The Influential Role of Relational Messages in Group Interaction
Joann KeytonNorth Carolina State University
Stephenson J. BeckNorth Dakota State University
Relational communication is often neglected in group interaction research; this studyargues for a need to refocus our efforts on the relational dimension of group commu-nication. Specifically, this article uses the Interaction Process Analysis (IPA) codingscheme to investigate implications of the relational communication model. Usingsequences of talk in breast cancer support group interaction, the authors examine howtask messages preceding and following relational messages in conjunction with therelational messages accomplish different relational (and task) goals. Results confirmearlier claims that relational influence stems from both task and relational messages. Anexamination of the overlapping nature of task and relational dimensions of communi-cation, as well as the multidimensional nature of relational messages, is likewiseexplored.
Keywords: relational, socioemotional, interaction analysis, group, communication
Predominantly, group research has focusedon task groups and task behavior or interaction(Hollingshead et al., 2005), in part because ofthe goal-orientation society has ascribed togroups and teams. Groups are supposed to dosomething (i.e., make decisions, solve prob-lems, create ideas). Task group interaction hasbeen addressed by many different research per-spectives with a primary focus on decision mak-ing, leadership, and the influence of participa-tion on information sharing. In general, theselines of work have focused on how task mes-sages influence subsequent member behaviorsand group outcomes.
Groups with goals that are socially oriented(i.e., creating and maintaining relationships,providing social support) have received consid-erably less attention. This is ironic in that allgroups are contextually and socially situated bythe interpersonal and intragroup relationshipsdeveloped by members through their group in-
teractions. Despite a distinction that can bemade between groups that pursue task-orientedgoals (e.g., decision-making groups, self-managing teams) and groups that pursue socialor relational goals (e.g., family groups, friend-ship groups, social support groups), all groupmembers will engage in relationally orientedinteractions or will interpret other group mem-bers’ interactions as facilitating or hinderingrelationship building (Keyton, 1999). When re-lational or social communication has been ex-plored, it has often examined task-orientedgroups by subordinating relational interaction(e.g., talk that produces cohesion) to task inter-action (e.g., talk that moves a group closer to itsgoal). Thus, the focus on task groups and task-oriented interactions leaves the relational aspectof group interaction undertheorized.
Relational Perspectives onGroup Interaction
Initial attempts at understanding the relational1
influence of groups began in the early 1950s(Bales, 1950a, 1953; Bion, 1948; Homans,
1 Even though there is debate about the socioemotionalvalidity of agrees and disagrees messages (Ridgeway & John-son, 1990), analyzing them for their relational functioning stillsupports our argument that relational elements must be under-stood in relationally oriented messages before we can under-stand the relational elements of task-oriented messages.
Joann Keyton, Department of Communication, NorthCarolina State University; Stephenson J. Beck, Departmentof Communication, North Dakota State University.
This article is based on a manuscript presented at theconference of the Interdisciplinary Network for Group Re-search (INGRoup), East Lansing, MI, July 2007.
Correspondence concerning this article should be addressedto Joann Keyton, Department of Communication, North Caro-lina State University, 201 Winston Hall, Campus Box 8104,Raleigh, NC 27695-8104. E-mail: [email protected]
1950; Lewin, 1951). Bion (1948) argued thatgroups always were in one of four states ofrelational form: fight, flight, pairing, depen-dency. Homans (1950) focused on the rela-tionship between communication, task, andsentiment or relational issues in group inter-action. Lewin (1951) connected relationalfactors to group member motivation and sub-sequent accomplishment of goals.
Among this initial highlighting of the rela-tional dynamic of groups, two foundational per-spectives on relational interaction emerged andare largely respected and accepted. The firstperspective, and most broadly construed, is thatof Watzlawick, Beavin, and Jackson (1967)who argue that relational information occurs inall messages in all contexts. In the form ofaffirming information given by the speaker,showing support to the speaker, and signaling tothe speaker to continue, relational messages canstand alone or simultaneously with task mes-sages that give and receive information. Perhapsbecause of the simplistic accompanying meth-odology, its grounding in interpersonal commu-nication, or its obviousness, the work ofWatzlawick and colleagues has failed to gener-ate research applications in the group context.Most importantly, it has failed to explain howtask and relational dimensions coexist in con-versation and why one dimension may domi-nate the other.
The second perspective—and one adopted byscholars in a variety of disciplines—is that ofBales (1950a) who posits that messages ingroup interaction are either socioemotional ortask—one message cannot assume the proper-ties of both. This theoretical position is opera-tionalized by the twelve functions of InteractionProcess Analysis (IPA), requiring a thought unitto be coded as accomplishing a task function orrelational function. Over time, research has gen-erally demonstrated that groups create moretask messages than socioemotional or relationalones (Bales, 1950b, 1999). This is not surpris-ing as giving and asking for suggestions, opin-ions, and information would be central to avariety of group goals, task or socially oriented.It is interesting to note that Bales (1950a) didprovide evidence of a position that task andrelational messages coexist. First, his codingprocedures for IPA instruct coders to look forthe surface meaning of a message from the viewof the generalized other, and not for more in
depth meanings. Second, he recommends thatcoders favor relational over task codes when amessage appears to be equally relational andtask. Other scholars (Fisher & Ellis, 1990) haveargued that these coding artifacts are represen-tative of a belief that Bales held a less radicalperception of the task-relational dichotomy; infact, he believed they coexisted, but that onealways predominated over the other.
Despite the prevalence of these two relationalperspectives, the role of relational interactionamong group members has often been studiedas members’ perceptions of cohesion that eitherinfluence task completion (see, e.g., Chang &Bordia, 2001) or as satisfaction that results fromtask completion (see, e.g., Keyton, 1991). Inpart, this may be due to the more informal (andoften perceived as secondary) role of relationalmessages. Early on (Benne & Sheats, 1948) andmore recently (Mudrack & Farrell, 1995) re-searchers have identified group maintenanceroles that help members define relationships. Inthese roles, members deliver messages thatpraise, demonstrate agreement, and reconciledisagreement—all acts that contribute to rela-tionships and relational climate in the group.Also early on, Schutz (1966) argued that indi-viduals join groups to satisfy relational needs,such as inclusion, control, and affection—needsthat cannot be completely satisfied by task mes-sages. Other research also acknowledges thatgroup members’ liking (i.e., based on their re-lationships with one another) influences mem-ber identity with their group (Henry, Arrow, &Carini, 1999) and that strong relationships (i.e.,liking one another) positively influences groupsynergy (Salazar, 1995). These effects are at-tributed to relational messages among groupmembers.
Beyond these studies, two have directlyidentified relational communication as facili-tating and hindering task and decision-making groups. In their exploration of deci-sion-making groups, Scheerhorn, Geist, andTeboul (1994) found that relating with others isa primary activity in group meetings. Alter-nately, DiSalvo, Nikkel, and Monroe (1989)found that group members reported that inter-personal problems were the primary reason fortheir work groups’ ineffectiveness.
Despite two seminal positions and ongoingresearch concerning relational constructs, rela-tional dynamics in groups are not well
15RELATIONAL MESSAGES
understood; nor is there consensus as to howrelationships are created among group mem-bers. One reason for this difficulty is the com-plexity of the relational dynamics of communi-cation. Embedded in all conversation contexts,including groups and teams, is a relational foun-dation upon which messages are created anddelivered. A group member consciously or un-consciously assesses the nature of his or herrelationships with other members, and uses thisassessment as a premise for creating, delivering,listening to, and interpreting messages. More-over, when messages from a sender are consis-tent with a receiver’s relational assessment, thereceiver is more likely to attribute validity andcredibility to the sender’s message. If messagesconflict with the receiver’s assessment, the re-ceiver may revise his or her relational assumptionsin his or her future messages. In both ways,interaction among group members provides in-formation from which relational inferences aredrawn (e.g., see Knobloch & Solomon, 2005).
Keyton’s Organizing Framework ofRelational Communication in Groups
In her review of relational communication,Keyton (1999) created a framework of howrelational messages function in groups that con-nected relational messages, processes, and out-comes (see Figure 1). There are three points ofemphasis to this model. First, the model drawsattention to the relational elements of all mes-sages in groups, regardless of their task or re-lational predominance. Several aspects of themessage, including the content, verbals, andnonverbals, all contribute to the relational value
of a message and the social fabric of the group.Second, the model draws attention to the rela-tional processes that occur as a result of groupcommunication. Because of the complexity ofgroup interaction, it is unclear whether inter-personal findings on relational processes arereplicated in groups. Keyton (1999) considersfurther examination of how relational commu-nication creates processual outcomes as animportant avenue for future relational commu-nication research. Third, the model draws atten-tion to traditional, positively and negatively va-lenced relational outcomes. In doing so, themodel hopes to create a more complete under-standing of how relational messages influencerelational outcomes.
The model works in a cyclical fashion in thatrelational outcomes affect subsequent task andrelational messages, which, in turn, affect sub-sequent relational processes, and so forth. Thus,the model refutes the assumption that groupmembers use only task messages to accomplishtask purposes, and use only relational messagesto accomplish relational purposes. Both taskand relational messages are used interdepen-dently to accomplish both task and relationalgoals.
While the model illustrates process, it is notintended as a representation of longitudinalgroup development in which relational out-comes occur only at the end of a process ormeeting. Relational outcomes are accom-plished, changed, and evolved throughout theinteractions in a group’s history. Thus, thismodel can reflect the effects of relationalcommunication at the beginning of their first
Figure 1. Relational messages, processes, and outcomes in groups.
16 KEYTON AND BECK
meeting, during periods of conflict and transi-tion, as well as at the termination of the group.
An example is used to illustrate the point.Suppose group member B asks group leader Ato take a look at a memo she recently wroteduring a meeting. Group leader A responds, “Idon’t have time for that right now.” Groupleader A is intending this message to be task-oriented, stating that the present time would notbe best for consideration of the memo. Eventhough there is clear task value for group mem-ber B, relational value may also be inferred.Group member B could believe that groupleader A does not value her contributions, orthat group leader A does not want to work withher. This message was delivered as part of thegroup’s discussion at a meeting (group process),and may affect several relational outcomes(e.g., satisfaction, communication climate, co-hesiveness). These relational outcomes will af-fect future task and relational content, as well asa group’s relational process, creating a cyclicaldevelopment pattern.
The purpose of this study is to validate andelaborate upon Keyton’s (1999) organizingmodel of relational communication and demon-strate with data the influential role of relationalmessages in group interaction. Human interac-tion, which by definition is a social process,cannot exist without relational informationabout and from others. Furthermore, messageshave a relational influence on all interactionwhether or not messages are intended to haverelational consequences. Thus, we argue thatrelational messages need to be accounted formore broadly theoretically, answering the callof researchers to wed group theories and re-search practices to a new relational agenda ingroups (Barker et al., 2000).
Bales’ (1950a) IPA coding scheme, whichlabels messages in terms of their surface levelcommunicative function, was used in this studybecause the coding methodology allows re-searchers to isolate relationally oriented acts.Before attempting to examine the relational di-mensions of all messages, a more complete un-derstanding of how relationally oriented mes-sages function is needed. Discovering howthese predominantly relational acts function cancreate a blueprint from which to analyze allmessages for relational influence. Further, italso allows for a comparative foundation to seeif relational elements function differently in
some messages as opposed to others. Thus, thisarticle examines how relational messages areused in and influence group interaction with theultimate objective of creating a premise to in-vestigate the relational dimension of all groupinteraction. Because task interaction in taskgroups has dominated group research, wesought to examine relational communication ina context in which relational interaction wouldbe obvious and necessary to the goals of thegroup. Thus, we selected a support group forthis microanalysis of relational interaction. Inaddition, we elected to look at how relationalmessages functioned in sequence with task-oriented messages, in order to focus on howrelational messages worked in interaction asopposed to how they might be interpreted inisolation. The following research questionswere posed, in hopes of creating a foundation toinvestigate Keyton’s (1999) larger claims thatthere are relational elements in all messages andthat relational communication creates relationalprocesses themselves:
RQ1: How do relationally oriented messagesfunction in support group interaction?
RQ2: How are relationally oriented messagesinterdependent with task messages?
Method
Data for this project are drawn from fiveweekly meetings of a breast cancer supportgroup. Social support was the primary goal ofthe group; social support is defined as “respon-siveness to another’s needs and, more specifi-cally, as acts that communicate caring . . . orthat facilitate adaptive coping with problemsthrough the provision of information, assis-tance, or tangible resources” (Cutrona, 1996, p.10). This type of group activity has often beenidentified as social or relational in nature (Cline,1999; Hollingshead et al., 2005). In other groupsettings, providing information is often viewedas a task activity, and, certainly, breast cancersurvivors can ask for information in other set-tings (e.g., from nurses or physicians; on theInternet). But the type of caring or relationalsupport that accompanies asking for and givinginformation in a support group setting makes itless clear how the two types of activities areinterdependent.
Group members were women who ranged inage from late 30s to early 70s and were at
17RELATIONAL MESSAGES
various stages of cancer survivorship (e.g.,newly diagnosed, posttreatment, 5–10 years outfrom remission). A volunteer member facili-tated the group conversation; her behaviormatched leader-approved norms identified byLieberman, Golant, and Altman (2004). Mem-bers of the support group sat in a circle forma-tion allowing free-flowing conversation. Thegroup had open membership, was self gov-erned, met weekly, did not charge fees, and didnot follow a formal curriculum.
With permission of the group and informedconsent of each participant, meetings were au-diotaped; a coauthor attended every meeting.Tapes were professionally transcribed and ver-ified by one author. The vernacular, grammarinconsistencies, and speaking nonfluenciescommon in spoken interaction were retained.Following the recording of the meetings, groupmembers were interviewed; they overwhelm-ingly indicated that they felt social support fromhaving attended the meetings.
Interaction analysis is both a theory andmethodology for examining conversation mes-sages sequentially, in order to understand theirrelation and effects on other messages. It isdescribed as “an interdisciplinary method forthe empirical investigation of the interaction ofhuman beings with each other and with objectsin their environment” (Jordan & Henderson,1995, p. 1). Bales’ Interaction Process Analysiswas deemed an appropriate coding scheme forthis study due to its ability to distinguish be-
tween relationally oriented and task-orientedthought units. IPA is an established and foun-dational methodology (McGrath, 1984), hasbeen tested for representational validity (Poole& Folger, 1981), and was chosen for this projectbecause Bales argued it was appropriate forsocial and recreational groups (1950). Further-more, IPA is an inclusive coding scheme thatallows all utterances in a group to be coded,thus eliminating the possibility of message se-lection bias.
Each act is identified according to one ofBales’ twelve functions. The underlying as-sumption is that groups have two goals: main-tenance and task performance. Maintenancegoals are characterized by socioemotional, orrelational, function codes. Task performancegoals are characterized as instrumental and in-formation-oriented. Bales set the two groupfunctions up as a contradiction—that is, agroup’s act-by-act interaction is either mainte-nance-goal oriented or task performance-goaloriented (see Table 1 for IPA coding). Groupinteraction necessitates the balancing of main-tenance (positive and negative evaluations ofother group members) and task performance(asking questions and providing answers aboutthe task) functions in communication. Balesclaims that effective groups are those that en-gage in more positive relational talk than neg-ative relational talk. Yet, on the whole, taskmessages account for nearly three fourths of agroup’s conversation with relational messages
Table 1Bales’ Interaction Process Analysis Functional Codes
Socioemotional areaPositive reactions
1. Shows solidarity/seems friendly: Any act that shows positive feelings towardanother person
2. Shows tension release/dramatizes: Any act that reduces the anxiety that a person orgroup may be experiencing
3. Agrees: Any act that shows acceptance of what another person has saidTask area: Attempted answers 4. Gives suggestions: Any act that offers direction/action for how to engage the task
5. Gives opinions: Any act that advances a belief or value that is relevant to the task6. Gives orientation/information: Any act that reports factual observations or
experiencesTask area: Questions 7. Asks for orientation/information: Any act that requests factual observations or
experiences8. Asks for opinions: Any act that requires a belief or value that is relevant to the task9. Asks for suggestions: Any act that requests direction/action for how to engage the
taskSocioemotional area 10. Disagrees: Any act that shows rejection of what another person has saidNegative reactions 11. Shows tension: Any act that indicates that a person is experiencing anxiety
12. Shows antagonism/seems unfriendly: Any act that shows negative feelings towardanother person
18 KEYTON AND BECK
accounting for the remainder (Bales, 1950a,1999).
Across the five weekly meetings, 256 pagesof transcript were produced. Coders workedwith both transcripts and audio recordings dur-ing coding. Coders were trained to identifythought units (i.e., complete thoughts rangingfrom single words to multiple sentences).Working in teams of two, coders unitized allpages of the meeting transcripts; coefficient re-liability ranged from .90 to .97. Next, using onerandomly selected group transcript, assistantswere trained in Bales’ IPA coding scheme;Scott’s � was .86 to .94 for the IPA categories.Differences in coding application were re-viewed and discussed with one of the authorsuntil consistency was achieved; audio record-ings were consulted. Thought units per meetingranged from 550 to 803 (M � 692.20). Thenumber of speaking turns per meeting rangedfrom 349 to 437 (M � 399.63). The number ofmeeting participants ranged from 7 to 15(M � 11.38). Meeting time ranged from 50to 80 minutes (M � 62.00). A detailed break-down of the IPA analysis is shown in Table 2.
Using IPA in its standard protocol provides away to identify types of acts and produces fre-quencies of types of acts. IPA captures frequen-cies of functions; sequence of functions can also
be retained. To analyze the interdependency ofmessage functions, we followed the proceduresof Fisher and Hawes’ (1971) Interact SystemModel in which the contiguous nature of actsare retained. Acts are also identified by speaker.Doing so is key to maintaining message inter-dependence (see Bonito, 2002), understandingsocial structure, and identifying relational pat-terns (see Courtright, Fairhurst, & Rogers,1989). Maintaining speaker identity and mes-sage interdependence, and considering the con-textual features of the meeting interaction, aidin the discovery of strategies. As Pomerantz andMandelbaum (2005) argue, conversation analyt-ical research has demonstrated that individuals“use their understandings about the activities,motives, rights, responsibilities, and competen-cies” (p. 152) associated with the role relation-ships of their conversational partners. Thus,how group members design and interpret othermembers’ messages depend on these relationalattributes. Our interpretations of group membermessages are based on the descriptive findingsof the interaction in context.
Combining Fisher and Hawes’s (1971) inter-action analysis methodology with Bales’(1950a) IPA, we are able to consider contiguousacts, or interacts, as the unit of analysis, andidentify whether the acts are contributed by
Table 2Interaction Process Analysis Coding of Support Group Meetings
Thought units 803 550 671 644 793Speaking turns 437 349 434 380 355Meeting length, min 65 60 55 50 80Thoughts per minute 12.35 9.17 12.20 12.88 9.91Speaking turns per minute 6.72 5.82 7.89 7.60 4.44
Note. IPA � Interaction Process Analysis; IPA codes 1 through 3 and 10 through 12 � relational categories; IPA codes4 through 9 � task categories.
19RELATIONAL MESSAGES
different group members or by one group mem-ber. By preserving the order of talk, analysis canidentify the different ways in which each type ofrelational talk works in conjunction with typesof task-oriented talk. Instead of simply labelinga thought unit as a function, this analysis exam-ines how the thought unit is performing a role inthe accomplishment of the support group’s re-lational goal. To identify data for this study,each act (i.e., thought unit) coded with one ofthe six relational codes was identified; the actsbefore and after it were also preserved. Thus,the minimum unit of analysis was three acts: anact coded as relational in sequence with the taskacts that preceded and succeeded it. In the dataanalysis that follows, messages coded as rela-tional are presented in italics.
Data Analysis
Consistent with other research using IPAmethodology (as an exception, see Pena &Hancock, 2006, and their adapted use of IPA),the breast cancer support group meetings wereprimarily task oriented (81.16%). The majorityof thought units served the function of givinginformation, with about five times as many at-tempted answers (67.63%) as questions(13.53%). Of the attempted answer acts, groupmembers more frequently gave orientation orinformation followed by opinions; of the ques-tions, members more frequently asked for ori-entation or information. Of the relationalthought units, the interaction was more positive(14.55%) than negative (4.29%).
Slightly more than half (53.85%) of relationalcodes were situated between task talk creatingtask-relational-task talk sequences. Identifica-tion of these sequences revealed that a singlerelational code was commonly isolated betweenlonger strings of coded task talk. In other in-stances, multiple relational codes were se-quenced between and among task talk. Three ofthe six relational codes predominated (agree-ment, disagreement, and shows solidarity) andwere found in all five meetings. Further testsinvestigated whether these three codes wererepresented equally across the five meeting ac-tivities listed by Scheerhorn et al. (1994). Thecodes were initially applied to topic changesthroughout the meetings with separate codesapplied to separate meeting activities within atopic (see Beck, 2008). A 5 meeting activity
(information dissemination, problem-solving,coordination, affiliation, and motivation) � 2IPA code (present, not present) �2 test of inde-pendence was computed for each of the threepredominant relational IPA codes. The test foragreement was not significant, �2(3) � 7.42,p � .060, suggesting that meeting activity typedid not influence the frequency of agreementcodes. This suggests that there is somethingfundamental about the way this support groupused agreement messages. The tests for dis-agreement, �2(3) � 23.50, p � .001, V � 070,and show solidarity, �2(3) � 91.69, p � .001,V � .138, were significant, suggesting thatmeeting activity type created differences in thefrequency of shows solidarity and disagreementcodes. Next, the transcripts were analyzed interms of how the relational code was function-ing in each interaction sequence. In order toaccomplish this purpose and respond to ourresearch questions, and in order to appropriatelymatch our method to our research questions(O’Keefe, 2004), an iterative, qualitative anal-ysis process was used. The second author ex-amined the interaction sequences in terms ofhow the relational code influenced precedingand succeeding messages, writing a brief de-scription for each sequence. Then the first au-thor separately examined each sequence andcompared her interpretation to that of the sec-ond author. When interpretive agreement wasachieved for each sequence, all sequences werecompared for similarity in function and catego-rized accordingly. Last, a final read throughconfirmed that the separate functions were con-sistent with their categorical placement. Thefollowing sections on each emergent categorydetail task-relational-task talk sequences thatoccurred 10 times or more. Considering thatfrequency of use may obscure some functions(Krippendorff, 2004), the analysis then movesto those task-relational-task talk sequences thatoccurred least frequently. The final analysisfocuses on interaction sequences in whichmultiple relational codes were sequenced be-tween task talk.
Frequent Relational Codes
Agreement (IPA 3, positive relational), dis-agreement (IPA 10, negative relational), andshowing solidarity (IPA 1, positive relational)were the most commonly used relational codes
20 KEYTON AND BECK
in task-relational-task interaction sequences inthe support group meetings. The variety of waysin which the relational functions were used inthese meetings is described below.
Agreement
Agreement was identified 136 times acrossthe five meetings and was used in a variety oftask-relational-task interaction sequences. Themost prevalent involved an agreement messagefollowing information dissemination (61%).Other high frequency messages that precededagreement included asking questions (20%), of-fering an opinion (11%), and giving a sugges-tion (8%). Agreement acts were used primarilyin three ways: It (a) confirmed information fromanother speaker (and sometimes extended orqualified it), (b) signaled the direction of theconversation (continuance or transition), and (c)signaled identification of other members. Thesethree functions were not mutually exclusive,and in some agreement messages there wasevidence of all three. Each is described in detail.
Agreement to confirm. Speakers usedagreement messages to confirm information ordecisions offered by other group members. Of-ten these messages were short (i.e., yes, sure), tothe point, and unaccompanied by other mes-sages. There were no indications semanticallythat the speaker was trying to convey anythingelse besides confirming information from an-other group member, or acknowledging that theprevious information was heard.
Other times affirmations were accompaniedby an extension or qualification of the agree-ment. As such, these operated in a way to pro-vide added information to the agreement:
Excerpt #1
MEMBER A. Well, I went back to Advanced HomeCare, and the Jodee manager, whoever, the represen-tative. She is supposed to be finding me some bras thathas—like a deeper band.
MEMBER B. Yeah, I told you. I wear Jodee bras, andyou get a nice wide band.
In this instance, B’s agreement is accompa-nied by language framing A’s information assimilar to what B previously said. Here agree-ment gives added value, in that it not onlyconfirms information, but that information sup-
ports or confirms the previously stated opinionof the second member.
In other interaction, agreement qualificationsintentionally weakened the overall meaning ofthe message. Excerpt #2 is an example:
Excerpt #2
FACILITATOR. Is it hard to do with one hand?
MEMBER. Yeah, sort of. But the wrapping is kind ofelastic wrapping and it grabs.
In this exchange, the message answers affir-matively to the question, but also qualifies theresponse. Due to the uncertainty of the agree-ment qualifier statement, further task thoughtunits (disseminating information) provides amore accurate understanding of the agreement-confirmation thought unit.
Agreement-confirmation thought units wereimportant in how they clarified meaning for othergroup members. When simple agreement mes-sages would not provide an accurate depiction,qualifiers would be added to tailor the messagewith more depth than a simple “yes.” Thus theyplayed a key role in clarifying the previous taskthought units in order to provide informationalsupport, which has been identified as a key ele-ment of social support (Alexander, Peterson, &Hollingshead, 2003; Cutrona, 1996). In addition,agreement-confirmation messages provided cred-ibility to previous group member statements.Credibility confirmations are likely to build groupmember rapport and encourage group members toparticipate. Consistent agreement-confirmationthought units are likely to create a supportivecommunication climate where individuals believetheir participation is respected and recognized(Gibb, 1961).
Agreement signals structure. The secondmost frequent use of agreement was instances inwhich agreement messages signaled the discus-sion structure to other members. An obvious useof agreement to signal direction is when theagreement act signals another member tocontinue speaking. These messages were oftenshort and kept the focus on the first speaker:
Excerpt #3
MEMBER A. Okay, I’ll ask them for it tomorrowwhen I go in.
MEMBERS. Sure; Sure.
21RELATIONAL MESSAGES
MEMBER A. I know they’ve got it, because the firsttime they gave me a couple was when they were inOttawa and they had to get it from their main office inOlathe. And they brought, you know, this big pot ofit—in this huge . . . .
This agreement message not only affirmsMember A’s first statement, but as can be de-termined from A’s second comment, it alsosignaled for her to continue talking. In thisexcerpt it appears that the agreement-structuremessage also signals that nothing was wrongwith her message, thus showing signs of agree-ment-confirmation as well.
An agreement message can also be used tosignal a transition in the conversation, such aswhen someone attempts to change topics:
Excerpt #4
MEMBER A. I’ll have to pay for it out of my pocket.
MEMBER B. Yeah, and I’m having trouble with mybras rolling under . . . .
In this example, B’s “yeah” is used as a bothan acknowledgment of A, but also as a transi-tion signal or pivot, as the rest of the sentenceconcerns a different topic. In essence, B uses asimple agreement to structure closure to A’scomments and to take control of the direction ofthe conversation. The agreement-structurethought unit appears to be an appropriate way tosignal topic change and bring abrupt closure tothe topic without offending A.
Agreement-structure thought units providesubtle structural cues as to the flow of conver-sation. Agreement-structure thought units canmaintain or change topic, as well as confirm thata message is considered appropriate or consis-tent with conversation. Thus multiple tasks canbe accomplished within the normal flow of con-versation. Agreement-structure thought unitsalso prevent more explicit means of structuringconversation topics, such as through interrup-tion. More subtly, agreement-structure thoughtunits have the potential to eliminate relationaltensions. It appears that by using more subtleagreement-structure messages, group memberscan negotiate personal conversation desireswhile allowing other group members the sameopportunity, and at the same time, creating asupportive communication climate.
Agreement signals identification. Agree-ment messages can also signal identification
with other group members; this was the thirdmost frequent use of agreement. Signaling iden-tification was characterized by the use of inclu-sive agreement messages that help group mem-bers to identify with one another. Although afew messages were explicit in indicating sup-port through identification, the majority focusedon implicit inclusive means to offer support.
Excerpt #5
MEMBER A. On a dreary day, I always wear some-thing bright, you know . . .
MEMBER B. I do, too. Now that is one thing as arule, you know, I will usually will force myself toget out.
B’s agreement does not simply affirm A’sstatement, but further affirms that the secondspeaker is able to identify with the experience ofA. This use of agreement provides supportby acknowledging that the first speaker’s opin-ions are not exclusive to her; in other words, Ais not alone, as B voices similar experience.Other agreement-identification messages fo-cused on similarities in diagnosis, side effects,and bonding through the use of humor.
Agreement-identification thought units canbe an important way to accomplish the goal ofproviding social support in the breast cancersupport group meetings. Agreement-identifica-tion thought units worked in conjunction withthe task thought units to accomplish this pur-pose. While the task thought units focused onproviding information, the agreement-identifi-cation thought unit aligned this information toemphasize the relational structures amonggroup members, thus providing social supportin the form of social cohesion or inclusion.Agreement-identification implies greater sym-metrical relational structure among group mem-bers. Group members’ authentic use of agree-ment-identification could help to establish asupportive relational climate of openness forfuture comments (Gibb, 1961).
Disagreement
Disagreement was identified 39 times andwas used in two types of task-relational-taskinteraction sequences. The most prevalent in-volved a common or normative response to aquestion (71.79%). Disagreement was also used
22 KEYTON AND BECK
in refuting information from a previous speaker(28.21%). These two functions were not mutu-ally exclusive. Each is described in detail.
Normative response. The most commonuse of negation came after a closed-ended pre-sumption that a yes/no response would follow.Sometimes these responses were the extent of aparticipant’s contribution to the conversation.
Excerpt #6
MEMBER A. So your bottle didn’t tell you to taketwo?
MEMBER B. Nope.
MEMBER A. Sometimes if they don’t have that 20mg, they . . . .
When B’s requested contribution was com-pleted, then A continued on with the conversa-tion. In this circumstance the disagreement-normative response was task focused in that itprovided information needed for A’s contribu-tion to continue.
Other times the normative response was qual-ified by additional messages. These qualifica-tions were similar to those used in agreementmessages and often elaborated upon thestrength or intensity of the disagreement.
Excerpt #7
MEMBER A. So, I remembered that the doctor tookher off liquid with a meal. Remember that?
MEMBER B. No, I don’t. At all.
MEMBER A. So, I decided, well . . . (laughter).
In this excerpt, B stated her unfamiliaritywith A’s previous message, and then empha-sized this disagreement in absolute terms. Sincean absolute disagreement is clearly contrary towhat A had hoped for, the subsequent taskstatement ignored B’s comments, leading to themembers’ laughter.
Disagreement-normative response thoughtunits were primarily task oriented, as the re-ceiver was simply providing information. Sincethe information had been requested, there wasno clear relational value to the thought units.The interaction continued on without interrup-tion, as if the disagreement-normative responsethought unit was simply another task-orientedpiece of information. It did not appear to add or
subtract from the group’s communicationclimate.
Refute information. In refuting information,disagreement was used to dispute claims madein previous task statements (opinion, sugges-tions, information) by other group members.Unlike the disagreement-normative responsethought units, where the disagreement messagesimmediately responded (in sequence and topic)to the first statement, refuting information ofteninterrupted conversation.
Excerpt #8
MEMBER A. Yeah. I don’t know if you can getseedless, like, blackberry jam or something. But black-berries have a natural anti—diarrhea.
MEMBER B. No, they don’t. – and, uh, I just keep thefrozen ones. You know, if the kids are at home andthey get diarrhea (laughter, multiple conversations).
In this excerpt, A presents a claim as if itwere fact, but B interjects that the claim is false.She does not provide any evidence for her ref-utation and continues on with her conversation.If there was any tension build-up as a result ofthe disagreement-refutation thought unit, it wasprobably minimized through the humor in B’smessages.
Reading the disagreement-refutation thoughtsunits in context suggests that group members didnot intend to negatively affect other groupmembers. This of course is not surprising con-sidering the nature of the group (there were onlythree thought units labeled as showing antago-nism across all meetings). The disagreement-refutation thought units appear to be attempts atclarifying or verifying previous task thoughtunits. Nonetheless, disagreement-refutationthought units can create relational tensions dueto the potential attack on a group member’scredibility. This could potentially discouragegroup member participation. Contrary to agree-ment-identification thought units that implysymmetrical relational structure, disagreement-refutation has the potential to polarize groupmembers and create asymmetrical relationalstructures. An abundance of disagreement-refutation thought units could potentially createan unequal power or relational structure withinthe group and lead to less subtle shifts in con-versation structure, ultimately being detrimentalto the purpose of a breast cancer support group.This may be why the disagreement thought units
23RELATIONAL MESSAGES
in the examples were followed by attempts todiffuse any such tension (e.g., through humor). Inthese meetings, acts that refuted informationwere not as frequent as disagreement-normativeresponse thought units.
Shows Solidarity
Showing solidarity was used in 53 task-relational-task interaction sequences; in all ofthe sequences the solidarity statement followedinformation dissemination. The three highestfrequency sequences were categorized as offer-ing evaluation (26.19%), displaying gratitude(19.05%), and showing concern (16.67%).Showing solidarity was also exemplified bymessages of familiarity, apology, joke, contin-uance, transition, showing excitement, andshowing surprise. These functions were not mu-tually exclusive.
Support as evaluation. One message strat-egy used to show solidarity was to provide anassessment or evaluation of what another mem-ber had said. Oftentimes this was done in theform of a compliment toward another groupmember.
Excerpt #9
MEMBER A. So, it was just the small world . . .
MEMBER B. You seem to have that ability, just toconnect . . .
MEMBER A. Well, I tend to wear something when Itravel, and people will make comments.
In this excerpt, B interjects her assessment ofA’s positive quality. This assessment could becoded as task function, except that it is clearlyintended to be a compliment and show supportto A. Member A does not directly recognize thecompliment, although it is clear her followingresponse is a way of acknowledging the com-pliment without speaking to it directly.
The majority of support-evaluation state-ments in the support group meetings focused oncompliments to others. Such support-evaluationthought units would follow task-oriented mes-sages, such as telling a story or disclosing per-sonal information on the status of an individu-al’s cancer. Based on the subsequent thoughtunit, these compliments were welcome and con-tributed to a supportive communication climate.Compliments such as this excerpt would be an
important mechanism for accomplishing socialsupport.
Support as gratitude. Relational messagesthat displayed gratitude as a function of show-ing solidarity oftentimes followed group mem-bers’ opinions. Subsequent support-gratitudemessages were often considered compliments.In response to the compliment, they would say“thank you.” In this example, even though thefirst member does not clearly state her assess-ment as a compliment, another group memberinterprets it in this light and shows gratitude forthe compliment.
Excerpt #10
MEMBER A. And I’m doing much better, and one ofthe big things that has helped me is this group. It hasreally, really helped. And that was what I was . . . .
MEMBER B. Thank you.
MEMBER C. I was telling [name] today. You need tobe here.
In this excerpt, gratitude is shown for thecontent of the previous message. Moreover, Bappears to be expressing gratitude from herselfas well as on behalf of other group members.
A support group is dependent upon participa-tion in order to accomplish social support; thusit is not surprising that most relational thoughtunits in the meetings contributed to a supportivecommunication climate.
Excerpt #11
MEMBER A. So there is help in the community, too,for folks.
FACILITATOR. Thank you. Yes, I had forgotten thatDr. [name] had been doing that.
In this excerpt, the gratitude is not for thecontent of the first message but for A’s partic-ipation in general. In addition, it provided clo-sure to earlier comments and led to a transitionfor the facilitator.
Support-gratitude thought units were impor-tant in expressing thanks and ascertaining ex-plicit benefits of the support group conversa-tion. Support-gratitude thoughts units could beboth an important part of the flow and structureof conversation, as well as a side note as demon-strated in the first excerpt. The support-gratitudecomments fostered a supportive communicationclimate, and provided an opportunity for group
24 KEYTON AND BECK
members to express thanks specifically to othergroup members or to the group as a whole.
Support as concern. Another aspect ofshowing solidarity was showing concern forothers. Although relational thought units in gen-eral were short, support-concern thought unitswere the shortest. In this relational function,expressions of concern took a variety of verbaland nonverbal forms. For example, not even anentire word was needed to convey concern inthis excerpt.
Excerpt #12
MEMBER A. They ended up, you know, with all ofthis down here—and there was something—I think anerve that was clipped, because I’m still numb right allthe way down to the knee.
FACILITATOR. Ooh.
MEMBER A. But when I had my C-section, there wasalso this, I think the same nerve, that something hap-pened.
Through this utterance the speaker was ableto convey empathy for the plight of A. The useof an emotive utterance may have been partiallydue to the difficulty in expressing this feeling ofconcern in words. Clearly, nonverbals such asprosody and kinesics may have likewise beenimportant to the message.
In addition, support-concern thought units at-tempted to relate or identify with group mem-bers. This would often involve attempts at stat-ing out loud potentially similar thoughts andfeelings as the group members.
Excerpt #13
MEMBER A. They had me scheduled for therapy, butmy chemo was always such a hard thing for them to dothat it would take eight hours. I would be there foreight hours trying to get this done, and it was alwaystoo late . . . .
FACILITATOR. Hard thing for you to do, too . . .
MEMBER A. Do you remind you also not to do bloodpressure on that arm?
In this excerpt, the facilitator takes the posi-tion of A in order to provide a relational mes-sage that empathizes with A’s plight. This isalso helpful in that it confirms that the facilitatoris genuinely listening to what A is saying.
Support-concern messages are often attempt-ing to convey emotion and sympathy to other
group members. They sometimes do not pro-vide any informational value beyond emotions,which can be difficult to convey. As a groupmember attempts to convey emotion, an oppor-tunity is provided for other group members towitness the emotion displayed verbally and re-late to it. The use of support-concern messagescreates an empathic and positive group climate,in which group member are willing to experi-ence the same feelings as the speaker.
Rarely Used Relational Codes
To avoid the fallacy of letting frequency deter-mine importance of talk sequences (Krippendorff,2004), other relational codes were also identi-fied, although with considerable less frequency.Releasing and showing tension will be consid-ered briefly in turn.
Releases tension. Releasing tension thoughtunits were of two general varieties. The firsttype typically involved a statement summariz-ing or assessing the positive aspects of a poten-tially negative situation.
Excerpt #14
MEMBER. They knew, and were, you know, prettyactive in pursuing ruling things out so they ended up,actually, with a diagnosis that had been missed. So,well, all is well. Have you noticed a difference in howyou feel?
In this statement, a member summarizes herown story with a short positive statement. Thestatement brings closure to the story, and allowsher to change the focus to someone else.
The second way tension release was used wasattempting to diffuse the seriousness of sober-ing conversation.
Excerpt #15
MEMBER A. Oh, I’m glad that’s over.
MEMBER B. Oh, that’s so much fun, isn’t it? (laughter).
MEMBER A. My port wouldn’t work half the time.
In this excerpt, a joke is used to release thetension of an otherwise serious matter. Similarutterances that drew laughter from other groupmembers were used, often interrupting anothermember’s interaction. In this excerpt B’s ten-sion release statement plays on A’s disclosureof treatment status.
25RELATIONAL MESSAGES
In both examples, the relational message ofreleasing tension was used to portray a positivevalence of the situation. Clearly, as membersrecount and explain difficult bouts of cancerthere are many opportunities for tension to in-crease. These potentially could be destructive toa supportive communication climate, and a re-lease tension thought unit may be a way toresolve the problem.
Shows tension. Messages that showed ten-sion often involved discussion of how cancerhad debilitated members of the group. Thesecomments tended to be self-directed and oftenwere very explicit, adding to the gravity of theconversation. The shows tension thought unit isembedded in one member’s lengthy comment,which is different from most of the earlier ex-cerpts in which conversation moved quicklybetween or among members.
Excerpt #16
MEMBER A. I think it was more an infection after thesurgery; about a month in from the surgery. I am verydisfigured from that. That was part of it, but I think Iwas just so weak and it just took me so long for me toeven be able to just walk from one side of the room tothe other.
It may be that other group members found itdifficult to interject or interrupt this extendedsegment by one member, particularly given theseriousness of the information being shared. Or,it may be that group members recognized theimplicit (and potentially healing) need for amember sharing this type of information tovoice her own negative evaluation of the cancerexperience. In this case, showing tension is arelease of tension.
Show tension thought units were often em-bedded in retrospective reports, attempting tohelp other support group members dealing withsimilar experiences.
Excerpt #17
MEMBER A. You know, I was just not in very goodshape, I guess. You know, when it happens to you, youdon’t realize how bad you are. My mother keepssaying, “Joy,” how awful it was.
Clearly support groups are designed to helpindividuals with difficult experiences, so nega-tive experiences are an important part of meet-ing discussions in order to accomplish socialsupport. How these negative experiences are
introduced into conversation, and in what termsthey are dealt with may rely heavily on howgroup members negotiate relational messages.
Discussion
To answer the first research question rela-tional messages, as coded with Bales’ (1950a)socioemotional codes, were found to demon-strate three primary functions: agreement, dis-agreement, and showing solidarity. Groupmembers’ use of agreement and showing soli-darity, two positive relational acts, replicatedMacKenzie’s (1983) conceptualization of coun-seling group climate (especially the use of pos-itive relational messages after the sharing ofinformation) and demonstrated that group mem-bers were engaged in supportive conversation(Cline, 1999) with one another. This study alsoreplicated earlier studies that claimed disagree-ment has both task and relational dimensions(Kelly & Spoor, 2007; Ridgeway & Johnson,1990).
In response to the second research question,which sought to describe how relational mes-sages are interdependent with task messages,we found that task-oriented acts dominated thesupport group’s regular meeting interaction andgenerally were used as bookends to relationalacts. However, the frequency of positive to neg-ative relational acts rendered the overall emo-tional tone and attitude of group members aspositive and similar to that found in more for-mal counseling groups (Yalom, 1995). Further-more, the lower frequency of relational mes-sages to task messages demonstrates the powerof relational messages on the overall tone orclimate of a group’s interactions. It is not sur-prising that task dominates because talk doesthings. Talk is a performative act (Grice, 1999)that creates social meaning, and thus, best un-derstood in adjacency pairs (Tracy, 2002).
At the beginning of this article, we noted twoapproaches to identifying task and relationalmessages: (a) Watzlawick et al.’s (1967) posi-tion that relational information occurs in allmessages; and (b) Bales (1950a) position thatmessages are either relational or task. We be-lieve that these data argue for a different posi-tion. That is, task and relational dimensionscoexist in conversation (interchanges amonggroup members) not in a singular message orthought unit. While messages can be labeled as
26 KEYTON AND BECK
relational or task, meaning in context is derivednot from isolated messages, but from interde-pendencies among relational and task messages.Thus, thought unit or message sequence is im-portant to the interpretation. In the case of thissupport group, relational talk was regularly in-serted between task messages. In that way, re-lational messages organized the group’s conver-sation into support and caring whereas the tasktalk moved forward the group’s objective ofhelping one another cope with breast cancer.
Specifically, the data demonstrate how agree-ment acts can function to create a positive com-munication climate. By confirming informationfrom other group members, signaling the direc-tion of the conversation, and signaling identifi-cation with one another, members used agree-ment acts in a variety of ways. Comparingagreement acts with disagreement acts, it ap-pears that agreement produces greater relationalor relational influence whereas disagreementacts are more tightly linked to task-orientedacts. This finding may be a result of the supportgroup context. Alternately, this finding could beinterpreted as the more subtle nature of dis-agreement. Disagreement tied to answeringquestions or refuting information gives the ap-pearance that disagreement is anchored ontasks. However, repetitive disagreement to onegroup member or from one group member maycreate inhibiting relational influences. If dis-agreement is infrequently used and not isolatedto one or a few members, then it may have littleeffect on group member relationships. But itdoes appear that agreement and disagreementmessages can influence a group’s climate, andperhaps ultimately, group members’ cohesionand satisfaction.
Group members showed solidarity by offer-ing evaluation, displaying gratitude, and show-ing concern for another member’s contributionor position. Moreover, the many ways in whichthese subfunctions were delivered suggests thatshowing solidarity is a highly dynamic and robustconstruct. More importantly, these messageswere linked tacitly to task messages demonstrat-ing that providing support is an interactive goal(Wilson, 2007) and generally not interpretablein isolation. Indeed, affirmative and solidarityacts move the thoughts and feelings of individ-uals into existence for other group members(Gergen, Gergen, & Barrett, 2004), and in es-
sence, provide the social support and psycho-social benefits members seek in such a group.
The group’s use of releasing and showingtension was infrequent. However, these interac-tion sequences revealed tender and sometimesironic moments in the group’s conversations. Itmay be that the infrequent use of such actsplaces them as having high value to group mem-bers, as these acts were personal and often ca-thartic—features that are central to recovery andacceptance. These findings also replicateCawyer and Smith-Dupre’s (1995) study of asupport group in which relationship messagescommunicated to heal, prepared members forlife changes, helped members vent emotions,and acted as a mechanism for changing society.
Across relational research in the group con-text (see Keyton, 1999, for a review), research-ers have not agreed whether relational influencein groups is created through how a messagefunctions in a group’s conversations, or how amessage is delivered by a group member. Thisis an important distinction. In the first, research-ers would focus on what the message is and howit is connected and interdependent with othermessages (by the same member, as well asmessages of other group members). In the sec-ond, researchers would focus on how a messageis communicated (e.g., its form, member’s de-livery) by a specific group member. From agroup perspective, we advocate that the formerposition is more important in identifying howgroups work; it is also a superior position ashow a message functions in a group’s interac-tions as it is based in part on how a message isdelivered by a speaker and interpreted by areceiver. A simple task/relational dichotomy isdeficient in its ability to accurately account forhow messages can accomplish both task andrelational outcomes.
The findings from this study provide initialsupport for several assumptions of Keyton’s(1999) relational communication model, withtwo standing out. First, it demonstrates howboth task and relational messages are used inconjunction to accomplish relational purposes.In doing so, the microprocesses of relationalmessages have been explored, revealing subtleways relational messages influence interaction.This provides a foundational step to explorehow relational elements function in all mes-sages. Second, this analysis clearly shows thatrelational outcomes (as well as task outcomes)
27RELATIONAL MESSAGES
are being accomplished throughout group pro-cess, and cannot simply be accounted for at theend of an interaction. Treated as such, relationalsubtleties of interaction are oversimplified.Solely attempting to understand the final rela-tional product of an interaction treats commu-nication as static instead of dynamic, and cog-nition as simple instead of multidimensional.Truly, “relational outcomes can be and arepresent throughout a group’s history” (Keyton,1999, p. 197).
Future Research
Using one breast cancer support group for aninitial examination of the role of relational actsrelative to task acts is certainly a limitation. Weacknowledge this limitation and offset it byconsidering the group’s conversations acrossmultiple meetings. We also argue that our use ofa support group as a critical case study waspurposeful because we expected that relationalinteraction in such a group would be legitimateand heightened. Certainly, the same type ofstudy (using IPA and retaining interaction se-quences) should be conducted in task-orientedgroups for comparison. Of course, another lim-itation is that this support group consisted ofwomen. Other researchers have examined malesupport groups (van Wormer, 1989; Whatule,2000). However, methodological differencesconfound what could be otherwise contributableto differences in sex (see Eisenberg & Lennon,1983). Context may also be more importantthan sex in detecting relational message differ-ences (Goldsmith & Dun, 1997).
Support group research is difficult due toconstraints in gaining access to face-to-facemeetings. Considerably more research has beenconducted on Internet-based support groups. In-teresting to note, few of these studies (, e.g., seeAlexander et al., 2003) have examined the tex-tual conversation among participants. Studiessimilar to the one presented here should also beconducted with online support groups. In doingso, researchers should acknowledge three im-portant distinctions between face-to-face andonline support groups: First, what informationand how it is shared may differ between the twoenvironments. Second, immediacy of relation-ships in a virtual environment may be perceiveddifferently in asynchronous and synchronousmedia. Third, conversations differ in the two
environments. In a face-to-face setting, conver-sation allows for more give and take whereasconversation in the online environment is char-acterized by long textual passages.
Conclusion
In sum, we advocate for group researchers tonot isolate relational and task messages, becausethey cannot be interpreted separately. Thus, meth-odologies must acknowledge both and interpretfindings in light of both. Second, interdependenceor sequence of messages must be retained. Groupsare dynamic entities and task accomplishment andgroup sentiment likely change across a group’sspecific interaction or duration. Third, relationalmessages are infrequent but influential. Simplylooking at frequency counts will provide an in-complete understanding into how relational mes-sages influence conversation.
Analyzing how relational messages functionnot only increases our understanding of rela-tional messages, but also provides a foundationupon which researchers can examine relationalinfluences in all messages. Group members cangenerally agree on the content of a task messageeven if they hold differing evaluations of thecontent. Relational messages, on the other hand,are more subjective and relationship specific(with other group members or the group as anentity). Delivering and responding to relationalmessages can also produce vulnerability forgroup members and the group. Understandinghow relational messages function and the waysin which relational messages influence the di-rection of group conversation are essential tounderstanding how groups complete their tasks.
References
Alexander, S. C., Peterson, J. L., & Hollingshead,A. B. (2003). Help is at your keyboard: Supportgroups on the internet. In L. R. Frey (Ed.), Groupcommunication in context: Studies of bona fidegroups (pp. 309–334). Mahwah, NJ: Erlbaum.
Bales, R. F. (1950a). Interaction process analysis: Amethod for the study of small groups. Cambridge,MA: Addison Wesley.
Bales, R. F. (1950b). A set of categories for theanalysis of small group interaction. American So-ciological Review, 15, 257–263.
Bales, R. F. (1999). Social interaction systems:Theory and measurement. New Brunswick, NJ:Transaction.
28 KEYTON AND BECK
Barker, V. E., Abrams, J. R., Tiyaamornwong, V., Seibold,D. R., Duggan, A., Park, H. S., et al. (2000). Newcontexts for relational communication in groups. SmallGroup Research, 31, 470–503.
Beck, S. J. (2008). The communicative creation ofmeetings: An interaction analysis of meetingthought units and meeting activities in three natu-ral meeting contexts. Unpublished doctoral disser-tation, University of Kansas, Lawrence.
Benne, K., & Sheats, P. (1948). Functional roles ofgroup members. Journal of Social Issues, 4, 41–49.
Bion, B. R. (1948). Experiences in groups. HumanRelations, 1, 314–320.
Bonito, J. A. (2002). The analysis of participation insmall groups: Methodological and conceptual is-sues related to interdependence. Small Group Re-search, 33, 412–438.
Cawyer, C. S., & Smith-Dupre, A. (1995). Commu-nicating social support: Identifying supportive ep-isodes in HIV/AIDS support group. Communica-tion Quarterly, 43, 243–258.
Chang, A., & Bordia, P. (2001). A multidimensionalapproach to the group cohesion-group perfor-mance relationship. Small Group Research, 32,379–405.
Cline, R. J. W. (1999). Communication in socialsupport groups. In L. R. Frey, D. S. Gouran, &M. S. Poole (Eds.), The handbook of group com-munication theory and research (pp. 516–538).Thousand Oaks, CA: Sage.
Courtright, J. A., Fairhurst, G. T., & Rogers, L. E.(1989). Interaction patterns in organic and mech-anistic systems. Academy of Management Jour-nal, 32, 773–802.
Cutrona, C. E. (1996). Social support in couples.Thousand Oaks, CA: Sage.
DiSalvo, V. S., Nikkel, E., & Monroe, C. (1989).Theory and practice: A field investigation andidentification of group members’ perceptions ofproblems facing natural work groups. Small GroupBehavior, 20, 551–567.
Eisenberg, N., & Lennon, R. (1983). Sex differencesin empathy and related capacities. PsychologicalBulletin, 94, 100–131.
Fisher, B. A., & Ellis, D. G. (1990). Small groupdecision making: Communication and the groupprocess. New York: McGraw-Hill.
Fisher, B. A., & Hawes, L. (1971). An interact systemmodel: Generating a grounded theory of smallgroups. Quarterly Journal of Speech, 57, 444–453.
Gergen, K. J., Gergen, M. M., & Barrett, F. J. (2004).Dialogue: Life and death of the organization. In D.Grant, C. Hardy, C. Oswick, & L. Putnam (Eds.),The Sage handbook of organizational discourse(pp. 39–59). London: Sage.
Gibb, J. R. (1961). Defensive communication. Jour-nal of Communication, 11, 141–148.
Goldsmith, D. J., & Dun, S. A. (1997). Sex differ-ences and similarities in the communication ofsocial support. Journal of Social and PersonalRelationships, 14, 317–337.
Grice, H. P. (1999). Logic and conversation. In A.Jaworski & N. Coupland (Eds.), The discoursereader (pp. 76–88). London: Routledge.
Henry, K. B., Arrow, H., & Carini, B. (1999). A tripar-tite model of group identification: Theory and mea-surement. Small Group Research, 30, 558–581.
Hollingshead, A. B., Wittenbaum, G. M., Paulus,P. B., Hirokawa, R. Y., Ancona, D. G., Peterson,R. S., et al. (2005). In M. S. Poole & A. B.Hollingshead (Eds.), Theories of small groups:Interdisciplinary perspectives (pp. 21–63). Thou-sand Oaks, CA: Sage.
Homans, G. C. (1950). The human group. Orlando,FL: Harcourt Brace.
Jordan, A., & Henderson, B. (1995). Interaction anal-ysis: Foundations and practice. Journal of theLearning Sciences, 4, 39–103.
Kelly, J. R., & Spoor, J. R. (2007). Naı̈ve theoriesabout the effects of mood in groups: A preliminaryinvestigation. Group Processes and IntergroupRelations, 10, 203–222.
Keyton, J. (1991). Evaluating individual group mem-ber satisfaction as a situational variable. SmallGroup Research, 22, 200–219.
Keyton, J. (1999). Relational communication ingroups. In L. R. Frey, D. S. Gouran, & M. S. Poole(Eds.), The handbook of group communication the-ory and research (pp. 192–222). Thousand Oaks,CA: Sage.
Knobloch, L. K., & Solomon, D. H. (2005). Rela-tional uncertainty and relational information pro-cessing: Questions without answers? Communica-tion Research, 32, 349–388.
Krippendorff, K. (2004). Content analysis: An intro-duction to its methodology. Thousand Oaks, CA:Sage.
Lewin, K. (1951). Field theory in social science:Selected theoretical papers. New York: Harper &Row.
Lieberman, M. A., Golant, M., & Altman, T. (2004).Therapeutic norms and patient benefit: Cancer pa-tients in professionally directed support groups.Group Dynamics: Theory, Research, and Prac-tice, 8, 265–276.
MacKenzie, K. R. (1983). The clinical application ofa group climate measure. In R. R. Dies & K. R.MacKenzie (Eds.), Advances in group psychother-apy: Integrating research and practice (pp. 159–170). New York: International Universities Press.
McGrath, J. E. (1984). Groups: Interaction and per-formance. Englewood Cliffs, NJ: Prentice Hall.
Mudrack, P. E., & Farrell, G. M. (1995). An exam-ination of functional role behavior and its conse-
29RELATIONAL MESSAGES
quences for individuals in group settings. SmallGroup Research, 26, 542–571.
O’Keefe, D. J. (2004). The unity of argument acrossmethodological divides. Paper presented at themeeting of the International Communication As-sociation, New Orleans, LA.
Pena, J., & Hancock, J. T. (2006). An analysis ofsocioemotional and task communication in onlinemultiplayer video games. Communication Re-search, 33, 92–109.
Pomerantz, A., & Mandelbaum, J. (2005). A conver-sation analytic approach to relationships: Theirrelevance for interactional conduct. In K. Fitch &R. E. Sanders (Eds.), Handbook of language andsocial interaction (pp. 149–171). Mahwah, NJ:Erlbaum.
Poole, M. S., & Folger, J. P. (1981). A new methodof establishing the representational validity of in-teraction coding schemes: Do we see what theysee? Human Communication Research, 8, 26–42.
Ridgeway, C., & Johnson, C. (1990). What is therelationship between socioemotional behavior andstatus in task groups? The American Journal ofSociology, 95, 1189–1212.
Salazar, A. J. (1995). Understanding the synergisticeffects of communication in small groups: Makingthe most out of group member abilities. SmallGroup Research, 26, 169–199.
Scheerhorn, D., Geist, P., & Teboul, J. B. (1994).Beyond decision making in decision-making
groups: Implications for the study of group com-munication. In L. R. Frey (Ed.), Group communi-cation in context: Studies of natural groups (pp.247–262). Hillsdale, NJ: Erlbaum.
Schutz, W. C. (1966). The interpersonal underworld.Palo Alto, CA: Science & Behavior Books.
Tracy, K. (2002). Everyday talk: Building and re-flecting identities. New York: Guilford Press.
Van Wormer, K. (1989). The male-specific group inalcoholism treatment. Small Group Research, 20,228–242.
Watzlawick, P., Beavin, J., & Jackson, D. (1967).Pragmatics of human communication: A study ofinteractional patterns, pathologies, and para-doxes. New York: Norton.
Whatule, L. J. (2000). Communication as an aid toresocialization: A case study of a men’s angergroup. Small Group Research, 21, 424–440.
Wilson, S. R. (2007). Communication theory and theconcept of “goal.” In B. B. Whaley & W. Samter(Eds.), Explaining communication: Contemporarytheories and exemplars (pp. 73–104). Mahwah,NJ: Erlbaum.
Yalom, I. D. (1995). The theory and practice ofgroup psychotherapy (4th ed.). New York: BasicBooks.
Received July 24, 2007Revision received July 8, 2008
