Constructive conflict at work

Journal of Organizational Behavior J. Organiz. Behav. 20, 475±491 (1999) Constructive con¯ict at work EVERT VAN DE VLIERT*, AUKJE NAUTA, ELLEN GIEBELS AND ONNE JANSSEN Department of Psychology, University of Groningen, Grote Kruisstraat 2/1, 9712 TS Groningen, The Netherlands Summary Two ®eld studies of interpersonal con¯ict in organizations and a simulated dyadic negotiation show that problem solving and forcing are frequently combined simultaneously and sequentially. As a robust ®nding, conglomerations of problem solving and forcing appear to enhance the parties' joint e€ectiveness. The best substantive and relational outcomes for the organization are reached through forcing followed by problem solving. Copyright # 1999 John Wiley & Sons, Ltd. Introduction Con¯ict behavior is so omnipresent in organizational life that we too easily take it for granted. Con¯ict arises at work when an employee is obstructed or irritated by another person or a group. The con¯ict can be handled in either a constructive or a destructive way. Workers usually react by choosing a well-trodden path: they ignore a reproach, they accommodate a poor plan, they compromise on a price or solve a problem, and sometimes they force an opponent to yield to their demands (e.g. Blake and Mouton, 1964; Rahim, 1992; Rubin, Pruitt and Kim, 1994; Thomas, 1992; Van de Vliert and Kabano€, 1990). The present article addresses the question of how these daily con¯ict behaviors relate to organizational e€ectiveness. It focuses on the e€ectiveness of two reactions, problem solving and forcing, in situations of dyadic interpersonal con¯ict and negotiation. Speci®cally, it challenges both the classic assumption that problem solving is the most constructive reaction (e.g. Blake and Mouton, 1970, 1981) as well as the modern contingency assumption that the e€ectiveness of both problem solving and forcing depends on situational characteristics (e.g. Rahim, 1992). The ®rst section provides de®nitions of problem solving, forcing and e€ectiveness, and an overview of four perspectives on dyadic e€ectiveness. In the next section, three studies are presented that demonstrate the usefulness of a `complexity perspective' by examining the dyadic e€ectiveness of simultaneous and sequential combinations of problem solving and forcing. * Correspondence to: Evert Van de Vliert, Department of Psychology, University of Groningen, Grote Kruisstraat 2/1, 9712 TS Groningen, The Netherlands. Tel: (31) 503 636 386. Fax: (31) 503 636 304. E-mail: E.Van.de.Vliert@ppsw. rug.nl CCC 0894±3796/99/040475±17$17.50 Copyright # 1999 John Wiley & Sons, Ltd. Accepted 23 March 1997 476 E. VAN DE VLIERT ET AL. Theoretical Background De®nitions Con¯icting or negotiating parties attempt to further their own interests, their opponent's interests, or both. Problem solving is reconciling both parties' basic interests. Forcing is furthering one's own interests by contending the opponent in a direct way. Whereas problem solving appears to be a very agreeable, de-escalative, and e€ective mode of handling con¯ict at work, forcing is a very disagreeable, escalative, and ine€ective alternative (Andrews and Tjosvold, 1983; Burke, 1970; Likert and Likert, 1976; Rahim, 1983; Tjosvold, 1991; Volkema and Bergmann, 1989; Van de Vliert and Euwema, 1994; Van de Vliert, Euwema and Huismans, 1995). Con¯ict behavior is viewed as personally e€ective to the extent that an individual succeeds in realizing the bene®ts desired for oneself. Personal e€ectiveness may or may not be in accord with dyadic e€ectiveness. Dyadic e€ectiveness, or constructiveness, is the extent to which con¯ict behavior produces better outcomes for the organizational dyad by resolving the con¯ict, improving the relationship between the parties, or both (cf. Janssen and Van de Vliert, 1996; Nauta, 1996; Thomas, 1992; Tjosvold, 1991). Perspectives on dyadic e€ectiveness A one-best-way perspective has been advocated by a number of scholars who focus on behavioral strategies (e.g. Blake and Mouton, 1970, 1981; Fisher and Ury, 1981; Pneuman and Bruehl, 1982). Their work demonstrates that, compared to avoidance, accommodation, compromise, and force, problem solving is a more constructive mode of con¯ict management because it unites people against the issue; it stimulates creativity and is therefore more mutually bene®cial. In contrast, proponents of the contingency perspective contend that what is e€ective can best be determined in the light of situational realities. Each mode of con¯ict management is appropriate under some circumstances because speci®c situations require unique and appropriate responses (e.g. Axelrod, 1984; Hocker and Wilmot, 1991; Rahim, 1992). Thomas (1992) made a big leap forward by integrating the preceding perspectives through the help of short- and long-term time horizons. In his time perspective he asserts that the contingency approach provides answers to the short-term question of how best to cope with the here and now, while the `one-best-way' approach deals with the longer term task of creating desirable future circumstances for the organization. Though the one-best-way, contingency, and time perspectives on constructive con¯ict are theoretically di€erent, they have two implicit assumptions in common. The ®rst is that people behave in one pure way and not in any other way. That is, they react to con¯ict issues with either problem solving or forcing, or another mode of handling con¯ict, but not with two or more behaviors at the same time. The second assumption is that people reach better substantive and relational outcomes by means of pure behavior, not by a combination of behaviors. To supplement the prevailing three perspectives, we propose the addition of a complexity perspective, which is based on the following alternative assumptions. The ®rst assumption holds that a reaction to a con¯ict issue consists of multiple behavioral responses rather than a single and pure mode of behavior. Indeed, combinations of avoiding, accommodating, compromising, problem solving, and forcing are the rule rather than the exception (Falbe and Yukl, 1992; Knapp, Putnam and Davis, 1988; Rubin et al., 1994; Yukl, Falbe and Young Youn, 1993). `Tacit coordination' is such a combination of reactions: people stand ®rm and withhold relevant information while expressing obligingness and their real interests through nonverbal cues Copyright # 1999 John Wiley & Sons, Ltd. J. Organiz. Behav. 20, 475±491 (1999) CONSTRUCTIVE CONFLICT 477 (e.g. Boriso€ and Victor, 1989; Pruitt, 1981; Putnam, 1990). We have coined the term `conglomerated con¯ict behavior' for the simultaneous or sequential aggregation of various degrees of several modes of con¯ict behavior (Van de Vliert, 1997; Van de Vliert et al., 1995). The second, and related, assumption holds that conglomerated modes of con¯ict behavior have joint e€ects on the substantive and relational outcomes of the con¯ict. For example, `log-rolling,' the combination of being exacting about a bene®t or cost that is important to oneself but unimportant to the opponent, while being obliging about a bene®t or cost that is unimportant to oneself but important to the opponent, is an e€ective form of conglomerated con¯ict behavior (cf. Fisher, Ury and Patton, 1991; Rubin et al., 1994; Tutzauer and Rolo€, 1988). The above line of reasoning is supported by an unpublished laboratory experiment (Euwema and Van de Vliert, 1996) that showed that senior nurses in Dutch hospitals tend to handle con¯icts more e€ectively when they combine problem solving and forcing into a behavioral compound that might be seen as a form of `constructive controversy' (cf. Johnson, Johnson and Smith, 1989; Tjosvold, 1985, 1991). Dyadic e€ectiveness of simultaneous as well as sequential conglomerations of problem solving and forcing make sense for at least two reasons. First of all, most con¯ict issues are so complex that problem solving might be more appropriate for some aspects of the discord, while forcing might be more appropriate for other aspects. For example, problem solving is especially appropriate for merging insights as well as working through a negative feeling, whereas forcing is especially appropriate for responding to an emergency as well as implementing an important but unpopular decision (Rahim, 1992; Thomas, 1992). As a consequence, e€ectiveness might well increase if problem solving is competently interspersed with forcing, or vice versa. In the second place, each component of conglomerated con¯ict behavior comes with its own risks in terms of ine€ectiveness. Among the risks of problem solving is the possibility that the parties' negative mutual relationship will block high joint outcomes, or that the ultimate solution is not worth the necessary investment of time and energy, or that so-called win±win outcomes are out of the question (Hocker and Wilmot, 1991). The risks of forcing include escalation beyond acceptable cost limits, a deterioration of the relationship, and a stalemate resulting from the failure of contentious tactics (Rubin et al., 1994). The juxtaposition or alternation of problem solving and forcing might reduce both types of risks. To paraphrase Walton, CutcherGershenfeld, and McKersie (1994, p. 337), who developed a similar argument and reported some empirical support for it: the combination of the two strategies can enhance e€ectiveness by minimizing the tendency for problem solving to produce stagnation and the tendency for forcing to produce escalation. Thus, we felt con®dent hypothesizing that the dyadic e€ectiveness of conglomerated con¯ict behavior is a positive function of problem solving (hypothesis 1), a negative function of forcing (hypothesis 2), and a positive function of problem solving in combination with forcing (hypothesis 3). All three hypotheses were tested in each of the three studies, albeit in very di€erent ways. Study 1 Overview The ®rst investigation was carried out in a national management development center of the police force. As part of their regular training activities, dyads of unit managers handled an easily Copyright # 1999 John Wiley & Sons, Ltd. J. Organiz. Behav. 20, 475±491 (1999) 478 E. VAN DE VLIERT ET AL. recognizable, standardized con¯ict of interest. The role play was videotaped for both feedback and research purposes. Neutral observers later recorded the occurrence of problem solving and forcing, and assessed the e€ectiveness of the conglomerated con¯ict behavior. Method Participants Sixty-three male and 15 female unit managers from Dutch police organizations participated in the study. Their age ranged from 24 to 55 years (M ˆ 41 years). No dyad members were co-workers in daily life. Con¯ict simulation Pilot studies resulted in the following con¯ict scenario. The participant is a manager of a police unit in a large city. The chief of the police district has `created' a budget to appoint a clerical sta€ member to assist the unit manager. The participant (A) and another unit manager (B) are badly in need of an assistant, albeit for di€erent reasons (the scenario then speci®es these reasons for both A and B). The chief of the district has given A and B the following option to obtain the extra sta€ position: they can both hand in a joint written request for an assistant or jointly orally motivate their request in a meeting with the Board of the District. The choice between a written proposal and an oral presentation complicates the situation. A, who is a good writer, prefers a written proposal, whereas B, who is good at oral presentations, prefers it the other way. A and B have made an appointment to discuss the issue and to decide in favor of either the written proposal or the oral presentation. The participants were randomly assigned to dyads and then to the roles of A and B. When they arrived in the simulation room, the experimenter had them read the con¯ict scenario and identify with their roles. The dyad then had about 10 minutes to enact the con¯ict. Each interaction was videotaped and later coded. Measurements Four trained observers, who were unaware of the goals of the study, assessed each individual's use of problem solving and forcing. The following operationalizations were used. Problem solving: exchanges accurate information; works through di€erences; deals with each party's concerns. Forcing: makes an e€ort to get one's way; uses power to win; makes authority decisions. Each coder independently indicated the use of problem solving and forcing by completing a single 7-point rating scale anchored by not at all (1) and to a great extent (7). The ratings were done for the ®rst and second half of the interaction, and for A and B separately (the correlations between A's and B's behavior ranged from 0.66 to 0.85, Mr ˆ 0.75), after which each observer's average scores for the dyad were calculated. The coders' average ratings were then combined into one measure for dyadic problem solving (Cronbach's a ˆ 0.73) and one for dyadic forcing (a ˆ 0.92). The substantive and relational outcomes of the total con¯ict interaction were independently rated by the same observers. They used 7-point rating scales. As speci®ed below, each of these scales is anchored by appropriate operationalizations of very ine€ective (1) and very e€ective (7). The substantive outcomes were measured with the following ®ve items addressing: the ultimate number of con¯ict issues (very large (1)±very small (7)), the severity of these issues (very severe Copyright # 1999 John Wiley & Sons, Ltd. J. Organiz. Behav. 20, 475±491 (1999) CONSTRUCTIVE CONFLICT 479 (1)±not severe (7)), proximity to a solution (very remote (1)±very near (7)), the chances of recidivism (increased considerably (1)±decreased considerably (7)), and the quality of concerted task performance (got much worse (1)±got much better (7)). Likewise, the relational outcomes were measured with ®ve items addressing: the attention given to commonalities (very little (1)Ð very much (7)), the ultimate amount of mutual distrust (increased considerably (1)±decreased considerably (7)) and mutual understanding (decreased considerably (1)±increased considerably (7)), as well as the ultimate atmosphere (much more unfriendly (1)±much more friendly (7)) and personal relationship (got much worse (1)±got much better (7)). Averaged within coders (a ranges from 0.95 to 0.97) and then across coders (a ˆ 0.93), these 10 items were additively combined into one scale of dyadic e€ectiveness. Validity Immediately after the role play, the participants completed the DUTCH questionnaire (Janssen and Van de Vliert, 1996) regarding their own problem solving (a ˆ 0.80), their own forcing (a ˆ 0.71), and the 10 dyadic e€ectiveness items given above (a ˆ 0.92). Within dyads, A's and B's perceptions were summed (between As and Bs r ˆ 0.51, p 5 0.001 for problem solving, r ˆ 0.18, n.s. for forcing, and r ˆ 0.71, p 5 0.001 for e€ectiveness). The subjective dyadic measures were signi®cantly related to the observers' corresponding ratings for problem solving (r ˆ 0.56; p 5 0.001), forcing (r ˆ 0.73; p 5 0.001), and dyadic e€ectiveness (r ˆ 0.72; p 5 0.001), which attests to the validity of the observational measures of con¯ict behavior and e€ectiveness. Results Table 1 provides the means, standard deviations, and intercorrelations of problem solving, forcing, and e€ectiveness. Note that problem solving and forcing are negatively interrelated and have opposite positive and negative relations with e€ectiveness, respectively. Since problem solving and forcing are continuous variables, the hypothesized interactive e€ect was best analyzed through hierarchical regression (Cleary and Kessler, 1982; Cohen and Cohen, 1983). First the main e€ects of problem solving and forcing were entered into the regression equation, after which the two-way interaction term was included. To facilitate interpretation and minimize problems of multicollinearity, the analysis was conducted with standardized variables (Aiken and West, 1991). Supporting hypotheses 1 and 2, problem solving (b ˆ 0.28, p 5 0.05) and forcing (b ˆ ÿ0.65, p 5 0.001) together account for 72 per cent of the variance in dyadic e€ectiveness (F ˆ 45.62, p 5 0.001). The interaction e€ect over and above the main e€ects is also signi®cant (b ˆ 0.31, Table 1. Means, standard deviations, and intercorrelations of problem solving, forcing, and e€ectiveness: Study 1 1. Problem solving 2. Forcing 3. E€ectiveness M S.D. 1 2 2.95 3.06 3.05 0.94 1.00 1.15 Ð ÿ0.59* 0.66* Ð ÿ0.82* N ˆ 39; scales range from 1 to 7. * p 4 0.01. Copyright # 1999 John Wiley & Sons, Ltd. J. Organiz. Behav. 20, 475±491 (1999) 480 E. VAN DE VLIERT ET AL. DR2 ˆ 6 per cent, F ˆ 9.77, p 5 0.01). In agreement with hypothesis 3, e€ectiveness is a positive function of problem solving in combination with forcing. Not surprisingly, dyads were especially ine€ective if they handled the con¯ict predominantly with forcing behavior without sucient problem solving. However, they prevented ine€ectiveness when they combined their forcing moves with higher levels of problem solving. This ®nding con®rms our claim that a combination of problem solving and forcing might reduce the risks inherent in the latter behavior in terms of e€ectiveness reduction. Additional analyses with A's or B's individual use of problem solving and forcing as predictors of dyadic e€ectiveness yielded the same pattern of results. Discussion Managers of Dutch police units who handled a con¯ict with a colleague appeared to produce good substantive and relational outcomes for their organization when they used as little force as possible. However, if contentious management cannot be avoided, these police managers are advised to manifest conglomerated con¯ict behavior. The data clearly demonstrate that e€ective unit managers who used power to get their way combined forcing and problem solving. The ®ndings do not completely support the one-best-way view that problem solving serves the organization better than do other reactions because the e€ect of problem solving is quali®ed by the joint e€ect of problem solving and forcing. Moreover, using a little force is an even better option, given that it takes less time, energy, and skill than problem solving. Likewise, the ®ndings do not completely support our inference from the contingency view that either problem solving or forcing will be most appropriate under the simulated circumstances. The complexity perspective fares better inasmuch as police managers who used force were more e€ective to the extent that they also exchanged information and views dealing with both parties' concerns (cf. Janssen and Van de Vliert, 1996; Johnson et al., 1989; Tjosvold, 1985). Indeed, Study 1 supported our point of departure that problem solving and forcing do have joint e€ects. However, Study 1 leaves several other questions unanswered. To begin with, is it problematic that this ®eld experiment employed a single group of observers for the assessment of the conglomerated con¯ict behavior and its e€ectiveness? A common source bias cannot be excluded with conviction. We therefore replicated Study 1, ®rst in a laboratory experiment that enabled us to use an objective measure of e€ectiveness (Study 2), then in another ®eld experiment that used independent groups of observers for the assessment of conglomerated con¯ict behavior and e€ectiveness (Study 3). In the second place, are the ®ndings about dyads working in police organizations generalizable to dyads working in free enterprises and other types of organizations? To answer this question, business students drawn from di€erent specializations were used in Study 2, while employees from a wide variety of organizations were used in Study 3. In considering the complexity perspective, the most serious queries concern the distinction between simultaneous and sequential conglomerations of problem solving and forcing. Was the e€ectiveness of the combination of problem solving and forcing, that was found, primarily caused by the juxtaposition or the alternation of problem solving and forcing? And if alternation makes conglomerated con¯ict behavior more e€ective, is this because problem solving precedes or follows forcing? Not only were the three hypotheses retested in Studies 2 and 3, the studies also examined how problem solving and forcing can best be conglomerated to enhance dyadic e€ectiveness. Copyright # 1999 John Wiley & Sons, Ltd. J. Organiz. Behav. 20, 475±491 (1999) CONSTRUCTIVE CONFLICT 481 Study 2 Overview Business students participated in a face-to-face simulation of a dyadic, interpersonal buyer±seller negotiation with integrative potential; that is, a negotiation in which joint gains beyond simple compromise were hidden but possible. Each role play was audiotaped and later rated for the occurrence of problem solving and forcing by two trained observers. The extent to which an integrative agreementÐa high joint outcome for the dyadÐwas reached, served as the measure of e€ectiveness of the conglomerated con¯ict behavior. Method Participants Sixty-four dyads of business students from the University of Groningen participated in the study. Their age ranged from 20 to 24 years (M ˆ 21 years). There were 31 male±male dyads and 33 female±female dyads. Dyads were formed on a random basis and care was taken that no dyad members were acquainted with each other. Con¯ict simulation Each participant was randomly assigned the role of either buyer or seller, and was led to either the `buyer room' or the `seller room.' After they were given oral and written role instructions, the buyer and seller were seated at the same table in the negotiation room to reach agreement in 20 minutes at the most. The negotiation task, dealing with the trade of household appliances, was similar to those employed in much previous research on bargaining (for details, see Pruitt and Carnevale, 1993). The dyads negotiated a deal on four issues, to wit: price, delivery time, payment by instalments, and term of guarantee. Buyer and seller were given their own pro®t schedule regarding price, delivery, payment, and guarantee. The task had integrative potential in that the most valuable issue for the buyer (payment) was the least valuable for the seller, whereas the most valuable issue for the seller (price) was the least valuable for the buyer. If the dyad members made larger concessions on their less valuable issues and smaller concessions on their more valuable issues, the joint pro®t exceeded the obvious but suboptimal settlement of a 50±50 split on all issues. Measurements Each negotiation was audiotaped and later transcribed. A transcript contained 82 speaking turns on average, a speaking turn being de®ned as everything a person says between the other's last utterance and the other's next utterance. Two observers, who were unaware of the goals of the study, were trained to record the occurrence of problem solving and forcing. Problem solving included making a cooperative statement, asking for information about the other's pro®t positions and priorities, giving information about one's own pro®t positions and priorities, and showing insight into the other's pro®t positions or priorities. Forcing was arguing one's own standpoint and sticking to it, making threats, and being condescending to the opponent. Transcripts scored by both judges produced interrater agreement in 83 per cent of the cases. Copyright # 1999 John Wiley & Sons, Ltd. J. Organiz. Behav. 20, 475±491 (1999) 482 E. VAN DE VLIERT ET AL. The joint outcome, the sum of the individual outcomes of each of the dyad members, varied from 990 to 1160 (M ˆ 1099, S.D. ˆ 70). Predictors Each speaking turn was categorized as re¯ecting no problem solving versus problem solving, as well as no forcing versus forcing. Next, we created three types of categorized speaking turns. First, pure problem solving (P): problem solving with no forcing. Second, pure forcing (F): forcing with no problem solving. Third, concurrence (C): problem solving combined with forcing. For each dyad, we then computed the percentage of speaking turns coded as P, F, and C, respectively. These three predictors dealt with the absence or presence of simultaneous conglomeration of problem solving and forcing within speaking turns. A number of additional predictors dealt with the sequential conglomeration of problem solving and forcing in one dyad member's speaking turn at time t and problem solving and forcing in the same dyad member's next speaking turn at time t ‡ 2 (after the other dyad member's speaking turn at t ‡ 1). The fourth predictor was each dyad's percentage of two subsequent own speaking turns coded as P at time t and F at time t ‡ 2 (PF). Conversely, the ®fth predictor was each dyad's percentage of two subsequent own speaking turns coded as F at time t and P at time t ‡ 2 (FP). Four similar predictors indicated P at time 1 and C at time 2 (PC), F at time 1 and C at time 2 (FC), C at time 1 and P at time 2 (CP), and C at time 1 and F at time 2 (CF). Validity Immediately after the role play, the participants completed a questionnaire using 5-point rating scales anchored by not (1) and very much so (5). Four items measured their opponent's problem solving (e.g. attempted to search for an optimal solution; a ˆ 0.86 and 0.87 for As and Bs, respectively). Similarly, four items measured their opponent's forcing (e.g. attempted to impose his/her conviction; a ˆ 0.73 and 0.77 for As and Bs, respectively). Within dyads, A's and B's perceptions of the other's problem solving (r ˆ 0.14, n.s.) and forcing (r ˆ 0.15, n.s.) were summed. The subjective dyadic measures were signi®cantly related to the observers' corresponding ratings for problem solving (r ˆ 0.48, p 5 0.001) and forcing (r ˆ 0.59, p 5 0.001), which attests to the validity of the observational measures of con¯ict behavior. Results Table 2 contains the descriptive statistics. Note that pure forcing is used much more than pure problem solving, concurrence, and sequences of problem solving and forcing. As in Study 1, pure problem solving and pure forcing are negatively interrelated and have opposite positive and negative relations with e€ectiveness, respectively. In addition to pure problem solving, forcing followed by concurrence has a signi®cant positive relation with e€ectiveness. Regression analysis based on the z-scores of all the variables showed that pure problem solving (DR2 ˆ 17 per cent, b ˆ 0.42, F ˆ 12.87, p 5 0.001) and pure forcing followed by concurrent problem solving and forcing (DR2 ˆ 5 per cent, b ˆ 0.23, F ˆ 4.00, p 5 0.05) account for 22 per cent of the variance in e€ectiveness. These results imply a rejection of hypothesis 2 because, in and of itself, pure forcing does not contribute to e€ectiveness. The two e€ects support hypotheses 1 and 3, respectively. The more dyad members used problem solving behavior, and the more they used forcing behavior that subsequently added both problem solving and forcing Copyright # 1999 John Wiley & Sons, Ltd. J. Organiz. Behav. 20, 475±491 (1999) Copyright # 1999 John Wiley & Sons, Ltd. Table 2. Means, standard deviations, and intercorrelations of problem solving, forcing, and e€ectiveness: Study 2 Simultaneous Pure problem solving (P) Pure forcing (F) Concurrence (C) Sequential P ®rst, F next (PF) F ®rst, P next (FP) P ®rst, C next (PC) F ®rst, C next (FC) C ®rst, P next (CP) C ®rst, F next (CF) S.D. F C PF 11.23 21.75 1.14 9.84 14.05 1.70 ÿ0.55* 0.22{ ÿ0.16 ÿ0.04 0.20 0.09 1.29 1.53 0.22 0.16 0.11 0.31 1.31 1.69 0.70 0.45 0.43 0.71 1098.91 70.17 FP PC FC CP 0.19 0.29{ 0.12 0.33* ÿ0.16 0.58* 0.20 ÿ0.04 0.42* ÿ0.09 0.03 0.35* 0.47* ÿ0.08 0.13 0.13 0.31* 0.07 0.21 0.02 ÿ0.09 0.02 CF E 0.04 0.08 0.50* 0.42* ÿ0.28{ 0.17 ÿ0.05 0.14 0.19 0.33* ÿ0.04 ÿ0.15 0.17 0.12 0.30* ÿ0.17 0.07 N ˆ 64 dyads; scales indicate percentages of speaking turns (P, F, C) or transitions of preceding speaking turns to following speaking turns (PF, FP, PC, FC, CP, CF). * p 5 0.01; { p 5 0.05. CONSTRUCTIVE CONFLICT 483 J. Organiz. Behav. 20, 475±491 (1999) E€ectiveness (E) M 484 E. VAN DE VLIERT ET AL. behavior, the more e€ectively they handled their con¯ict. Or in words that underscore the relevance of the complexity perspective re¯ected in hypothesis 3: the dyadic e€ectiveness of conglomerated con¯ict behavior is a positive function of both problem solving and problem solving preceded and accompanied by forcing. As in Study 1, additional analyses with A's or B's individual use of problem solving and forcing as predictors of dyadic e€ectiveness yielded the same pattern of results. Discussion Study 2 successfully introduced the time dimension by examining not only simultaneous but also sequential manifestations of behaviors. Additionally, it focused on negotiation where the con¯icting parties want to reach an agreement in the end. And it used an objective standard for dyadic e€ectiveness, which is dicult to obtain outside the laboratory. It is especially dicult to think of an e€ectiveness criterion in organizational situations that is not confounded with other factors from the myriad of strategic, structural, and cultural variables. The laboratory approach adds extra support to the conclusion that the one-best-way image of problem solving and the contingency perspective of the situational appropriateness of problem solving need to be changed. People do react to con¯ict with simultaneously or sequentially assembled behaviors, which do have joint e€ects. Speci®cally, it has been shown that the simultaneity of problem solving and forcing, preceded by pure forcing, enhances substantive e€ectiveness. Unlike Study 1, in Study 2, no attention at all was paid to the relational outcomes of the conglomerated con¯ict behavior. Also, neither Study 1 nor Study 2 was directed at dyad members having a work relationship in daily life. To obviate both restrictions and to exclude a common source bias in the behavior and e€ectiveness ratings, which was a possible problem in Study 1, we planned a ®nal phase in this line of research. Study 3 Overview The ®nal investigation concerned work dyads from a wide variety of industries, consultancy ®rms, and governmental institutions. We designed a general and comprehensible con¯ict issue to be acted out by the dyads as part of a workshop on con¯ict management that we organized for them. Each simulation was videotaped and later rated by two independent groups of observers. The ®rst group assessed the occurrence of problem solving and forcing, while the second group assessed the e€ectiveness of the conglomerated con¯ict behavior. Method Participants Twenty-four Dutch dyads participated in the study, including 11 male±male, nine male±female, and four female±female dyads. The age of the participants ranged from 25 to 52 years (M ˆ 39 years), and 64 per cent of them held a managerial position. Their experience in the Copyright # 1999 John Wiley & Sons, Ltd. J. Organiz. Behav. 20, 475±491 (1999) CONSTRUCTIVE CONFLICT 485 present organization ranged from a few months to 30 years (M ˆ 7 years). All dyad members had either a superior±subordinate or a lateral co-working relationship in daily life. Con¯ict simulation Pilot studies (Nauta, 1996) resulted in the following role script: `In the space of 2 months, two persons, A and B, form a selection committee to hire a capable candidate for a new job. Because of illness during a vacation in a foreign country, A has been absent for 3 weeks. When A returns, it appears that B has continued screening candidates and there are only two left. Both A and B feel frustrated because things have not proceeded satisfactorily. B believes that A let B do all the work alone. A feels bypassed by B because the two remaining candidates do not meet the requirements that A ®nds most important. They have made an appointment to talk things over'. A 1-day workshop on con¯ict management for co-working dyads was advertized and organized. In separate rooms, the experimenters instructed one dyad member to prepare for role A, while the other dyad member was told to prepare for role B. The participants had 8 minutes to read the role script and to have ambiguities clari®ed, followed by 12 minutes to handle the con¯ict. Measurements As in Study 2, each con¯ict interaction was videotaped and later transcribed. A transcript contained 55 speaking turns on average. Four observers were trained to assess the occurrence of problem solving and forcing. Problem solving was operationalized through six prototypical examples of diagnostic explorations, integrative initiatives, creative suggestions, and breaking impasses. Similarly, forcing was operationalized through six prototypical examples of blaming, claiming, threatening, and imposing one's will on the opponent. The observers independently completed 3-point rating scales (not at all ˆ 1; a little ˆ 2; to a great extent ˆ 3). The ratings were done for each of A's and B's speaking turns separately. The interobserver reliability coecients were good for A's and B's problem solving (a ˆ 0.78) as well as forcing (a ˆ 0.90). The correspondence between A's and B's average scores across observers and speaking turns (problem solving r ˆ 0.35, p 5 0.05; forcing r ˆ 0.71, p 5 0.001) allowed us to treat the dyad as a whole, instead of di€erentiating between A and B. After combining the behavioral ratings 2 (`a little') and 3 (`to a great extent'), we used the same procedure as in Study 2 to create types of simultaneous combinations and sequences of problem solving and forcing. A completely independent group of four observers watched the videotapes once again to assess the substantive and relational outcomes with the help of seven 7-point rating scales anchored by appropriate operationalizations of very ine€ective (1) and very e€ective (7; see for details Study 1). The interrater agreement was satisfactory (a ranges from 0.76 to 0.87). Averaged across coders, the 7-item scale provided a reliable overall criterion of dyadic e€ectiveness (a ˆ 0.96). Validity Immediately after the role play, the participants were asked to rate the destructiveness versus constructiveness of their opponent's behavior. They used four 9-point rating scales (a ˆ 0.83 and 0.92 for As and Bs, respectively; for details, see De Dreu, Nauta and Van de Vliert, 1995, Study 3). Within dyads, A's and B's perceptions of the other's constructiveness (r ˆ 0.54; p 5 0.01) were summed. The resulting subjective constructiveness index was signi®cantly related Copyright # 1999 John Wiley & Sons, Ltd. J. Organiz. Behav. 20, 475±491 (1999) 486 E. VAN DE VLIERT ET AL. to the observers' ratings of pure forcing (r ˆ ÿ0.52; p 5 0.01), pure problem solving (r ˆ 0.51; p 5 0.01), and e€ectiveness (r ˆ 0.73; p 5 0.001), which attests to the validity of the observational measures of con¯ict behavior and e€ectiveness. Results Table 3 contains the descriptive statistics. Note that, once again, pure forcing is used much more than pure problem solving, concurrence, and sequences of problem solving and forcing. As in Studies 1 and 2, pure problem solving and pure forcing are negatively interrelated and have opposite positive and negative relations with e€ectiveness, respectively. In addition, concurrence of problem solving and forcing, problem solving followed by forcing, forcing followed by problem solving, and concurrence followed by problem solving, all have signi®cant positive relations with e€ectiveness. Regression analysis based on the z-scores of all the variables showed that pure forcing (DR2 ˆ 56 per cent, b ˆ ÿ0.75, F ˆ 28.14, p 5 0.001) and pure forcing followed by pure problem solving (DR2 ˆ 10 per cent, b ˆ 0.31, F ˆ 5.99, p 5 0.05) account for 66 per cent of the variance in e€ectiveness. These results imply a rejection of hypothesis 1 because, in and of itself, pure problem solving does not predict e€ectiveness. The two e€ects support hypotheses 2 and 3, respectively. The less dyad members used forcing behavior, and the more they used forcing behavior followed by problem solving, the more e€ectively they handled their con¯ict. Thus, in line with the results of Study 1, dyads prevented ine€ectiveness when they combined their forcing moves with problem solving moves. Once again, the ®ndings con®rm our presumption that a mixture of problem solving and forcing reduces the risks of forcing. As in Studies 1 and 2, additional analyses with A's or B's individual use of problem solving and forcing as predictors of dyadic e€ectiveness yielded the same pattern of results. Discussion As in Studies 1 and 2, the results of the third study support the conclusion that problem solving and forcing have interdependent e€ects on con¯ict outcomes. More importantly, however, Study 3 solidi®es, generalizes, and deepens our understanding of dyadic e€ectiveness by employing di€erent groups of observers for behaviors and behavioral e€ectiveness, by moving beyond the boundaries of particular types of organizations, and by making it clear that dyadic e€ectiveness depends not only on what combination of tactics is used but also on how the tactics are sequenced throughout the con¯ict interaction. It is important to note that the predictive power of forcing followed by problem solving rests on the observation of repeated behavioral patterns rather than once-only behavioral shifts. For that reason, the above conclusion leads to more than the recommendation that one should start competitive and end cooperative. An additional recommendation is that the sequence of `®rst forcing then problem solving' may have to be repeated to enhance e€ectiveness. Supplementary analyses of the total `wave' of forcing±problem solving±forcing±problem solving indicate that e€ectiveness is especially sensitive to ending with problem solving. The recommendation that one should alternate forcing and problem solving until a stage of problem solving produces satisfactory outcomes represents a crucial speci®cation of the corollary that employees are more e€ective to the extent that they use combinations of pure problem solving and pure forcing. Copyright # 1999 John Wiley & Sons, Ltd. J. Organiz. Behav. 20, 475±491 (1999) Copyright # 1999 John Wiley & Sons, Ltd. Table 3. Means, standard deviations, and intercorrelations of problem solving, forcing, and e€ectiveness: Study 3 S.D. F C PF FP PC FC CP CF E Simutaneous Pure problem solving (P) Pure forcing (F) Concurrence (C) 2.56 66.01 20.70 3.11 15.97 8.42 ÿ0.71* 0.41{ ÿ0.78* 0.89* ÿ0.64* 0.33 0.41{ ÿ0.05 0.04 0.39{ ÿ0.32 0.31 0.14 ÿ0.51* 0.71* 0.78* ÿ0.79* 0.46{ 0.04 ÿ0.35{ 0.59* 0.64* ÿ0.75* 0.51* Sequential P ®rst, F next (PF) F ®rst, P next (FP) P ®rst, C next (PC) F ®rst, C next (FC) C ®rst, P next (CP) C ®rst, F next (CF) 1.63 1.01 0.37 13.95 0.64 11.56 2.44 1.32 0.93 4.74 1.35 3.71 0.68* 0.00 0.34 0.20 0.05 ÿ0.00 ÿ0.01 0.69* 0.03 0.55* 0.35{ 0.33 0.23 0.56* 0.04 E€ectiveness (E) 4.81 1.30 0.19 0.08 0.53* 0.26 ÿ0.24 ÿ0.14 N ˆ 24 dyads; scales indicate percentages of speaking turns (P, F, C) or transitions of preceding speaking turns to following speaking turns (PF, FP, PC, FC, CP, CF). * p 5 0.01; { p 5 0.05. CONSTRUCTIVE CONFLICT 487 J. Organiz. Behav. 20, 475±491 (1999) M 488 E. VAN DE VLIERT ET AL. General Discussion The present series of studies contributes to the ®eld in several regards. First, at the level of the organizational dyad, the occurrence of problem solving and the occurrence of forcing were studied togetherÐas twinsÐrather than as separate entities. Breaking with the prevailing assumption of con¯ict behaviors as mutually isolated modes of reaction, problem solving and forcing were considered components of complex con¯ict behavior (cf. Falbe and Yukl, 1992; Knapp et al., 1988; Rubin et al., 1994; Yukl et al., 1993; Van de Vliert et al., 1995). Second, the e€ectiveness of problem solving and the ine€ectiveness of forcing were likewise studied together and not separately. Challenging the prevailing assumption of con¯ict behaviors as mutually isolated determinants of substantive and relational outcomes, it was demonstrated that problem solving and forcing are more e€ective, or less ine€ective, in combination than they are in isolation (cf. Falbe and Yukl, 1992; Putnam, 1990; Van de Vliert et al., 1995). Third, for the ®rst time, the simultaneous and the sequential occurrence of problem solving and forcing were studied in tandem. The second study, in particular, revealed that the simultaneity of problem solving and forcing, preceded by pure forcing, is an e€ective behavioral combination. The ®ndings provide additional and more speci®c empirical support for the crude behavioral recommendation that con¯ict at work pays (De Dreu and Van de Vliert, 1997; Johnson et al., 1989; Tjosvold, 1985, 1991; Van de Vliert, 1985). As outlined in our theoretical introduction, problem solving and forcing may supplement each other with respect to appropriateness for speci®c aspects of con¯ict as well as reduction of risks in terms of ine€ectiveness. For those reasons, a combination of problem solving and forcing might be best. The ®ndings in the last two studies provide a more re®ned operationalization of the rule of thumb that integrative and distributive moves should be mixed (cf. Putnam, 1990; Lax and Sebenius, 1986). In order to be e€ective, distributive controversy in the form of forcing should precede integrative exploration in the form of problem solving. This behavioral sequence resembles the well-documented e€ectiveness of the `reformed sinner strategy' (Deutsch, 1973; Harford and Solomon, 1967), and the within-person `black-hat/white-hat routine' (Hilty and Carnevale, 1993; Rafaeli and Sutton, 1991). In all such cases, a phase of di€erentiation precedes a phase of integration (Walton, 1987). On closer consideration, however, the reformed sinner strategy and the strategy of taking o€ one's black-hat and putting on one's white-hat address once-only behavioral shifts. In contrast, the ®ndings in Studies 2 and 3 are based on the alternation of pure forcing and either pure problem solving or problem solving combined with further forcing. Apparently, the move from forcing to problem solving has a more de-escalatory and e€ective nature than the reversed move from problem solving to forcing has an escalatory and ine€ective nature. Future research should be directed at determining why exactly de-escalatory shifts increase e€ectiveness more than escalatory shifts decrease e€ectiveness. Considerable progress would be made if the explanation could be ®ne-tuned to speci®c operationalizations of problem solving and forcing. A shortcoming of the present studies is the broad domain of coding instructions. Problem solving covers such divergent ways of behaving as exchanging accurate information, working through di€erences, dealing with each party's concerns, and suggesting creative breakthroughs. Similarly, forcing covers acts of blaming, claiming, threatening, using power to win, and imposing one's will. It is unlikely that each alternation of any problem solving element and any forcing element improves the e€ectiveness in the same way or to the same extent. Another weakness of our studies is that they relied predominantly on outsiders' assessments and a priori standards of con¯ict behavior and e€ectiveness, although the correspondence between objective Copyright # 1999 John Wiley & Sons, Ltd. J. Organiz. Behav. 20, 475±491 (1999) CONSTRUCTIVE CONFLICT 489 and subjective measures reported in the validation sections suggests that it is not a major ¯aw. It is relevant to determine whether the actors themselves experience the apparent e€ectiveness of pure forcing followed by at least some problem solving. Finally, con¯ict behavior that might be e€ective between Dutch employees may be less e€ective between employees in other countries, especially warmer countries which have more hierarchically di€erentiated and centralized organizations, and higher levels of extraversion and mutual competition (Van de Vliert and Van Yperen, 1996). Counterbalancing these limitations of the present series of studies are several notable strengths dealing with di€erences in samples, con¯ict situations, and operationalizations of behaviors and e€ectiveness. The robust pattern of results reported here also inspires con®dence in the following three tentative conclusions. First, simultaneous combinations of problem solving and forcing seem e€ective if sequential combinations are left out of consideration (Studies 1 and 2), but not if sequential combinations are also examined (Studies 2 and 3). Second, forcing is ine€ective unless it is followed by a combination of problem solving and forcing (Study 2), or pure problem solving (Study 3). Third, repetition of forcing followed by problem solving enhances organizational e€ectiveness most. References Aiken, L. S. and West, S. G. (1991). Multiple Regression: Testing and Interpreting Interactions, Sage, Thousand Oaks, CA. Andrews, I. R. and Tjosvold, D. (1983). `Con¯ict management under di€erent levels of intensity', Journal of Occupational Behaviour, 4, 223±228. Axelrod, R. (1984). The Evolution of Cooperation, Basic Books, New York. Blake, R. R. and Mouton, J. S. (1964). The Managerial Grid, Gulf, Houston, TX. Blake, R. R. and Mouton, J. S. (1970). `The ®fth achievement', Journal of Applied Behavioral Science, 6, 413±426. Blake, R. R. and Mouton, J. S. (1981). `Management by grid principles or situationalism: Which?', Group and Organization Studies, 6, 439±455. Boriso€, D. and Victor, D. A. (1989). Con¯ict Management: A Communication Skills Approach, Prentice Hall, Englewood Cli€s, NJ. Burke, R. J. (1970). `Methods of resolving superior±subordinate con¯ict: The constructive use of subordinate di€erences and disagreements', Organizational Behavior and Human Performance, 5, 393±411. Cleary, P. D. and Kessler, R. C. 1982. `The estimation and interpretation of modi®er e€ects'. Journal of Health and Social Behavior, 23, 159±169. Cohen, J. and Cohen, P. (1983). Applied Multiple Regression/Correlation Analysis for the Behavioral Sciences, Erlbaum, Hillsdale, NJ. De Dreu, C. K. W. and Van de Vliert, E. (Eds, 1997). Using Con¯ict in Organizations, Sage, Thousand Oaks, CA. De Dreu, C. K. W., Nauta, A. and Van de Vliert, E. (1995). `Self-serving evaluations of con¯ict behavior and escalation of the dispute', Journal of Applied Social Psychology, 25, 2049±2066. Deutsch, M. (1973). The Resolution of Con¯ict: Constructive and Destructive Processes, Yale University Press, New Haven, CT. Euwema, M. C. and Van de Vliert, E. (1996). `The e€ectiveness of conglomerations of problem solving and forcing by senior nurses'. Unpublished manuscript, University of Utrecht, The Netherlands. Falbe, C. M. and Yukl, G. (1992). `Consequences for managers of using single in¯uence tactics and combinations of tactics', Academy of Management Journal, 35, 638±652. Copyright # 1999 John Wiley & Sons, Ltd. J. Organiz. Behav. 20, 475±491 (1999) 490 E. VAN DE VLIERT ET AL. Fisher, R. and Ury, W. (1981). Getting to Yes: Negotiating Agreement Without Giving In, Hutchinson, London. Fisher, R., Ury, W. L. and Patton, B (1991). Getting to Yes: Negotiating Agreement Without Giving In, 2nd edn, Houghton Mi‚in, Boston. Harford, T. and Solomon, L. (1967). ` ``Reformed sinner'' and ``lapsed saint'' strategies in the prisoner's dilemma game', Journal of Con¯ict Resolution, 11, 104±109. Hilty, J. A. and Carnevale, P. J. (1993). `Black-hat/white-hat strategy in bilateral negotiation', Organizational Behavior and Human Decision Processes, 55, 444±469. Hocker, J. L. and Wilmot, W. W. (1991). Interpersonal Con¯ict, 3rd edn, William C. Brown, Dubuque, IA. Janssen, O. and Van de Vliert, E. (1996). `Concern for the other's goals: Key to (de-)escalation of con¯ict', International Journal of Con¯ict Management, 7, 99±120. Johnson, D. W., Johnson, R. T. and Smith, K. (1989). `Controversy within decision making situations'. In: Rahim, M. A. (Ed.) Managing Con¯ict: An Interdisciplinary Approach, Praeger, New York, pp. 251±264. Knapp, M. L., Putnam, L. L. and Davis, L. J. (1988). `Measuring interpersonal con¯ict in organizations: Where do we go from here?', Management Communication Quarterly, 1, 414±429. Lax, D. A. and Sebenius, J. K. (1986). The Manager as Negotiator: Bargaining for Cooperation and Competitive Gain, Free Press, New York. Likert, R. and Likert J. G. (1976). New Ways of Managing Con¯ict, McGraw-Hill, New York. Nauta, A. (1996). `Battle and bossiness in business: Interaction patterns in interpersonal con¯ict within bureaucratic and organic organization departments'. Unpublished dissertation (in Dutch), University of Groningen, The Netherlands. Pneuman, R. W. and Bruehl, M. E. (1982). Managing Con¯ict, Prentice-Hall, Englewood Cli€s, NJ. Pruitt, D. G. (1981). Negotiation Behavior, Academic Press, New York. Pruitt, D. G. and Carnevale, P. J. (1993). Negotiation in Social Con¯ict, Open University Press, Buckingham, London. Putnam, L. L. (1990). `Reframing integrative and distributive bargaining: A process perspective'. In: Sheppard, B. H., Bazerman, M. H. and Lewicki, R. J. (Eds) Research on Negotiation in Organizations, Vol. 2, JAI, Greenwich, CT, pp. 3±30. Rafaeli, A. and Sutton, R. I. (1991). `Emotional contrast strategies as means of social in¯uence: Lessons from criminal interrogators and bill collectors', Academy of Management Journal, 34, 749±775. Rahim, M. A. (1983). Rahim Organizational Con¯ict Inventories, Consulting Psychologists Press, Palo Alto, CA. Rahim, M. A. (1992). Managing Con¯ict in Organizations, 2nd edn, Praeger, Westport, CT. Rubin, J. Z., Pruitt, D. G. and Kim, S. H. (1994). Social Con¯ict: Escalation, Stalemate, and Settlement, McGraw-Hill, New York. Thomas, K. W. (1992). `Con¯ict and negotiation processes in organizations'. In: Dunnette, M. D. and Hough, L. M. (Eds) Handbook of Industrial and Organizational Psychology, 2nd edn, Consulting Psychologists Press, Palo Alto, CA, pp. 651±717. Tutzauer, F. and Rolo€, M. E. (1988). `Communication processes leading to integrative agreements: Three paths to joint bene®ts', Communication Research, 15, 360±380. Tjosvold, D. (1985). `Implications of controversy research for management', Journal of Management, 11, 221±238. Tjosvold, D. (1991). The Con¯ict-Positive Organization: Stimulate Diversity and Create Unity, AddisonWesley, Reading, MA. Van de Vliert, E. (1985). `Escalative intervention in small-group con¯icts', Journal of Applied Behavioral Science, 21, 19±36. Van de Vliert, E. (1997). Complex Interpersonal Con¯ict Behaviour: Theoretical Frontiers, Psychology Press, Hove, England. Van de Vliert, E. and Euwema, M. C. (1994). `Agreeableness and activeness as components of con¯ict behavior', Journal of Personality and Social Psychology, 66, 674±687. Van de Vliert, E., Euwema, M. C. and Huismans, S. E. (1995). `Managing con¯ict with a subordinate or a superior: E€ectiveness of conglomerated behavior', Journal of Applied Psychology, 80, 271±281. Van de Vliert, E. and Kabano€, B. (1990). `Toward theory-based measures of con¯ict management', Academy of Management Journal, 33, 199±209. Van de Vliert, E. and Van Yperen, N. W. (1996). `Why cross-national di€erences in role overload? Don't overlook ambient temperature!', Academy of Management Journal, 39, 986±1004. Copyright # 1999 John Wiley & Sons, Ltd. J. Organiz. Behav. 20, 475±491 (1999) CONSTRUCTIVE CONFLICT 491 Volkema, R. J. and Bergmann, T. J. (1989). `Interpersonal con¯ict at work: An analysis of behavioral responses', Human Relations, 42, 757±770. Walton, R. E. (1987). Managing Con¯ict: Interpersonal Dialogue and Third-Party Roles, 2nd edn, AddisonWesley, Reading, MA. Walton, R. E., Cutcher-Gershenfeld, J. E. and McKersie, R. B. (1994). Strategic Negotiations: A Theory of Change in Labor±Management Relations, Harvard Business School Press, Boston, MA. Yukl, G., Falbe, C. M. and Young Youn, J. (1993). `Patterns of in¯uence behavior for managers', Group & Organization Management, 18, 5±28. Copyright # 1999 John Wiley & Sons, Ltd. J. Organiz. Behav. 20, 475±491 (1999)