relational conflicts (Caughlin, Basinger, & Sharabi, in press). Some individuals claim that they look up information on their smartphones to help resolve differences; others report that they consult their phones for evidence in text messages that can prove their point (e.g., “see, you DID say that you would definitely be there”). Some individuals claim that by looking at a phone or other screen, they are able to gather their thoughts so they can say the right thing during conflict; others say they look at their phones to avoid having to look at the other person. Although we are just beginning to understand what the ubiquity of such devices means for interpersonal communication, it is clear that many people are using them as tools for communicating, and this has potentially important implications for assessing interpersonal communication. The traditional “gold standard” for studies of conflict, for instance, has been to have dyads come to a laboratory and discuss conflict issues. Whether explicitly proscribed or not, the nature of the studies makes it exceedingly unlikely that people would incorporate their communication technologies into their discussions. Yet if that is something they typically do when conflicts arise, not encouraging them to have their communication tools makes the traditional laboratory study even less naturalistic to these participants than it was before such communication devices became widespread. Because communication technologies have become part of interpersonal communication, our assessments of interpersonal communication need to do more to take this into account. Although we still know very little about the use of communication technologies in everyday interpersonal communication, the lessons of research on observations of other behaviors provide some insight into what we should consider when trying to assess the use of communication technologies in interpersonal communication. We know, for example, that the connection between behaviors and meaning is not one-to-one, and we would expect that the same would be the case for the use of technologies. For example, understanding the role of the communication technologies is likely to involve more nuanced conceptualization than would be implied by simply counting behaviors, such as how often one looks at a smartphone during a discussion. As the examples above illustrate, one can look at a phone to facilitate attempts at constructive conflict, one can look at a phone to gain evidence for a distributive strategy, or one can look at the phone to avoid the conflict. It would be fairly easy to count an objective behavior such as how often a person looked at a smartphone, but such counts probably would not tell us much without careful consideration of what that behavior means (and can mean). Consequently, as with other measures of interpersonal communication, before researchers start counting behaviors related to technology use, it will be important to do the theoretical work that helps identify key constructs. Measure selection and justification Measures are never inherently valid; they are instead more or less valid for a given purpose. As such, evaluating the relative validity of a survey measure, coding scheme, or any measure requires at least implicitly evaluating an argument about the suitability of the measure for a specific purpose. A word of caution is warranted, then, about the selection and/or adaption of measures for one’s own study that were designed for other purposes. For instance, psychologists have designed a number of assessments of personality or relationship traits that are of interest in interpersonal communication research (e.g., narcissism, depression). Although these measures may be valid for clinical assessment, they may not be appropriate for non-clinical samples. Or, measures developed to assess relationship characteristics between spouses may not make sense to use in studies of parent-child relationships. In cases like these, the importance of face validity, carefully assessing whether a measure actually seems to capture the concept of interest (and nothing else), cannot be overestimated. There must be a logic behind the selection of measurement tools beyond their prolific use in previous research. A separate, but related issue, is that researchers sometimes operationalize the same concept in multiple ways. Conflict, for example, may be operationalized by assessing frequency, intensity, management, satisfaction, or a number of other characteristics of conflict-laden interactions. Certainly, each of these may be “valid” representations of conflict, but it is worth noticing that if different scholars are conceptualizing the same word in different ways, they might actually be assessing slightly different constructs or different aspects of a larger construct. Thus, even when different scholars purport to be assessing the same construct, it may not be possible to compare the results of the various studies because the conceptualizations or operationalizations may be vastly different. Consequently researchers should be careful about which measure of a construct they choose; a case should be made for choosing one assessment tool over the other, based on the purpose of the study. The goals of one’s own research and the purpose of the measure in question should always be of primary interest in making an argument for the validity of a measurement tool. Another frequently cited piece of evidence for validity involves statistical tools. Although communication researchers tend to report and interpret statistics in similar ways, our rules of thumb for providing “proof of validity” should not be followed blindly. Rather, researchers should pay attention to the theoretical meaning behind their statistically-based decisions. There is sometimes a good theoretical justification for retaining items in a factor that do not load cleanly with other items or for separating items that do correlate highly. For instance, Thompson and Walker (1982) have pointed out that spouses’ attitudes generally correlate highly, but they should still be treated as separate scores because attitudes are individually formed, rather than characteristics of a couple. Statistical tools sometimes provide us with insight into our data, but basing decisions purely on the math behind these tools may lead us to make illogical or invalid decisions. In sum, the goal in the selection of a measurement tool ought to be related to the purpose of the study, and the argument for the validity should include a balance of reason, theory, research, and available statistical analyses. SEE ALSO: Attributions in Conflict; Conflict Avoidance; Conflict Styles and Strategies; Family Typologies; Measurement Bias; Physiological Measurement References Aue, T., Lavelle, L. A., & Cacioppo, J. T. (2009). Great expectations: What can fMRI research tell us about psychological phenomena? International Journal of Psychophysiology, 73, 10-16. doi: 10.1016/j.ijpsycho.2008.12.017 Cappella, J. N. (1991). Mutual adaptation and relativity of measurement. In B. M. Montgomery & S. Duck (Eds.), Studying interpersonal interaction (pp. 103-117). New York, NY: Guilford Press. Caughlin, J. P., Basinger, E. D., & Sharabi, L. L. (in press). The connections between communication technologies and relational conflict: A multiple goals and communication interdependence perspective. In J. A. Samp (Ed.), Communicating interpersonal conflict in close relationships: Contexts, challenges and opportunities. New York: NY: Routledge. Caughlin, J. P., & Basinger, E. D. (2014). Measuring social interaction. In In P. J. Schultz and P. Cobley (Series Eds.) & C. R. Berger (Vol. Ed.), Handbooks of communication science: Vol 6: Interpersonal Communication. (pp. 103-126). Berlin, Germany: De Gruyter Mouton. Denes, A. (2012). Pillow talk: Exploring disclosures after sexual activity. Western Journal of Communication, 76, 91-108. doi: 10.1080/10570314.2011.651253 Feeney, J. A., & Noller, P. (2012). Perspectives on studying family communication: Multiple methods and multiple sources. In: Anita L. Vangelisti (Ed.), Routledge handbook of family communication (2nd ed., pp. 29-45). New York, NY: Routledge. Frey, L. R., Botan, C. H., & Kreps, G. L. (2000). Investigating communication: An introduction to research methods (2nd ed.). Boston, MA: Allyn and Bacon. Hawkins, M. W., Carrere, S., & Gottman, J. M. (2002). Marital sentiment override: Does it influence couples’ perceptions? Journal of Marriage and Family, 64, 193-201. doi: 10.1111/j.1741-3737.2002.00193.x Huston, T. L., & Robins, E. (1982). Conceptual and methodological issues in studying close relationships. Journal of Marriage and the Family, 44, 901-925. doi: 10.2307/351454 Levine, T. R. ( 2011). Quantitative social science methods of inquiry. In M. L. Knapp & J. A. Daly (Eds.), The SAGE handbook of interpersonal communication (4th ed., pp. 25-57). Los Angeles, CA: Sage. Reis, H. T. (1994). Domains of experience: Investigating relationship processes from three perspectives. In R. Erber & R. Gilmour (Eds.), Theoretical frameworks for personal relationships (pp. 87-110). Hillsdale, NJ: Erlbaum. Sillars, A. L. (1986, April). Procedures for coding interpersonal conflict (revised) (Manual). Missoula: University of Montana, Department of Interpersonal Communication. Sillars, A.L. (1991). Behavioral observation. In B.M. Montgomery & S. Duck (Eds.), Studying interpersonal interaction (pp. 197-218). New York, NY: Guilford Press. Singleton Jr., R. A., & Straits, B. C. (2005). Approaches to social research (4th ed.). New York, NY: Oxford University Press. Thompson, L. & Walker, A. J. (1982). The dyad as a unit of analysis: Conceptual and methodological issues. Journal of Marriage and the Family, 44, 889-900. doi: 10.2307/351453 Zeller, R. A., & Carmines, E. G. (1980). Measurement in the social sciences. New York, NY: Cambridge University Press. Further reading Baesler, E. J., & Burgoon, J. (1987). Measurement and reliability of nonverbal behavior. Journal of Nonverbal Behavior, 11, 205-233. doi: 10.1007/BF00987254 Metts, S., Sprecher, S. & Cupach, W. R. (1991). Retrospective self-reports. In B. M. Montgomery & S. Duck (Eds.), Studying interpersonal interaction (pp. 162-178). New York, NY: Guilford Press.