Measuring Social Interaction

1 Measuring Social Interaction John P. Caughlin and Erin D. Basinger University of Illinois at Urbana-Champaign 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. (doi: 10.1515/9783110276794.103) Final Draft: This version may differ from published version due to copy editing. 2 Abstract This chapter concerns the measurement of concepts related to interpersonal communication. We begin with a discussion of general measurement principles and argue that there are multiple useful ways to conceptualize and measure interpersonal communication. We then review strengths and weaknesses of common measurement techniques, including selfreports, observations, in-depth interviews, and physiological measures. Because every measurement technique has limitations with respect to assessing interpersonal communication, we argue that it is often useful for multiple techniques to be used either within or across studies. We conclude with a discussion of special considerations for designing measures in studies of interpersonal communication. Specifically, interpersonal scholars should pay particular attention to: (a) the need to consider a timeframe that provides an adequate sampling of ongoing interaction; (b) the possibility that new communication technologies are changing the nature of interpersonal communication and what should be observed, even in face-to-face settings; (c) the pitfalls of using measures designed for other purposes to assess interpersonal communication constructs, and (d) the potential for confusing statistical information for definitive proof of validity. Key Words: research methods, interpersonal interaction, validity, reliability, measurement, operationalization 3 Chapter 5. Measuring Social Interaction 1. Introduction Measurement is “the process of determining the existence, characteristics, size, and/or quantity of changes or differences in a variable through systematic recording and organization of the researcher’s observations” (Frey, Botan, and Kreps 2000: 83). The choices we make about how to measure social interactions determine not only the quality of our data, but also the conclusions we can reasonably draw because our methods have assumptions built into them (Duck and Montgomery 1991; Reinard 2008). In fact, Levine (2011: 44) asserts that “the path to verisimilitude in quantitative research always goes through measurement.” Given the significance of measurement in social interaction research, this chapter focuses on issues surrounding the measurement of interpersonal communication concepts. Unfortunately, there are no simple guidelines for creating ideal measures of interpersonal communication. Part of the difficulty in measuring interpersonal communication concepts is that there is no consensus about what even counts as interpersonal communication. Over the past several decades, there have been periodic calls to focus on observable behaviors exchanged between people (Knapp and Daly 2011: 12). Implicit in these calls is the idea that one cannot fully know what happened in social interaction just by examining what people think about it. Yet, understanding the significance of those observable behaviors may depend on knowing something about the interactants’ expectations, plans, and interpretations, as well as how those behaviors fit into the history of interactions between the people involved. People involved in interaction often interpret the meaning of such interactions in terms of each other’s plans or goals (Berger and Palomares 2011; Wilson 2002); for example, rather than focusing on the fact that a partner stated “that is your third drink,” a person may describe the partner’s message as “she 4 wants me to stop drinking after this one.” If one focuses only on the observable behaviors and not the meanings attributed to them, would what is communicated be apparent to observers? This query implies that there is no simple answer to the question of whether it is best to conceptualize social interaction as something that happens apart from such inference processes (and is therefore fully observable) or whether such inferences are an inherent part of what it means to have social interaction. Scholars rarely take explicit positions on such issues, but their implicit stances shape what is presumed to be a valid measure of interpersonal communication. For example, some studies use the phrase “actual communication” synonymously with “behavior,” which conceptualizes all the interpretive aspects of interaction as apart from communication (e.g., see Le Poire and Yoshimura 1999). In such studies, what would be considered the most valid communication measures would be different from studies in which communication is conceptualized as inherently involving the meanings of such behaviors to the communicators. Given that communication scholars implicitly disagree about the very nature of what constitutes interpersonal communication, there cannot be complete consensus about the best measurement techniques. Thus, rather than espousing one set of values about measurement, our goal is to familiarize the reader with important issues and conceptual problems that can inform better and worse research practices. That is, even though there are no correct and incorrect ways to measure interpersonal communication, there are nevertheless more and less appropriate choices based on the particular purposes of a given study. Our goal is to make some of the implicit assumptions about measurement more explicit and to encourage a reflective stance toward choosing measures of interpersonal constructs. Toward that end, we begin this chapter by discussing concepts in general and reviewing some traditional measurement concepts and 5 techniques. Then, we consider the use of multiple measurement techniques in assessing social interaction, and we conclude with a discussion of special considerations for studies of interpersonal communication. 2. Concepts as arbitrary The challenge of not having a definitive consensus about what constitutes interpersonal communication is compounded by the arbitrary nature of interpersonal constructs. Philosophers have long recognized the arbitrary relationship between concepts and the things they represent (Ogden and Richards 1923). Among communication researchers, a range of beliefs about this relationship is represented. Baxter and Babbie (2004: 111) argue that “concepts are only mental creations.” The problem, they say, is that we begin to attach real meaning to our concepts anyway, leading us to measure them in ways that are inaccurate. In addition, Baxter and Babbie (2004: 132) argue that we can “measure anything that exists,” but warn that some of the things we want to measure are concepts that we have just agreed upon as having meaning. In contrast, Reinard (2008) proposes that there are some concepts that we cannot fully capture in measurement because they are too abstract or because they are mental experiences. Cappella (1991) similarly claims that our measurement tools cannot be neutral; rather, they are constructions of the social world. Representing a different perspective, Surra and Ridley (1991) caution that communication has subjective and idiosyncratic meaning, as well as normative and conventionalized meaning, and that our measures ought to be sensitive to these differences. Although each scholar represents a different perspective, all of them recognize that the relationship between measures and the things they represent is somewhat subjective. Even though all constructs relating to interpersonal communication are somewhat arbitrary, that does not mean all constructs and measures thereof are equally useful. O’Keefe 6 (1987) presents a useful way of approaching the problem of arbitrary concepts in measuring interpersonal interaction. He suggests that messages are complex and that there are many concepts that could be used to analyze them; thus, there is no intrinsically correct way to describe a message or segment of social interaction. This adds weight to the choices that researchers inevitably make as they select a system of measurement from an enormously large number of possibilities. O’Keefe draws two conclusions that are particularly relevant to social interaction measurement. First, it is unlikely there can ever be a general interaction coding system that is appropriate or accurate for every given purpose; what may be a perfectly valid measure for one purpose may be invalid for another. This seems to be an obvious point, but it is common to encounter claims that the validity of a key measure has been established in prior studies, even when a close examination reveals that the measure is being used to assess a construct for which it was not originally intended and seems ill-suited. Second, because we make choices each time we measure something, our measurement tools cannot be valid or invalid in an overall sense. Rather, a measure is valid or invalid only with respect to specific purposes. 3. Traditional concepts of assessing measurement Researchers can make a large number of choices when they measure concepts; however, there is an accessible list of traditional measurement tools and concepts. Although this list is not exhaustive, it is a useful starting point for understanding measurement basics. Because there are many available references for these basic measurement concepts, we provide a very brief review here (for more extensive discussions, see Baxter and Babbie 2004; Frey, Botan, and Kreps 2000; Singleton and Straits 2005). 3.1. Conceptualization and operationalization 7 A fundamental part of measurement is considering a concept carefully and recognizing what is and is not part of that concept. Conceptualization is a mental process of developing concepts and making them more precise. Operationalization is the process of translating concepts into measurable variables and specifying their assessable characteristics. Although conceptualization and operationalization apply to any systematic attempt to measure a construct, there are some particular challenges for scholars studying interpersonal communication. For example, because two people in an interpersonal encounter often influence each other, their behaviors and thoughts can be intertwined; consequently, researchers must think carefully about questions of the proper unit of analysis, such as whether a construct is best thought of as a property of the individual or the dyad (Thompson and Walker 1982; see Chapter 6, Liu). Similarly, interpersonal communication scholars may wish to consider communicators’ subjective perceptions, as well as some of their more objective behaviors. Given these (and many other) unique constraints of studying interpersonal communication, scholars ought to be particularly sensitive to their conceptualization and operationalization processes. 3.2. Reliability and validity Reliability refers to the stability or consistency of a measure and whether a particular measure, administered repeatedly, will yield the same results each time. Often reliability is assessed at a single point in time by examining various indicators of the same construct, with the idea that if multiple items or ratings show similar findings, then the overall measure is probably reliable. As we discuss below, however, concurrent measures of reliability do not always provide a good sense of whether a measure of interpersonal communication behavior would yield similar findings if repeated multiple times. 8 Validity involves the congruence between the conceptual and operational definitions of a concept (Levine 2011). Having true validity in any given study means fully reflecting the construct of interest and nothing else (Levine 2011). Measures can only be valid to a certain extent and for a specific purpose because each representation (i.e., measurement) of behavior, communication, or interaction is less complex than the behavior, communication, or interaction itself (Cappella 1991). The translations of those things are what we actually study, and they are more or less inferential depending on how we choose to measure them (Cappella 1991). Assessing validity, then, must necessarily be a complex process that is unique to each particular study and its purpose. Validity cannot be assessed directly. Researchers must either subjectively evaluate whether the operationalization assesses the intended concept or compare the results of a measure to others it should relate to (Singleton and Straits 2005). Moreover, validity is not a binary construct (Levine 2011). That is, a measure is not either valid or invalid; rather any given measure reflects a range of validity. As a result, validity can be threatened in a number of ways including inappropriate sampling, memory distortions affecting recall of events, errors in mental processes by the participant, distortions in judgments, and lumping conceptually distinct constructs together (Huston and Robins 1982). This list of threats is not exhaustive, so we point the reader to Huston and Robins (1982) and Frey, Botan, and Kreps (2000) for a more thorough review of validity threats. 3.3. Assessing measures Considerations of operationalization, reliability, and validity should contribute to the decision of whether a measure is suitable for a specific study. That is, is the operationalization adequate, accurate, and clear (Frey, Botan, and Kreps 2000)? Is the unit of analysis appropriate 9 for the concept of interest? Is the measure reliable? Is it valid for the purpose of this study? It is important to emphasize that the answer to whether a measure is appropriate really does depend on the specifics of a given study. For instance, a measure that is perfectly valid for one sample may not work well with another sample (Reinard 2008). 4. Common measurement techniques in interpersonal communication research There are many options for measuring interpersonal communication constructs. In this section we discuss the common general techniques that interpersonal communication researchers use. Because our goal is to provide an overview, we do not focus on the considerable variation that exists within these categories of measures. For a more thorough discussion of various pertinent measurement techniques see Feeney and Noller (2012). 4.1. Self-Reports 4.1.1. Retrospective self-reports The most common self-reports involve participants reflecting on and answering questions about their lives or experiences. Researchers have asked people to report on many aspects of interpersonal communication, including their own behaviors, other people’s behaviors, their attributions for their own or other people’s behaviors, and various subjective evaluations of the communication and other communicators. In addition to being used to assess a wide range of constructs, retrospective self-reports can be used to assess interpersonal communication behaviors over various timeframes; for example, Young et al. (2005) asked about a single hurtful experience of family communication whereas Vangelisti and her colleagues (2007) investigated factors that led to a general environment of hurtful family communication. Taken together, these studies point to the range of uses of retrospective self-reports. 10 Given that the focus of self-report measures can vary so widely, scholars should attend closely to exactly what participants were asked when generating their responses. Often measures that are ostensibly assessing the same construct differ in some significant attribute that makes the findings not comparable. For example, Larson and Chastain’s (1990) measure of concealment asks about the tendency of certain individuals to be secretive, and secrecy on this measure is inversely related to relational quality (e.g., Finkenauer et al. 2009). However, that measure of secrecy is not equivalent to those used in other studies that examine people’s decisions and experiences regarding a particular secret (e.g., Caughlin et al. 2005). It is important to keep such differences in mind because even though the general topic of the measures may be the same, the details of the constructs assessed may differ enough that one should be very cautious about making broad claims based on any particular measure. For example, results indicating that being a generally secretive person is associated with dissatisfying relationships do not imply that people in relationships should reveal any particular secret that they have (Caughlin, Petronio, and Middleton 2012). The potential pitfalls highlighted in this example should caution researchers against making overly broad claims based on specific measures. 4.1.2. Diaries or logs Diaries and logs are forms of self-report measures, but they differ from typical retrospective self-reports in that they involve repeated reports of the same behaviors or experiences over some sample of time. The goal of such measures is to study individuals’ everyday experiences, including those with interpersonal relationships and communication (Reis and Gable 2000). Depending on the research purposes, diary or log measures can be very broad or more specific. The Rochester Interaction Record (Reis and Wheeler 1991), for example, involves asking research participants to record detailed information about every interpersonal 11 encounter lasting at least ten minutes. Other studies focus on a particular relationship; for example, in the PAIR Project (e.g., Huston et al. 2001) married couples were called on the phone on multiple occasions and asked about daily occurrences of specific behaviors, such as whether one’s spouse had complained or said “I love you.” Data from diaries or logs are often aggregated to provide baseline information about the frequency of interpersonal behaviors or experiences, but they also can be examined for temporal patterns, such as whether marital interaction patterns are influenced by the day of the week (Huston, McHale, and Crouter 1986). 4.1.3. Advantages and disadvantages of self-report measures As Feeney and Noller (2012: 30) noted, “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)” (also see Metts, Sprecher, and Cupach 1991). The most notable problems with self-report measures are that people may be biased in their reports (e.g., due to social desirability or recall errors) and that there are aspects of interpersonal communication that individuals may not even be aware of, such as many nonverbal behaviors (Baesler and Burgoon 1987). These are necessary issues to consider when using self-report measures, so it is good that they are widely recognized among communication scholars. Yet, scholars should not dismiss self-report measures just because they, like all measures, have limitations. Moreover, just because self-report measures are subject to various biases does not mean that every self-report is equally biased. Instead, it is useful to think about the various threats to validity with respect to particular measures. What people are asked to report on, how questions are worded, the timeframe being assessed, and the accessibility of the information all influence the extent to which a self-report can be trusted (Huston and Robins 1982). For example, if a questionnaire asks an individual to report on a relatively long period of time and do 12 mental calculations, it is likely that the report will be more problematic than one that asks about a recent and narrow timeframe and does not require complicated assessments. To the extent that the recall interval is short and the behavior is easily observable to participants, the report is more likely to be accurate. If a researcher called and asked readers of this chapter if they were reading a book chapter right now, they could probably answer that question accurately. In short, not all self-report measures should be considered equally problematic; it is important to consider what exactly participants are being asked and whether they are in a position to provide accurate information. The main advantage of self-report data is that they provide access to information that would be difficult, if not impossible, to observe. Self-reports can be used to gather information about individuals’ cognitions, such as whether they found a conversation enjoyable, their beliefs about why other people said what they did, their beliefs about why they said what they did, and so forth. Diary and log versions of self-reports can be particularly useful at learning about interpersonal communication phenomena in everyday life. For example, a typical observational measurement strategy involves asking dyads to engage in a particular type of interaction (e.g., a conflict), yet such procedures essentially control or eliminate the extent to which dyads vary in how often they engage in that type of interaction. It might be possible to design an observational technique that allowed researchers to independently assess the frequency of everyday interpersonal behaviors, but such a measure would require a level of surveillance that probably would preclude the recruitment of a representative sample typical dyads. A self-report measure, on the other hand, opens up possibilities for assessing these types of interactions by exploring individuals’ cognitions about a particular type of interaction, as well as its frequency. 4.2. Observations 13 Interpersonal communication can be observed in a number of ways. One common technique for doing so involves bringing people into a laboratory and asking them to engage in some sort of discussion. For example, one could observe married couples discussing conflict issues based on a list of topics researchers select for their discussion (e.g., Gottman 1994; Sanford 2012). Although conflict tasks are by far the most common stimulus for interactions in laboratories, researchers also use other tasks, such as asking partners to comment on each other’s positive or negative qualities (Smith et al. 2011), asking partners to be supportive of each other (Sullivan et al. 2010), or instructing partners to discuss their positive feelings for each other (Graber et al. 2011). Observational studies conducted in laboratory settings are so prominent in the research literature that some researchers have suggested that observing interaction is nearly synonymous with a laboratory or some other artificial setting (e.g., Reis and Gable 2000). However, it is possible to observe naturally occurring interpersonal interactions. Indeed, discourse analysts have observed face-to-face interpersonal interaction during workplace interactions, police interviews, medical encounters, and so forth (Beavin Bavelas, Kenwood, and Phillips 2002). Interpersonal communication sometimes takes place via media that create artifacts that can be observed, such as letters exchanged between relational partners, internet bulletin board discussions, or logs of instant messaging interactions (Beavin Bavelas, Kenwood, and Phillips 2002). Thus, there is a myriad of social contexts in which observational data can be obtained. Regardless of how observational data are gathered, collecting the sample of interaction is only the first step. The researcher must decide which specific behaviors to examine and how to operationalize variables pertaining to the constructs of interest. In some cases, researchers count the frequency of discrete behaviors. In other cases researchers rate the extent to which a segment 14 of talk exemplifies some construct; for example, Christensen and Heavey’s (1993) coding scheme for demanding and withdrawing behaviors relies on ratings of the extent to which interactants are demanding or withdrawing. Sometimes the researcher is interested in particular behaviors, but other times the sequence of the behaviors is also considered important; for instance, Gottman and his colleagues (1998) examined how husbands responded to their wives when the wives expressed modest amounts of negative affect. When researchers study sequences, the implicit theoretical assumption is that the combination of individuals’ behaviors reveals something that cannot be discerned just from the frequencies of each person’s behaviors. 4.2.1. Advantages and disadvantages of observational measures Observational methods are often touted as a means of gathering objective information about actual interpersonal communication. The utility of objective assessments of interpersonal communication is unquestionable given the fact that self-reports of communication may be biased. Despite this obvious strength, however, observational methods also have weaknesses, which are often overlooked or unrecognized, perhaps because so much attention has been paid to the weaknesses of self-report measures (Feeney and Noller 2012). First, the totality of the observed interaction can never be presented in a scholarly manuscript. Instead, researchers must choose something to examine in the data and then interpret what those observations mean. Clearly, choosing what to assess can be a selective process. Researchers often address this potential bias by using established coding systems such as the Specific Affect Coding System (SACS; Gottman 1994). Although choosing an existing coding or rating scheme is useful, it is important to keep in mind that reliance on existing ratings can also introduce bias because it focuses the researchers’ attention on certain aspects of the interaction (and away from others), which is problematic if the rating scheme was not originally developed 15 for precisely the purposes of the current study. Moreover, as a rating or coding scheme becomes firmly established, there may be a tendency for researchers to reify the variables derived from it, forgetting that the particular ratings extracted from the interaction are just one of many possible useful ways of examining the data, and that each particular way was undoubtedly shaped by the original researchers’ views and goals. Given that the creation of such schemes is an inherently interpretive process, all rating schemes have a particular perspective. In other words, the data gathered may be objective, but the process of reducing that data into information that can be studied systematically is inherently an inventive one and therefore biased in some ways. Another form of bias inherent in observational research involves the interpretation of the rated interactions. In one study, Gottman et al. (1998) followed 130 newlywed couples over time and reported: “only newlywed men who accept influence from their wives are winding up in happy and stable marriages.” This conclusion was based on observational data that was coded with the SACS and analyzed based on sequences to determine how partners responded to each other. This study was interpreted as meaning that husbands who want happy marriages should do what they are told, and not surprisingly, this study received a great deal of attention in the popular press (for a representative example, see Maugh 1998). Yet a careful reading of exactly what was observed and coded in the Gottman et al. (1998) study reveals multiple possible interpretations of the “accept influence” variable aside from husbands doing what they are told. What the study actually assessed was husbands’ responses when their wives engaged in mildly negative behaviors such as showing anger or whining. In previous research, Gottman and colleagues had found that violent husbands responded to low-level negativity from their wives with very intense negativity (e.g., belligerence, showing contempt). In the current study, the label “accepting influence” was used whenever the wives engaged in low intensity negativity that 16 husbands did not escalate the situation with high intensity negativity. For example, if a husband responded to his wife’s whining by showing that he was angry at her (without expressing contempt or belligerence), this was considered an instance of accepting influence. In addition to the problematic dichotomy of suggesting that all husbands are likely either to accept influence or escalate to violence, it is clear that labeling such a sequence as “accepting influence” is an interpretation. It is accepting of influence in the sense that it is not quashing it strongly, but such a sequence is hardly consistent with the notion that “the newest advice from psychologists is quite simple: Be willing to do what your wife says” (Maugh 1998). This example illustrates the larger point that even when the data allow for direct observations of interpersonal communication, researchers should still be mindful of the fact that the data are interpreted, which is not truly objective. Finally, interpersonal communication scholars should remember that, except in the special case of first encounters, interactions between individuals are shaped by the history of interactions between the individuals involved. It is common for the stream of a conversation to span more than one particular encounter, and multiple periods of interaction across time can be recognized as belonging to the same discussion (Agha 2007). Research by Roloff and his colleagues (e.g., Johnson and Roloff 1998; Reznik and Roloff 2011), for example, illustrates that the impact of serial arguments can really only be understood in the context of the history of the conflict episodes on any particular topic. Whenever researchers observe an interaction segment from an existing relationship, “it is important to remember that outsiders know little about the history of the relationships they observe” (Feeney and Noller 2012: 34). Bringing a dyad into a laboratory may allow researchers to extract a sample of interaction that can be objectively analyzed, but that objective analysis may miss what that interaction actually means. 17 4.3. In-depth interviews There are a variety of types of interviews. Some interviewing is essentially comparable to questionnaire studies. For example, in the PAIR Project, a longitudinal study of couples first contacted as newlyweds, the follow-up phase conducted 13 years after the initial one was conducted entirely through phone interviews because the researchers and many of the participants had moved from the initial study location (see Huston et al. 2001). Many of the measures were derived from interview questions that were taken from paper and pencil questionnaires. In this instance and in other examples of interviews utilizing primarily closedended questions, the data yielded are probably quite comparable to that of standard questionnaires. Yet interviews can also involve various procedures used for different purposes. Interviews composed of open-ended questions typically involve an attempt to gather in-depth information, with the researcher using probing questions to facilitate thoughtful responses. The typical goal of such interviews is to “understand the lived experience of other people and the meaning they make of that experience” (Seidman 2006: 9). The most obvious purpose of such interviews is to investigate phenomena that cannot be observed, but qualitative researchers also strive to treat in-depth interviews as collaborative encounters that can allow important questions and phenomena to emerge as participants and researchers discuss a given topic (Lindlof and Taylor 2011). That is, unlike closed response questionnaires, in-depth interviews have the potential to reveal aspects of communication that the researchers did not even set out to study. Of course, questionnaires can include open-ended questions so they have some potential to reveal unexpected information, but questions typically preclude the interactive and probing aspect of interviews that can elicit subtle insights. 18 Consider, for example, Goldsmith, Lindholm, and Bute’s (2006) study of cardiac patients and their partners. One common communication dilemma for partners that emerged from the interviews involved partners’ desire to encourage healthy lifestyle choices seeming to contradict their desire not to “nag” their partner. Not only is it important to recognize that the understanding of this dilemma emerged from the interviews, but it is worth noting that it could be the defining feature of an encounter while simultaneously being something that could not be understood by even the most well-placed observer. Imagine, for example, that a man recovering from a heart attack is resisting his physicians’ and family’s attempts to increase his activity level. One morning at breakfast, his wife says, “Wow—it’s just a nice day out today. Are you going to walk to work?” This could be the wife’s attempt to suggest that her husband should walk, and depending on how often his wife makes similar suggestions and his sensitivity, the husband may even hear that seemingly simple question as nagging. Regardless of how successfully the wife is able to influence without coming across as a nag, understanding the significance of that episode requires understanding what it means to those individuals. Whatever behaviors could be objectively recorded or coded from that interaction may provide other useful information (e.g., about the emotional expressions of both individuals), but observations would be unlikely to reveal why this encounter is important in that particular relationship. Sometimes the meaning that people attribute to their communication is the most meaningful thing one can know about it. 4.4. Physiological measures In recent years, there has been a marked increase in the use of physiological measurement among interpersonal communication researchers (for a review, see Floyd and Afifi 2011). Cardiovascular reactivity, for example, can be measured by assessing participants’ heart rate and blood pressure at baseline, in the presence of a stressor, and during a recovery period. These 19 physiological indicators can be used to assess responses such as being engaged, challenged, or stressed, which can provide insights into individuals’ emotional states, communication experiences, or individual characteristics (Goyal et al. 2008). 4.4.1 Advantages and disadvantages of physiological measures Physiological data reveal information about the internal processes of research participants that cannot be obtained through other techniques (Smith and Uchnio 2008). In some cases, this information may prove crucial to understanding communication; for example, Afifi (2011) noted that some adolescents discussing their parents’ divorce with one of the parents showed marked stress reactions in their cortisol levels, even when they reported that the conversations were not stressful and there were no obvious behavioral manifestations of strain. Such findings suggest that physiological measures sometimes can provide information about processes about which individuals are not consciously aware. Using an entirely different technique, Buck and Powers (2005) explored the use of fMRI measurement to assess biologically based emotions. Their findings revealed that individuals’ internal emotional experiences and their external expressions of those emotional states often differ. Furthermore, McRae and colleagues (2008) conducted a study of gender differences in emotion regulation. They found that, contrary to most communication research, which finds few differences in how males and females express emotion, the physiological responses between genders differed significantly. Taken together, these studies and others using similar fMRI measures (e.g., Ochsner et al. 2002) suggest that by not considering physiological responses, our understanding of individuals’ experiences and communication patterns is, at best, incomplete. Although physiological measures promise to add much to our understanding of what people experience during interpersonal communication, these techniques have limitations. Most 20 obviously, interpersonal communication, to a large extent, involves the exchange and negotiation of meaning. Physiological measures can show that individuals are aroused or experiencing some emotional state, but they cannot tell us what those responses mean to the individuals or to the interaction. Moreover, some scholars have argued that the claims made about some physiological data have been oversold. For instance, Legrenzi, Umilta, and Anderson (2011: 17) suggest that brain images have been presented as if the readings are much more precise and objective than they actually are and that scientific descriptions of which parts of brains are involved in certain processes oversimplify the extent to which multiple processes occur simultaneously. We are not suggesting that brain imaging is not useful, but it is worth recognizing that even purported experts are only beginning to understand what such measures really can and cannot tell us. Given that some physiological assessment tools are relatively new and technologically sophisticated, we should be cautious about expecting too much from them. It is important to separate the “gee whiz” aspect of such techniques from what can be learned about the substance of interpersonal communication. 5. The utility of using multiple measurement techniques As argued above, every measurement technique (and every specific measure) has strengths and weaknesses (Feeney and Noller 2012). One response to the realization that different measurement techniques have different strengths is to suggest that researchers try to match techniques to constructs based on these strengths. For instance, it is reasonable to suggest that researchers try to use observational methods when the construct involves overt behaviors and use self-report measures when the construct pertains to individuals’ psychological processes, such as memories, attitudes, or emotions (Levine 2011). There are certainly instances in which the focus of study clearly warrants the use of a particular measurement technique over others. 21 People may not even be aware of many of their nonverbal behaviors, for example, and they would not be able to articulate the intent or purpose of them (Burgoon, Guerrero, and Manusov 2011). Thus, if nonverbal behaviors are important constructs in a study, it would typically be best to use observations and coding rather than self-reports. Although there are plainly instances when one type of measurement is better than another, given that there is no single best way to conceptualize any phenomenon, it is often useful to use multiple measures to assess the same general construct. Studying the same phenomena with a variety of methods offers different perspectives and insights (Cappella 1991), and examining the same phenomena with different lenses can have various benefits. When various measures of a construct converge, it provides evidence of validity (Campbell and Fiske 1959), and when various methods are triangulated with each other, consistent results provide more confidence in the findings than is possible from any one method (Webb, et al. 1966). For example, there are various ways to assess the demand/withdraw pattern of communication, which involves one person nagging or criticizing while the other person tries to avoid the discussing the topic of criticism. Researchers can observe indications of demand/withdraw in laboratory conversations, but demand/withdraw is inherently a subjective pattern—the same behavior that counts as nagging in one couple may be viewed as simply discussing an issue in another. This problem can be addressed by asking participants to self-report on their demand and withdraw in a laboratory conversation, but demand/withdraw often plays out over the course of days, not within a single conversation (Christensen and Heavey 1993). Such lengthy behavioral patterns can be assessed with retrospective reports, but those are susceptible to various biases. None of these measurement strategies is perfect, but if a study includes all three and all three reveal similar results, this lends a level of credibility to the findings that none of the measures 22 could have conferred on its own. Indeed, some of our research has used this strategy, and the results from different assessments of demand/withdraw often evince very similar patterns (e.g., Caughlin and Malis 2004). For communication phenomena like demand/withdraw patterns, different measurement choices may point researchers to similar conclusions. Sometimes, however, data from multiple measures may highlight different patterns. Often this would impugn the validity of one or more of the measures. Alternatively, such discrepancies may be informative, perhaps revealing important conceptual distinctions that were not initially apparent. For example, in the aforementioned PAIR Project (e.g., Huston et al. 2001), there were two assessments of the frequency of conflict, and each was taken when the couples were newlyweds and again shortly after their first and second anniversaries. The first measure of conflict was a retrospective report in which participants were asked to report on the amount of conflict they had in their marriage in the past two months, and the second was based on the aggregated telephone diary reports (Caughlin and Huston 1996). Both the husbands’ and wives’ reports portrayed a similar story. Based on the diary assessments, there was a clear decline in the number of overt conflicts over the first three years of the marriages, and this was true regardless of whether the couples were happy or unhappy, or whether they remained married or divorced by their thirteenth anniversary. The retrospective reports of conflict, however, evinced a different pattern, with these reports of conflict frequently increasing over the first three phases of the study, particularly among couples who ended up divorcing. At first glance, these results may seem puzzling, but taken together they suggest that among couples who eventually divorce, the number of times they engage in overt conflicts goes down over time, yet they believe (or feel as if) they are continuing to have intense conflicts often. A likely explanation of this seeming inconsistency is that some couples experience serial arguments (Johnson and Roloff 23 1998) that do not get resolved, and they think about them, even on days they do not have an overt disagreement. This suggests that the there is a potentially important conceptual distinction between the number of ongoing conflicts a couple is experiencing and the frequency at which they overtly engage in communication about these issues. Obviously, neither measure alone could have suggested such a distinction, demonstrating that using multiple assessments of a general construct can provide insights into different aspects of that construct (or suggest that what was thought to be a single construct may more usefully be thought of as two). 6. Special measurement considerations for interpersonal communication scholars In many respects, the principles of sound measurement transcend research areas. Such issues as reliability and validity are not specific to the study of interpersonal communication, and in fact, most writing on these measurement issues has been produced by scholars in other disciplines. This is generally unproblematic, but the fact that much of the received wisdom about measurement is influenced by scholars studying other topics frames the discussion and thinking about measurement in ways that foreground some issues and background others. This implies that there are some particular considerations that scholars of interpersonal communication should be aware of, and we discuss four of these below. 6.1. What constitutes an adequate sampling of ongoing interaction? Interpersonal interaction researchers must choose not only what aspect of interaction to study but also the timeframe to study. Scholars who use observational methods, for example, typically record a sample of interaction that occurs on one occasion, and they usually assess the consistency across coders at that particular time (see e.g., Baesler and Burgoon 1987; Caughlin 2003). This aspect of reliability is important, and if the conceptual interest is in understanding that segment of communication, reliability during that interaction is the main concern. However, 24 there is another aspect of reliability that can be equally important in interpersonal communication but is usually ignored. If one is interested in the communication that occurs in ongoing interpersonal relationships, it is important to ask whether the sample of interaction events examined is adequate (Huston and Robins 1982). In close relationships, it is possible to have a measure that evinces the qualities of good reliability at a point in time, but is nevertheless not a reliable indicator of what happens generally, even if the measure itself is otherwise unbiased and valid. Conceptually, single observations are problematic whenever the interest is in how a dyad interacts in general (e.g., the frequency of a particular communication pattern), not just in one encounter. We know that there is variation in how people interact; for example, there is considerable variation in how much negativity spouses express to each other based on their experiences at work (Doumas, Margolin, and John 2003). Scholars must heed these kinds of patterns as they assess interpersonal interaction. In designing a diary study, for instance, researchers should consider how many entries are needed to compose a sufficiently reliable index of the communication variables of interest. As is the case with increasing the number of items on a scale, assessing communication on more occasions tends to increase the reliability of the assessment. Of course, how many samples of communication are needed to compose a reliable index of interpersonal interaction depends on factors such the fluctuation rate of the behaviors. If the behaviors vary greatly, more assessments are necessary for an adequate sampling. If there is little fluctuation, fewer samples of interaction may be sufficient. If couples’ interaction in response to a particular situation is highly routinized, a single assessment may be sufficient to make assessments about what happens in general. 6.2. Is the very nature of interpersonal communication changing? 25 In recent years, there has been a tremendous increase in the use of communication technologies to engage in interpersonal communication. To date, research on the use of those technologies has tended to study face-to-face communication separately from technologically mediated communication, either by focusing on a particular medium (e.g., texting) or by comparing face-to-face to technologically mediated communication (see Chapter 23, Walther and Lee). Yet, there have been recent calls for researchers to begin to study how people use both simultaneously (Baym 2009), and recent evidence suggests that in many close relationships the interconnections between face-to-face encounters and technologically mediated interaction is extensive and complex (Caughlin and Sharabi 2012). These interconnections between online and offline interpersonal communication present some potentially major challenges for current ways of assessing interpersonal communication. If it becomes the norm for relational partners to use their smartphones while interacting face-toface, would a traditional sample of laboratory interaction (which usually would preclude such technologies) be representative of usual interpersonal communication? Moreover, given that mediated communication allows people to extend streams of conversation even when they are apart, this could exacerbate the problem of observing a segment of interaction that is disconnected from the larger conversation about a particular topic. For example, because people in close relationships may have the expectation that they can always get some message through to a partner (even if it is only a brief text), does that change what it means for one person to try avoiding a conflict topic? Can “nagging” now continue remotely? It is not clear how profound these changes ultimately will be, but they do raise questions that interpersonal scholars should be mindful of when they make measurement decisions. It may be that classic observational methods will always tell us something useful, such as how skillful a 26 dyad can be when called on to demonstrate exemplary behavior in a laboratory (Reis and Gable 2000), but we should also recognize that as communication technologies become more embedded into the fabric of interpersonal interaction, the classic laboratory techniques may be less representative of how people actually engage in interpersonal communication. 6.3. Pitfalls of using measures originally developed for other purposes Before using even the most thoroughly-tested existing research measure, it is important to consider whether it is valid for the purposes of a particular investigation. This is always a potential issue, but it appears to be particularly salient for scholars of interpersonal communication because researchers from a number of allied fields have developed pertinent measures, but the purposes of the research often differ enough that the measure is not valid beyond its original use. One example that has been discussed extensively involves two measures that are commonly used to assess martial satisfaction: the Marital Adjustment Test (MAT; Locke and Wallace 1959) and the Dyadic Adjustment Scale (DAS; Spanier 1976). The original purpose of these measures was to assess how well spouses accommodate each other and to provide a tool that could be used to predict marital well-being (see Locke and Wallace 1959). That is, they were originally intended to serve as global indicators of martial functioning; consequently, they include questions on a wide range of topics, including both general questions about how happy spouses are and reports of specific communication behaviors, such as self-disclosure and conflict engagement. This broad scope is probably a valid overall assessment of marital well-being, but neither is a valid assessment of marital satisfaction. Marital satisfaction is usually conceptualized as a subjective evaluation of a marriage; that is, it is an attitude toward marriage. Because both the MAT and DAS include a mix of items reporting on such attitudes but also on communication behaviors within the marriage, they are plainly not valid measures of marital satisfaction 27 (Huston, McHale, and Crouter 1986; Norton 1983). Given that these measures confound satisfaction with questions about communication, communication scholars should be particularly leery of using the MAT and DAS. Using these measures makes it impossible to assess associations between relational communication and satisfaction because any covariances could be due to the confound in the measures. The MAT and DAS are particularly clear examples of measures that are probably valid for one purpose being misused for other purposes, but they illustrate the point that measures always need to be evaluated with respect to the specific constructs under investigation. Another prominent example of a well-established measure that is misused is the series of FACES (Family Adaptability and Cohesion Evaluation Scales) measures developed by Olson and his colleagues (Olson 2000; Olson et al. 1982). The FACES instrument measures the broad constructs of cohesion and adaptability in families, and there is abundant evidence that it provides useful diagnostic information about the functioning of families. Yet, these measures may not be ideal for use in studies of interpersonal communication in families. The original measure of cohesion, for instance, included assessments of widely varied constructs, including emotional bonding, boundaries, coalitions, and time spent together. This mix of affective, behavioral, and structural concepts is unlikely to be unidimensional; indeed, many of the items for the cohesion measure did not load strongly on a single dimension in Olson et al.’s (1982) original report, with factor loadings as low as .13. Given that at least some of the items involve reports of communication behaviors, communication researchers should be particularly cautious about using this cohesion measure for the same reasons they usually want to avoid the MAT and DAS. 6.4. Confusing statistical evidence with proof of validity 28 Our focus has been on the logic of measurement rather than the details of measurement development. There are, of course, sophisticated statistical tools for gathering evidence that relates to validity. Confirmatory factor analysis provides a technique for deciding whether the empirical findings from a measure are congruent with the conceptualized associations among items (Brown 2006). For instance, items intended to assess the same construct should covary highly with each other but should not vary strongly with items intended to assess conceptually distinct constructs. As useful as such techniques can be, it is important not to apply them blindly or indiscriminately. For example, if a researcher is attempting to index a set of behaviors, such as behaviors that contribute to health risks, factor analyses are inappropriate because they presume that there is some underlying construct that causes the various items to be intercorrelated (Bollen and Lennox 1996). An overall assessment of risky behaviors is composed of a number of specific behaviors, which means that the behaviors are not all caused by an underlying risky lifestyle construct. In such instances, the assumptions behind confirmatory analysis do not apply. Even when traditional statistical tools are applicable, empirical findings are only meaningful when used in conjunction with informed and thoughtful conceptualization. Decisions about measurement should always be rooted not only in statistical findings but also in the larger scholarly literature and sound theorizing. For instance, just because items covary highly in a given study with one sample at a particular time does not imply that those items should be considered part of the same construct or that they will always covary highly. Braiker and Kelley (1979), for instance, found that feelings of love and maintenance behaviors were highly and positively correlated early in heterosexual relationships but were empirically distinct in more committed relationships. If Braiker and Kelley had based their measurement on the findings 29 early in the relationships, they may have lumped love and maintenance together into a common index, but the empirical findings from another point in relationships are consistent with a conceptual distinction between love and maintenance. Unfortunately, researchers sometimes make conceptual decisions based purely on the statistical information from a single study. For example, one enduring problem in relationship research is the fact that some scholars use high correlations between spouses’ attitudes as a rationale for combining those attitudes into a single score. Thompson and Walker (1982) long ago pointed out that this is problematic. Consider the case of marital satisfaction. As argued above, marital satisfaction is a subjective evaluation, yet researchers sometimes average the scores of husbands’ and wives’ attitudes toward their marriage. Regardless of how correlated those two values are, we know that husbands and wives sometimes do differ in their attitudes toward marriage, and conceptually, an attitude is a property of an individual; thus, it does not make sense to create a combined measure of satisfaction (Thompson and Walker 1982). Reliance solely on statistical support can lead researchers to make invalid indices. Specific to interpersonal communication, scholars should be particularly mindful of this issue because some behaviors that are known to be conceptually distinct are highly correlated under some conditions. In fact, some studies that base measurement decisions purely on statistical covariance have collapsed an exceedingly wide range of communication behaviors into a single measure. For instance, in a highly cited study by Karney and Bradbury (1997), the covariances among observed behaviors were used to create a single measure of communication, with positive behaviors and negative behaviors simply viewed as two ends of a single dimension. Regardless of how high the correlation between negativity and positivity is in a particular observation, there is not a conceptually sound justification for combining these. In addition to 30 reducing the domain of interpersonal research into a single variable, this is problematic because numerous other studies have shown that negative and positive behaviors are often empirically distinct, both in terms of having low covariance and in terms of predicting different outcomes (for reviews, see Caughlin and Huston 2006; Gable and Reis 2001). Not only do positive and negative behaviors in relationships have distinct outcomes, but they also appear to moderate each other in some instances (e.g., Huston and Chorost 1994; Smith, Vivian, and O’Leary 1990), a finding that would be obscured if these behaviors were lumped together. Given that positive and negative behaviors are clearly distinct, it is unclear exactly what any findings relating to a combined measure of positive and negative behaviors even mean; are they due to high levels of one set of behaviors, low levels of the other, or some interaction between the two that was not systematically examined? As implied by O’Keefe’s (1987) argument about analyzing messages, there probably is not a right answer to how many constructs should be gleaned from interpersonal communication in dyads, but it is clear that reducing interpersonal interaction to one construct is conceptually indefensible. In short, interpersonal communication researchers should be cautious about forming indices of communication based solely on the statistical information from a single study. 7. Conclusion The goal of this chapter has been to provide a conceptual overview of important issues involved in measuring social interaction. We have argued that there is no single correct or best way to assess interpersonal communication constructs. No measure of interpersonal communication is perfect. Indeed, the only way to make the study of interpersonal communication entirely objective is to ignore the meaning of it. Even when researchers have elaborate and precise coding of actual interaction data, they often end up describing what they 31 observe in subjective terms, such as what the participants are trying to accomplish. Gottman’s interpretation of wives’ and husbands’ behaviors as seeking and accepting influence is just one of many examples (Gottman et al. 1998). Recognizing the inherent weaknesses of all measures does not mean that every assessment is equally useful. There are still better and worse practices, and more and less useful ways to assess social interaction. For example, rather than seeking a single valid measure, we suggest that researchers attempt to use multiple assessments when possible and also remain sensitive to the particular challenges inherent in measuring something as complex and dynamic as interpersonal interaction. 32 References Afifi, Tamara D. 2011, April. The impact of divorce and parents' communication skills on adolescents' and young adults' physiological stress response systems. Presentation to the University of Illinois Department of Communication Colloquium Series, Urbana, IL. Agha, Asif (2007). Language and Social Relations. New York, NY: Cambridge University Press. Baesler, Erland J. and Judee Burgoon (1987). Measurement and reliability of nonverbal behavior. Journal of Nonverbal Behavior 11: 205-233. Baxter, Leslie and Earl Babbie (2004). The Basics of Communication Research. Belmont, CA: Thomson Wadsworth. Baym, Nancy K. (2009). A call for grounding in the face of blurred boundaries. Journal of Computer-Mediated Communication. 14: 720-723. Beavin Bavelas, Janet, Christine Kenwood and Bruce Phillips. (2002). Discourse analysis. In: Mark.L. Knapp and John A. Daly (eds.), Handbook of Interpersonal Communication, 102-129. Thousand Oaks, CA: Sage. Berger, Charles, R. and Nicholas A. Palomares (2011). Knowledge structures and social interaction. In: Mark L. Knapp and John A. Daly (eds.), The SAGE Handbook of Interpersonal Communication, 169-200. Los Angeles, CA: Sage. Bollen, Kenneth and Richard Lennox (1991). Conventional Wisdom on Measurement: A Structural Equation Perspective. Psychological Bulletin 110: 305–314. Braiker, Harriet B. and Harold H. Kelley (1979). Conflict in the development of close relationships. In: Robert L. Burgess and Ted L. Huston (eds.), Social Exchange in Developing Relationships, 135-168. New York, NY: Academic Press. Brown, Timothy A. (2006). Confirmatory Factor Analysis for Applied Research. New York, NY: Guilford Press. Buck, Ross and Stacie R. Powers (2005). The expression, communication, and regulation of biological emotions: Sex and cultural differences and emotions. Psychologia 48: 335 353. Burgoon, Judee K., Laura K. Guerrero and Valerie Manusov (2011). Nonverbal signals. In: Mark L. Knapp and John A. Daly (eds.), The SAGE Handbook of Interpersonal Communication, 239-280. Los Angeles, CA: Sage. Campbell, Donald, T. and Donald W. Fiske (1959). Convergent and discriminant validation by the multitrait-multimethod matrix. Psychological Bulletin 56: 81-105. 33 Cappella, Joseph N. (1991). Mutual adaptation and relativity of measurement. In: Barbara M. Montgomery and Steve Duck (eds.), Studying Interpersonal Interaction, 103-117. New York, NY: Guilford Press. Caughlin, John P. (2003). Family communication standards: What counts as excellent family communication and how are such standards associated with family satisfaction? Human Communication Research 27: 5-40. Caughlin, John P., Walid A. Afifi, Katy Carpenter-Theune and Laura E. Miller (2005). Reasons for, and consequences of, revealing personal secrets in close relationships: A longitudinal study. Personal Relationships 12: 43-59. Caughlin, John P. and Ted L. Huston 1996, November. The development of conflict: Perception and frequency of conflict among couples with different marital outcomes. Paper presented at the annual convention of the Speech Communication Association, San Diego, CA. Caughlin, John P. and Ted L. Huston (2006). The affective structure of marriage. In: Anita L. Vangelisti and Daniel Perlman (eds.), The Cambridge Handbook of Personal Relationships, 131-155. New York, NY: Cambridge University Press. Caughlin, John P. and Rachel S. Malis (2004). Demand/withdraw communication between parents and adolescents: Connections with self-esteem and substance use. Journal of Social and Personal Relationships 21: 125-148. Caughlin, John P., Sandra Petronio and Ashley V. Middleton (2012). When families manage private information. In: Anita L. Vangelisti (ed.), Routledge Handbook of Family Communication (2nd ed.), 321-337. New York, NY: Routledge. Caughlin, John P. and Liesel Sharabi 2012, July. In close relationships, it’s both online and face to-face communication—Not online versus face-to-face communication. Paper presented at the biennial conference of the International Association for Relationship Research, Chicago, IL. Christensen, Andrew and Chris Heavey (1993). Gender differences in marital conflict: The demand/withdraw interaction pattern. In: Stuart Oskamp and Mark Costanzo (eds.), Gender Issues in Contemporary Society, 113-141. Newbury Park, CA: Sage. Doumas, Diana M., Gayla G. Margolin and Richard S. John (2003). The relationship between daily marital interaction, work, and health-promoting behaviors in dual-earner couples: An extension of the work-family spillover model. Journal of Family Issues 24: 3-20. Duck, Steve and Barbara M. Montgomery (1991). The interdependence among interaction substance, theory, and methods. In: Barbara M. Montgomery and Steve Duck (eds.), Studying Interpersonal Interaction, 3-15. New York, NY: Guilford Press. 34 Feeney, Judith A. and Patricia Noller (2012). Perspectives on studying family communication: Multiple methods and multiple sources. In: Anita L. Vangelisit (ed.), Routledge Handbook of Family Communication, 29-45. London/New York: Routledge. Finkenauer, Catrin, Peter Kerkhof, Fracesca Righetti and Susan Branje (2009). Living together apart: Perceived concealment as a signal of exclusion in marital relationships. Personality and Social Psychology Bulletin 35: 1410-1422. Floyd, Kory and Tamara D. Afifi (2011). Biological and physiological perspectives on interpersonal communication. In: Mark L. Knapp and John A. Daly (eds.), The SAGE Handbook of Interpersonal Communication, 87-127. Los Angeles, CA: Sage. Frey, Lawrence R., Carl H. Botan and Gary L. Kreps (2000). Investigating Communication: An Introduction to Research Methods (2nd ed.). Boston, MA: Allyn and Bacon. Gable, Shelly L. and Harry T. Reis (2001). Appetitive and aversive social interaction. In John H. Harvey and Amy E. Wenzel (eds.), Close Romantic Relationships: Maintenance and Enhancement, 169-194. Mahwah, NJ: Erlbaum. Goldsmith, Daena J., Kristin A. Lindholm and Jennifer J. Bute (2006). Dilemmas of talk about lifestyle changes among couples coping with a cardiac event. Social Science and Medicine 63: 2079-2090. Goyal, Tanya M., Daichi Shimbo, Elizabeth Mostofsky and William Gerin (2008). Cardiovascular stress reactivity. In: Linda J. Luecken and Linda C. Gallo (eds.), Handbook of Physiological Research Methods in Health Psychology, 133-158. Thousand Oaks, CA: Sage. Gottman, John M. (1994). Why Marriages Succeed or Fail. New York, NY: Simon & Schuster. Gottman, John M., James Coan, Sybil Carrere and Catherine Swanson (1998). Predicting happiness and stability from newlywed interactions. Journal of Marriage and the Family 60: 5-22. Graber, Elana, Jean-Philippe Laurenceau, Erin Miga, Joanna Chango and James Coan (2011). Conflict and love: Predicting newlywed marital outcomes from two interaction contexts. Journal of Family Psychology 25: 541-550. Huston, Ted L., John P. Caughlin, Renate M. Houts, Shanna E. Smith and Laura J. George (2001). The connubial crucible: Newlywed years as predictors of marital delight, distress, and divorce. Journal of Personality and Social Psychology 80: 237-252. Huston, Ted L. and Amy F. Chorost (1994). Behavioral buffers on the effect of negativity on marital satisfaction: A longitudinal study. Personal Relationships 1: 223-239. 35 Huston, Ted L., Susan M. McHale and Ann C. Crouter (1986). When the honeymoon’s over: Changes in the marriage relationship over the first year. In: Robin Gilmore and Steve Duck (eds.), The Emerging Field of Personal Relationships, 109-132. New York, NY: Erlbaum. Huston, Ted L. and Elliot Robins (1982). Conceptual and methodological issues in studying close relationships. Journal of Marriage and the Family 44: 901-925. Johnson, Kristen L. and Michael E. Roloff (1998). Serial arguing and relational quality: Determinants and consequences of perceived resolvability. Communication Research 25: 327-343. Karney, Benjamin. R. and Thomas N. Bradbury (1997). Neuroticism marital interaction, and the trajectory of marital satisfaction. Journal of Personality and Social Psychology 72: 1075 1082. Knapp, Mark L. and John A. Daly (2011). Background and current trends in the study of interpersonal communication. In: Mark L. Knapp and John A. Daly (eds.), The SAGE Handbook of Interpersonal Communication, 3-22. Los Angeles, CA: Sage. Larson, Dale G. and Robert L. Chastain (1990). Self-concealment: Conceptualization, measurement, and health implications. Journal of Social and Clinical Psychology 9: 439 455. Legrenzi, Paolo, Carlo Umilta and Frances Anderson (Translator) (2011). Neuromania: On the Limits of Brain Science. Oxford/New York: Oxford University Press. Le Poire, Beth A. and Stephen M. Yoshimura (1999). The effects of expectancies and actual communication on nonverbal adaptation and communication outcomes: A test of interaction adaptation theory. Communication Monographs 66: 1-30. Levine, Timothy R. (2011). Quantitative social science methods of inquiry. In: Mark L. Knapp and John A. Daly (eds.), The SAGE Handbook of Interpersonal Communication (4th ed.), 25-57. Los Angeles, CA: Sage. Lindlof, Thomas R. and Bryan C. Taylor (2011). Qualitative Communication Research Methods (3rd ed.). Thousand Oaks, CA: Sage. Locke, Harvey J. and Karl M. Wallace, K. (1959). Short marital-adjustment and prediction tests: Their reliability and validity. Marriage and Family Living 21: 251-255. Maugh, Thomas H. 1998, February 21. Study’s advice to husbands: Accept wife’s influence. Los Angeles Times. Retrieved from http://articles.latimes.com/1998/feb/21/news/mn-21412 36 McRae, Kateri, Kevin N. Ochsner, Iris B. Mauss, John J. Gabrieli and James J. Gross (2008). Gender differences in emotion regulation: An fMRI study of cognitive reappraisal. Group Process & Intergroup Relations 11: 143-162. Metts, Sandra, Susan Sprecher and William R. Cupach (1991). Retrospective self-reports. In: Barbara M. Montgomery and Steve Duck (eds.), Studying Interpersonal Interaction, 162 178. New York, NY: Guilford Press. Norton, Robert (1983). Measuring marital quality: A critical look at the dependent variable. Journal of Marriage and the Family 45: 141-151. Ochsner, Kevin N., Silvia A. Bunge, James J. Gross and John D. E. Gabrieli (2002). Rethinking feelings: An fMRI study of the cognitive regulation of emotion. Journal of Cognitive Neuroscience 14: 1215-1229. Ogden, Charles. K. and Ivor, A.Richards (1923). The Meaning of Meaning. Oxford, Hacourt. O’Keefe, Daniel J. (1987). Message description. Paper presented at the annual convention of the Speech Communication Association. Boston, MA. Olson, David. H. (2000). Circumplex model of marital and family systems. Journal of Family Therapy 22: 144-167. Olson, David H., Hamilton I. McCubbin, Howard Barnes, Andrea Larsen, Marla Muxen and Marc Wilson (eds.) (1982). Family Inventories: Inventories Used in a National Survey of Families across the Family Life Cycle. St. Paul, MN: Family Social Science, University of Minnesota. Reinard, John C. (2008). Introduction to Communication Research (4th ed.). Boton, MA: McGraw-Hill. Reis, Harry T. and Shelly L. Gable (2000). Event-sampling and other methods for studying everyday experience. In: Harry T. Reis and Charles M. Judd (eds.), Handbook of Research Methods in Social and Personality Psychology, 190-222. Cambridge/New York: Cambridge University Press. Reis, Harry T. and Ladd Wheeler (1991). Studying social interaction with the Rochester Interaction Record. Advances in Experimental Social Psychology 24: 269-318. Reznik, Rachel M. and Michael E. Roloff (2011). Getting off to a bad start: The relationship between communication during an initial episode of a serial argument and argument frequency. Communication Studies 62: 291-306. Sanford, Keith (2012). The communication of emotion during conflict in married couples. Journal of Family Psychology 26: 297-307. 37 Seidman, Irving (2006). Interviewing as Qualitative Research: A Guide for Researchers in Education and the Social Sciences (3rd ed.). New York, NY: Teachers College Press. Singleton Jr., Royce A. and Bruce C. Straits (2005). Approaches to Social Research (4th ed.). New York, NY: Oxford University Press. Smith, David A., Dina Vivian and Daniel K. O’Leary (1990). Longitudinal prediction of marital discord from premarital expressions of affect. Journal of Consulting and Clinical Psychology 58: 790-798. Smith, Timothy W., Matthew R. Cribbet, Jill B. Nealey-Moore, Bert N. Uchino, Paula G. Williams, Justin MacKenzie and Julian F. Thayer (2011). Matters of the variable heart: Respiratory sinus arrhythmia response to martial interaction and associations with marital quality. Journal of Personality and Social Psychology 100: 103-119. Smith, Timothy W. and Bert N. Uchino (2008). Measuring physiological processes in biopsychosocial research: Basic principles amid growing complexity. In: Linda J. Luecken & Linda G. Gallo (eds.), Handbook of Physiological Research Methods in Health Psychology, 11-33. Thousand Oaks, CA: Sage. Spanier, Graham B. (1976). Measuring dyadic adjustment: New scales for assessing the quality of marriage and similar dyads. Journal of Marriage and the Family 38: 15-28. Sullivan, Kieran T., Lauri A. Pasch, Matthew D. Johnson and Thomas N. Bradbury (2010). Social support, problem solving, and the longitudinal course of newlywed marriage. Journal of Personality and Social Psychology 98: 631-644. Surra, Catherine A. and Carl A. Ridley (1991). Multiple perspectives on interaction: Participants, peers, and observers. In: Barbara M. Montgomery and Steve Duck (eds.), Studying Interpersonal Interaction, 35-55. New York, NY: Guilford Press. Thompson, Linda and Alexis J. Walker (1982). The dyad as the unit of analysis: Conceptual and methodological issues. Journal of Marriage and the Family 44: 889-900. Vangelisti, Anita L., Kathryn C. Maguire, Alicia L. Alexander and Gretchen Clark (2007). Hurtful family environments: Links with individual, relationship, and perceptual variables. Communication Monographs 74: 357-385. Webb, Eugene, Donald T. Campbell, Richard D. Schwartz and Lee Sechrest (1966). Unobtrustive Measures: Nonreactive Research in the Social Sciences. Chicago, IL: Rand McNally. Wilson, Steve R. (2002). Seeking and Resisting Compliance: Why People Say What They Do When Trying to Influence Others. Thousand Oaks, CA: Sage. 38 Young, Stacy L., Tara L. Kubicka, Caralyn E. Tucker, Desi Chávez-Appel and Jamie S. Rex (2005). Communicative responses to hurtful messages in families. Journal of Family Communication 5: 123-140.