An Experimental Study on the Effects of Exitability and Correctability on Electronic Negotiation

Negotiation has been defined as a discussion between two or more parties seeking to reconcile incompatible goals (Pruitt and Carnevale 1993). Negotiation involves not only face‐to‐face communication but also communication using such media as the telephone and written documents; sometimes more than one medium is used within a single negotiation. Recently, as global computer networks have spread, the use of electronic media such as e‐mail or group decision support systems to facilitate negotiation has increased dramatically.

Several researchers have studied how the use of electronic media affects social interactions and joint decision making. For instance, Sara Kiesler, Jane Siegel, and Timothy W. McGuire (1984) found that negative emotions and rude behavior are more overtly expressed in computer‐mediated communication (CMC) than in face‐to‐face (FTF) interaction, which the authors attributed to the weakness of social cues expressed via CMC. Kiesler and her colleagues, as well as other researchers, have suggested that the electronic media may thus have a negative influence on negotiations.

Janet Solomon and Bharat Jain (1999), however, argued that the use of electronic media in negotiations can have positive effects. They argued that in electronic negotiation, participants are less likely to experience negative bias or rely on stereotypes activated by nonverbal cues such as physical appearance than they are in FTF negotiation.

Researchers have noted several specific ways in which CMC negotiations differ from FTF negotiations (e.g., Sproull and Kiesler 1991; Wheeler 1995):

• Electronic media are asynchronous, so that users are not bound by either time or space; they are able to send and receive messages at any time they like.

• Electronic media (CMC) are mainly text‐based communications that do not convey nontextual social cues such as tone of voice, facial expression, and body language.

• Users often experience a greater sense of anonymity, which we defined as a lack of personal and social information about the person with whom they are communicating, relative to FTF communication. When shopping on the Internet or using an Internet auction service, such as eBay, people deal with someone whom they do not know well and may never meet. At some point, that person's identity and name may be revealed, but initially such interactions often begin anonymously.

• Computer utility functions enable negotiators conducting electronic negotiations to review and correct their messages in advance of delivering them to others.

• The electronic media reduce a user's cognitive burden by electronically storing complete protocols; in other words the negotiator can reread what has happened so far in her negotiation to refresh her memory and clear up confusion.

• Finally, the electronic media are linked to world‐wide networks, whereby users are able to communicate with a virtually infinite number of people (Sproull and Kiesler 1991). Users are easily able to find a potential negotiator who shares their interests via the wide networks that CMC has created.

Among these characteristics of CMC, researchers have been especially interested in the relative anonymity it encourages, the limited access to the other person's identity that can inhibit the development of stable social relationships between electronic negotiators. Based on their observations of participants' responses in electronic negotiation, Don Moore et al. (1999) suggested that the relative anonymity associated with CMC leads to a decrease in trust between participants and increases the likelihood of an impasse in negotiation.

In an experimental study in which the medium and anonymity were independently manipulated, Terri Griffith and Gregory Northcraft (1994) found that the greater the anonymity available in an electronic negotiation, the more reluctant negotiators become to reveal their personal goals. These studies suggest that the anonymity of the electronic media has negative effects on negotiations.

But because negotiation is often accompanied by an exchange of tangible resources, maintaining anonymity can be difficult even within an electronic negotiation. In negotiations for merchandise trading or employment contracts, for instance, participants must disclose their names and addresses to each other at some stage of the negotiation. Although anonymity can be a salient characteristic of CMC, we do not think that it is the most important element to study in electronic negotiation.

Therefore, we focused on two other characteristics of CMC negotiations, asynchrony and functionality. We assumed that these characteristics contribute to the exitability and correctability of messages in electronic negotiation. The specific purpose of the present study was to examine the effects of exitability and correctability of messages on the processes and outcomes of electronic negotiations.

For this study we generated six hypotheses. With most electronic media, users convey their information in writing using text‐based word‐processing programs that allow them to revise messages in advance of sending them to others. Three of our hypotheses relate to this feature, which we refer to as correctability.

John Hayes and Linda Flower (1986) formulated three stages in the writing process (planning, sentence generation, and revision) and suggested that the act of writing itself helps the writer discover what to say and how to say it. In particular, the revision process helps the author elaborate not only his external verbal expressions but also his internal private thoughts (Murray 1978; Nystrand and Himley 1984).

Based on these ideas, we assumed CMC's correctability allows participants in negotiations to utilize both informational and social elaboration in developing their messages. Informational elaboration refers to one's ability to develop one's formal messages; that is, a negotiator can clarify and make his message more understandable via revision. We thus predicted that

A more articulate message may facilitate participants' mutual understanding of their concerns and may encourage them to interact constructively with each other to find good solutions. Therefore, our second hypothesis was that

Social elaboration refers to the participant's ability to elaborate on the contents of a message, that is, the degree to which a participant considers the other party's point of view when composing a message that includes an offer that is more acceptable to the other party. We further define social elaboration as the revision of the content of the message and informational elaboration as the linguistic revision of the message (wording, phrasing, etc.) In communication, generally a sender attempts to construct messages by anticipating how the recipient of the message will interpret it (Hayes and Flower 1986; Murray 1978). Yasuyuki Kawaura et al. (1989) suggested that the same cognitive process takes place in CMC. In a survey of Japanese attitudes on the use of electronic media, the researchers found that 30 percent of the respondents answered that they wrote messages by inferring the responses of the recipients. We assumed that participants are more likely to be engaged in such mental simulation in electronic negotiation than in FTF negotiation because writing evokes the revision processes more frequently than speaking does (Hayes and Flower 1986; Sommers 1980).

In negotiation, it is considered more acceptable to suppress naked expressions of self‐interest. Therefore, we predicted that

Further, we predicted that

In CMC, users can send and receive messages at any time they want (Sproull and Kiesler 1991). This frees them from many of the time and space constraints of FTF negotiation. It also increases one's opportunities to leave the negotiation, a phenomenon that we call exitability. But this ever‐present opportunity for exit seems to encourage the perception that electronic negotiation is inherently unstable because one can never be certain that the other party will reply to one's most recent message.

Exitability is a noteworthy feature of CMC. In one survey study (Kawaura et al. 1989), 40 percent of respondents claimed that during CMC they experienced anxiety about whether the people they were communicating with would read their messages or not. We hypothesized that their awareness of the ease with which the other party could exit the negotiation might put pressure on the participants in the negotiation to make quick agreements.

Though it has not been used in the CMC research, we regarded the ultimatum bargaining paradigm as a suitable vehicle for examining the effects of exitability. It is a one‐round negotiation between two people: one party makes an offer regarding reward allocation, while the other party decides whether to accept or reject it. If the latter accepts the offer, both parties receive the offered rewards, but if the latter rejects it, both get nothing. Ultimatum bargaining is highly exitable because negotiators are given only one chance to negotiate (“take it or leave it”). Werner Güth, Rolf Schmittberger, and Bernd Schwarze (1982) observed that in this situation, negotiators were more likely to make an acceptable offer to the other party.

In the present study, we observed negotiations under three sets of situations that varied according to their level of exitability. We predicted that

Because less demanding offers may be more acceptable to other participants, we predicted that

One hundred twenty Japanese undergraduate students (sixty men and sixty women) participated in the experiment. The same‐sex participants were randomly paired because we sought to eliminate any possibility that perceived gender‐related power differentials and gender stereotype or differences in negotiation style would be a factor in the negotiations (e.g., Walters, Stuhlmacher, and Meyer 1998; Stuhlmacher and Walters 1999). All pairs were previously unacquainted with their counterparts. After the experiment, the participants were given 500 yen each as a reward for their participation.

When each pair of participants arrived at the laboratory, the experimenter asked them to conduct a negotiation simulation with each other. To motivate them, the experimenter told participants that they should seek to get the highest possible personal scores and that the two participants who obtained the highest personal score in each situation would be given an additional 3,000 yen as a bonus. But we also told them that if a pair failed to reach agreement within a time limit, both participants would be awarded zero points.

The experimenter explained first that the task of participants was to negotiate the working conditions between a prospective employee and an employer; participants were randomly assigned to one of these roles. The experimenter took each participant into a separate booth, asked each one to negotiate with the other via computer, and explained how to use the chat software.

Each participant was asked to introduce him or herself to the other party briefly and then engage in a ten‐minute practice session. The practice session was immediately followed by the forty‐five‐minute test session. Participants were each given a warning at the forty‐minute mark. After the negotiation, participants were asked to complete a questionnaire.

The task used in this experiment was modified from paradigms developed by Dean Pruitt and Steven Lewis (1975) and Leigh Thompson and Reid Hastie (1990). Each participant was presented with a payoff schedule for the simulation, which listed the issues to be resolved and the points he or she would receive for achieving each of the possible alternatives for each issue. The four issues to be negotiated were wage per hour, transportation expenses, length of training period, and work hours. For each issue there were nine possible options available to the negotiators. (See the  Appendix.)

Wage per hour was a distributive, fixed‐sum issue; for example, gains for one party resulted in an equivalent loss for the other. Transportation expenses and the length of the training period, on the other hand, were issues that could be “log‐rolled” with the possibility that the parties could achieve joint gains via integrative negotiation techniques. Achieving his or her position on transportation expenses was more desirable for the employee, while achieving his or her goal on the length of training period was a more favorable outcome for the employer because of the way point values were awarded in the simulation. The parties' preferences regarding the issue of working hours were compatible and the point value assigned to each alternative was the same for both negotiators (i.e., a gain for one party represented an equal gain for the other).

The experiment was designed with two levels of correctability (correctable and uncorrectable) and three levels of exitability for a total of six possible simulation situations. Each pair was randomly assigned to one of the six situations. In the correctable situation, each participant was allowed to revise messages in advance of transmitting them, but in the uncorrectable versions, the participants were not allowed to do so. (We achieved this by specially programming the word processor to deliver the message immediately after the participant composed it.)

The three exitability situations were the unexitable, exitable, and one‐limited situations. In the unexitable situation, the experimenter instructed the participants that neither of the parties was allowed to exit the negotiation until they had an agreement or forty‐five minutes had passed. In the exitable situation, the experimenter told each participant separately that the other party was allowed to exit the negotiation at any time. In the one‐limited situation, both participants were allowed to negotiate only one time. First, the employee sent a message making offers on all four issues to the employer. If the employer replied that he or she accepted the offers, the negotiation concluded with an agreement. If the employer chose to reject those offers, he or she was able to send a message that included counteroffers to the employee. If the employee accepted these offers, the negotiation finished with an agreement at this point. If he or she rejected these counteroffers, however, the negotiation finished without an agreement. Because participants cannot expect to continue the negotiation after the first round of interaction, we consider exitability to be highest in the one‐limited situation. Clearly it was also higher in the exitable situation than in the unexitable situation.

The major dependent variables in this study were the level of demand of the participants' offers, the number of agreements reached, how integrative each agreement was, and the clarity of the messages. To measure level of demand, we used the points allocated in the payoff schedule to each initial offer. Our measure of level of agreement was the total number of pairs in each situation who agreed with all of the four issues within forty minutes — even if a pair reached agreement after the forty‐minute mark, we did not count it because we considered it to be forced by the five‐minute warning. We measured integrativeness by measuring the joint gain, that is, the total scores of all paired participants who reached agreement within forty minutes. In order to measure the clarity of the message, each participant was asked after the experiment to answer the question, “Were you able to make your points clear in advance of delivering the messages to the other party?” on a seven‐point scale ranging from “Not at all (1)” to “Definitely (7).”

Because four pairs did not negotiate on every issue, we eliminated the data from those eight participants and analyzed only the data from the remaining 112 participants.¹

As shown in Figure One, the participants in the one‐limited situation made significantly lower demands than in the other two situations but no significant difference was shown between the other situations.²

Employers made less demanding offers than the employees on the issue of transportation expense (mean score = 134 versus 572), but a higher demand than the employees on the training period issue (mean score = 487 versus 197).³

Furthermore, the wage per hour offers made by employers in the correctable situation were lower than their length of training period offers (mean score = 503 versus 389), but there was no difference statistically in offers among these issues in the uncorrectable situation (mean score = 470 versus 420).⁴ The payoff schedule was constructed so that the length of training period was a more important issue than wage per hour for employers. Employers in the correctable situation rightly put higher demands on the length of training period than on the wage per hour, while those in the uncorrectable situation did not.

We measured the degree of informational elaboration with the questionnaire, but we found no statistically significant difference in any of the comparisons among the means.

We found that the pairs reached the agreement within forty minutes more frequently in the correctable situation than in the uncorrectable situation (80 percent versus 53 percent) and more frequently in the one‐limited situation than in either the exitable or unexitable situations (90 percent versus 60 percent and 50 percent).⁵ There was no substantial difference between the exitable and unexitable situations.

To measure this variable, we eliminated the data from the twenty pairs who did not reach agreement within forty minutes, leaving forty pairs. The joint gain for integrativeness of the pairs in the one‐limited situation were lower than the either the exitable or unexitable situations (means of joint gain = 2,363, 2,598, and 2,676) but there was no significant difference between the latter two situations.⁶

To examine the effects of exitability on negotiation, we compared responses of the participants in three situations: the one‐limited situation where both parties were able to send a message one time only, the exitable situation where both parties were told that the other party had the right to leave the negotiation, and the unexitable situation where both parties were required to keep negotiating until the time limit had expired.

The results revealed that the participants in the one‐limited situation made less demanding offers and reached agreement more frequently than the participants in either the exitable or unexitable situations. Because the one‐limited situation was the most exitable, these results supported Hypotheses Five and Six. There were, however, fewer joint gains in the one‐limited situation than in the other situations.

The results in this experiment indicated that the exitability of the one‐limited situation exerted strong pressure on the participants to compromise with the other party, but the joint gain was low. This means that the high level of exitability had both positive and negative effects on the electronic negotiation; that is, the participants seemed to rush to agreement at the expense of quality. The reason may be that the pressure to agree forced the participants to choose simple strategies such as concession.

We assumed that the exitable situation would be the best simulation of real electronic negotiations because participants in real electronic negotiation are often afraid that the other party might exit. But we did not find a clear difference in the responses of participants between the exitable and unexitable situations. An interpretation is that the participants in the exitable situation did not think that the other party might use his/her right to exit because the other party obtained nothing by doing so. Another reason might have stemmed from the participants' status as students who may have believed that quitting would have been unacceptable or inappropriate.

In order to make the difference between the exitable and unexitable situations larger, then, we introduced the presence of a potential alternative negotiation partner as the exitable situation in Study Two (to be discussed later).

Among the hypotheses predicting the effects of correctability, our fourth hypothesis was supported because the number of pairs that reached agreement was significantly greater in the correctable situation than in the uncorrectable situation, suggesting that the parties who were given an opportunity to revise their messages were more likely to reach agreement than those who were not given this opportunity. On the other hand, our second hypothesis was not supported because the amount of joint gain did not differ significantly between these two situations.

Regarding the level of demand, there was no significant difference between the two correctability situations, a finding that was inconsistent with our third hypothesis. We found, however, that the participants who were assigned the role of employer in the correctable situation made high demands on the most important issue (length of training period). The payoff schedule given to participants indicates that achieving their desired length of training period was favorable for employers, while achieving their transportation expense goal was favorable for employees, suggesting that these issues were log‐rolled tasks. This suggests that the opportunity to correct messages allows participants to prioritize issues in the making of offers. On the other hand, correctability did not affect informational elaboration, which was inconsistent with our hypothesis that correctability would contribute to clarity.

Although correctability facilitated agreement, the mechanism with which it did so did not become clear to us. For this reason, we analyzed the content of the messages using coding systems based on the work of Ellen Giebels, Carsten K. W. De Dreu, and Evert Van de Vliert (2000) and Ken‐ichi Ohbuchi, Sachiko Chiba, and Osamu Fukushima (1996). Three raters independently placed the participants' messages into the following fourteen categories: rejection, assertion, demand, threat, explanation, entreaty, offer, persuasion, question, concession, consideration, acceptance, trade‐off, and sounding.

For each category we averaged the scores (1 or 0) given by the three raters and then converted the averaged scores into the relative scores. According to the definition of social elaboration, we regarded two of the categories, acceptance and trade‐off, as fitting within our definition of social elaboration techniques because by using these strategies, participants clearly indicated that they had received their counterpart's message. We tested the difference in the frequencies between the correctable and uncorrectable situations.⁷ We found that the parties in the correctable situation more frequently developed messages that included trade‐offs than did those in the uncorrectable situation (mean frequency = 0.13 versus 0.06). Correctability did not, however, seem to have a statistically significant effect on the use of acceptance as a technique in these negotiations. These results suggest that the opportunity to correct messages prompts agreements by allowing the participants to make more trade‐off offers.

We found a significant relationship between issue and role when it came to how demanding offers were, which we interpret as a result of our scoring system. The scores allocated to the transportation expense and training period issues were reversed depending on role. Thus, for the employee, the highest score was awarded for achieving a favorable transportation expense outcome while success in negotiating a desirable training period was much less richly rewarded; for the employer the opposite scoring system was in effect.

In the first study, we manipulated exitability in two ways: by limiting the number of offers that could be made and by giving both parties the power to exit the negotiation. We found that limiting the number of possible offers (the one‐limited situations) encouraged agreements, apparently because it caused negotiators to make relatively less demanding offers. But we also found that giving parties the power to exit the negotiation did not affect the negotiation; we believe this resulted from the participants' calculations that the other party would not exit because there was no explicit advantage to be gained by doing so. In our second study, therefore, we sought to reexamine the effect of exitability by refining our procedure to address this expectation.

In many types of real (not simulated) electronic negotiations, participants often have potential alternative partners because the Internet increases the opportunity to find others who share this interest. This may make it easier for negotiators to exchange partners in CMC negotiations than in FTF negotiations. We assumed that the availability of potential alternative partners is a crucial factor of exitability that makes electronic negotiation more unstable. Several factors could prompt exitability, such as time pressure, but we believe that the availability of alternative partners may be the most important one.

Thus, in our second study, we manipulated the exitable and the one‐limited situations by telling participants that “the other party had alternative partners to negotiate with.” In the one‐limited situation, we told each participant that both parties were allowed to negotiate only once because the other party had alternative negotiation partners. In the exitable situation, the participants were told that only the other party was allowed to exit the current negotiation if he or she chose in order to negotiate with alternative partners. In the unexitable situation, the participants were told that neither party was allowed to exit the current negotiation until they reached agreement or the time had expired.

For this study, we generated six hypotheses. We predicted that

Because the less demanding offers could be expected to be more acceptable to the other participants, we predicted that

In the second study, we manipulated correctability in the same way as in the first study. Thus we hypothesized again that

We also predicted, again, that

Further, we also hypothesized that

Because each participant would be better able to comprehend the other party's message and, consequently, better understand his or her underlying interests, we hypothesized that

One hundred twenty Japanese undergraduate students (sixty men and sixty women) participated in this experiment. (For Study Two we recruited a completely new set of students.) Groups of twenty were instructed to negotiate via computer with other university students. But, in fact, they were randomly paired with a same‐sex participant in the same room. None of the pairs were previously acquainted. After the experiment, the participants were given 500 yen each for their participation in the experiment.

When the twenty participants arrived at the laboratory, the experimenter explained the experimental procedure to them on a computer. Unlike Study One, they did not introduce themselves or practice the electronic negotiation. The exitability was manipulated by referring to the availability of alternative partners for the other party. The one‐limited situation was of the highest exitability; that is, both the participants were allowed to make only one offer because each party was informed that the other one had alternative partners to negotiate with. In the exitable situation, the experimenter told each participant separately that the other participant was allowed to exit the negotiation at any time because he or she had alternative partners. In the unexitable situation, the experimenter instructed the participants that neither party was allowed to exit from the negotiation until they either reached agreement or exceeded the time limit.

In both the exitable and unexitable situations, the negotiation ended when an agreement was reached on all four issues. To measure the impact of correctability, we constructed two situations, just as in Study One. Thus, the experiment was designed with two levels of correctability and three levels of exitability. Each pair was randomly assigned to one of the six situations.

The dependent variables in the second study were the same as in the first, except for the measure of informational elaboration. Because we determined that the negotiators' self‐rating of informational elaboration could have been influenced by the outcomes, in Study Two we used the raters' coding of verbal protocols as well as self‐rating. Three raters independently scored the verbal clarity of each message on a three‐point scale ranging from two (most elaborated) to zero (least elaborated). The most elaborated responses were the messages that involved specific offers. Somewhat elaborated responses (score of one) were the messages that did not involve specific offers and needed confirmation by the other party. The least elaborated responses (score of zero) were the messages that had inconsistent content or merely repeated a previous offer. For each participant, the total score of all the messages was divided by the number of messages, and then it was also averaged among the three raters. We eliminated the scores in the one‐limited situation from analysis because the participants could respond only once or not at all in the one‐limited situation, offering fewer opportunities to revise their messages.

After eliminating two participants who misunderstood the instruction and four pairs who did not negotiate on each issue, we analyzed the data from the remaining 110 participants.

As shown in Figure Two, the participants in the one‐limited and exitable situations made significantly lower demands than those in the unexitable situations, but this time there was no significant difference between the one‐limited and exitable situations.⁸ (Recall that in Study One the difference in level of demand between the one‐limited situation and the other two was significant, but there was no significant difference between the exitable and unexitable situations.)

Furthermore, the employers made a lower demand on transportation expenses than did the employees (mean score = 127 versus 579) but a higher demand on the training period than the employees (mean score = 503 versus 191).⁹ This pattern of responses was rational in terms of the payoff schedule.

Twenty‐seven pairs reached agreement within forty minutes. We analyzed the frequency data across correctability and exitability. As shown in Figure Three, the pairs in the one‐limited situation more frequently reached agreement than in the other situations, and the pairs in the exitable situation more frequently reached agreement than those in the unexitable situation.¹⁰

After eliminating two pairs who misunderstood the instructions and thirty‐one pairs who did not reach agreement within forty minutes, we analyzed the data from the remaining twenty‐seven pairs.¹¹ We found that the joint gain in the one‐limited situation (mean score = 2478) was significantly lower than in the exitable situation (mean score = 2696)¹² and unexitable situation (mean score = 2755).¹³

Informational elaboration was measured using both the questionnaire and coding of messages. After we eliminated the two participants who misunderstood the instructions and one participant who responded incompletely, data from the remaining 117 participants were analyzed. We found that participants in the correctable situation rated their messages as being clearer than in the uncorrectable situation.¹⁴

We also found the score was higher in the exitable situation than in the unexitable situation (mean score = 1.67 versus 1.45).¹⁵ Furthermore, in the correctable and exitable situation, participants' messages were clearer than were those of participants in the correctable but unexitable situation.¹⁶

In Study Two, we manipulated exitability in the simulated negotiation in a different way by introducing the idea that the other party had the opportunity to negotiate with an alternative partner. Our second hypothesis, that the availability of alternative parties would make participants more likely to reach agreements, was supported — the more exitable the negotiation was, the more frequently the parties reached agreements. Specifically, we found that the participants in the exitable situation reached agreement less frequently than those in the one‐limited situation, but the joint gain in the exitable situation was higher than in the one‐limited situation. These results suggest that the uncertainty of the exitable situation prompted agreement without reducing the quality of the agreement.

How was integrative settlement encouraged in Study Two? Consistent with our second hypothesis, the participants in the one‐limited and exitable situations made less demanding offers. This seems to reflect their fear of unstable negotiation, that is, the fear that the other party might exit from the negotiation if the offer were not attractive enough. The participants in the exitable situations made more demanding offers than those in the one‐limited situation on the issue of working time (mean score = 399 versus 364). This suggested that the negotiators in the exitable situation took more active action to attain a compatible result (working time).

Repeating a more demanding offer in the exitable situations seems likely to increase the risk of impasse, which is why we believe the participants in the one‐limited situation did not attempt it. In the exitable situation, however, the negotiators used more complicated strategies, such as combining assertive and conciliatory approaches. It seems that they believed they still had a chance to persuade the other party even under the pressure of exitability.

An essential feature of electronic negotiation is instability, and one kind of instability is generated by the potential availability of alternative partners. We simulated this feature in the laboratory as a component of the exitable situations. The finding that the participants in the exitable situations were more likely to reach integrative settlements suggests a potential strength of electronic negotiation.

We attempted to examine the effects of correctability in terms of informational elaboration, which we measured by a postexperimental questionnaire. The result was consistent with our fifth hypothesis that participants who are given an opportunity to revise messages in advance would produce more articulate messages than those who were not given this opportunity. The participants who were given an opportunity to revise their messages judged that they produced more concise messages than those who were not given the same opportunity. This means that the correctability that electronic negotiation offers can reduce redundancy and rambling in communication between negotiators.

In addition, the rated clarity of messages correlated with the joint gain in the correctable situation.¹⁷ These findings are partially consistent with our sixth hypothesis, that electronic negotiation's correctability supports more productive communication.

Those effects, however, were restricted by another variable, which was exitability. The analysis of coded informational elaboration indicates that correctability prompted participants to clarify their messages in the unstable situation where the other party could exit from the current negotiation. Furthermore, no pair reached agreement in the correctable and unexitable situation, suggesting that the participants in this situation might have used up their time perfecting their messages rather than coming to agreement because they were not motivated to respond quickly.

As we noted earlier, the interaction between issue and role was significant, but this is most likely an artificial effect produced by the payoff structure used in the experiment.

The present study experimentally examined the influence of two important characteristics of CMC on negotiation: correctability and exitability. A distinguishing feature of many forms of electronic communication — e‐mail in particular — is that they are asynchronous. In electronic negotiation, the participant is likely to feel uncertainty as to whether the other party will respond or not because the users are not sharing a space at the same time, which makes the process seem inherently unstable.

In the two studies reported in this article, we examined the effects of exitability on negotiation behavior and outcome by comparing the one‐limited, exitable, and unexitable situations. In Study One, we found that the one‐limited situation, in which exitability was highest, prompted agreements but reduced their quality. We believe the exitable situation was the most realistic, but the behavior of participants in this situation did not seem to be affected by exitability. We interpreted that the participants in the exitable situation might have thought that the other party would not exit because, in this exercise, exiting from negotiations did not pay.

Using electronic media expands one's field of potential negotiating partners (e.g., job openings that are posted online can expect to receive a greater number of applications). For many kinds of transactions, participants can change negotiation partners with little cost. This also creates instability in electronic negotiation. In Study Two, where the possibility that the other party could seek out an alternative negotiation partner was introduced, we found that the participants who negotiated in the exitable situation made less demanding offers and were more likely to reach agreement than in the unexitable situation.

The results of both studies support our hypothesis that the possible existence of an alternative partner increases exitability. Research on nonelectronic negotiations has also found the same effect. Robin Pinkley and her colleagues (Pinkley 1995; Pinkley, Neale, and Bennett 1994) demonstrated that participants who do not have a good best alternative to a negotiated agreement (BATNA) (Fisher and Ury 1981) tend to make more attractive offers than those with a good BATNA, and they further proposed that those who do not have an attractive BATNA are more strongly committed to the current negotiation. But while exitability can prompt agreement, if it is achieved by mere concessions, that agreement may not be regarded as a constructive resolution (Pruitt and Carnevale 1993).

Our results indicate, however, that exitability prompted more integrative agreements. Leigh Thompson and Reid Hastie (1990) argued that an integrative agreement is most likely to be reached by finding compatibility on issues of value. In Study Two, the participants in the exitable situation made more demanding offers on compatible issues than did those participants in the one‐limited situation. The participants in the exitable situation made concessions only on issues that were not important to them, but not on those issues that mattered more to them; that is, they did not make an overall concession. This provides one explanation for how they could reach agreement even in this unstable environment without losing joint gain.

In contrast, in the situation where both negotiators were allowed to make only one offer (the one‐limited situation), their offers were not very demanding, and there was minimal joint gain. It seems that the extremely high exitability prompted the participants to select a strategy of compromise on all issues, which ultimately sacrificed the quality of agreement. We must note, however, that the one‐limited situation is an artifice; “real” negotiations are unlikely to include such a limitation. Therefore, we conclude, despite the result from the one‐limited situations, that exitability in electronic negotiation has the potential to lead negotiators to reach constructive agreements, though this depends on the particular structure of the negotiation and its issues.

The present study provides some practical implications for electronic negotiation. Electronic negotiation, we believe, can be accomplished effectively when negotiators are strongly motivated to reach consensus or when the issues are relatively simple so that the main task is to make concessions. Even when the issues are complex, electronic negotiation may produce integrative outcomes if the structure of the issues enables participants to make selective concessions.

We also focused on correctability, which is another important characteristic of most electronic negotiations. Specifically, we attempted to examine the effect of correctability on social and informational elaboration of messages. Electronic media enable a continuous flow of communication between users that is more difficult to achieve via other text‐based media such as letters (Carnevale and Probst 1997; Hiltz 1982). In addition, word‐processing functions assist the generation, constitution, and expression of messages (Wheeler 1995).

Consistent with these observations, we found through these experiments that the ability to revise their messages prior to transmitting them helped participants to craft more socially acceptable and linguistically clearer messages in negotiation. The results of Study One indicated that correctability of the electronic media produced messages including trade‐offs and more frequently resulted in agreements.

As Lawrence Frase (1987) and Hayes and Flower (1986) both observed, a revision stage in the writing process is necessary for producing creative or logical ideas. Correctability prompts specific feedback by inducing participants to consider and revise their own message to take into account the other's point of view (perspective taking). Furthermore, we suggest that the opportunity to correct one's message prompts an individual to consider the perspective of the person to whom he or she is transmitting the message. Matthew Traxler and Morton Ann Gernsbacher (1992) observed that good writers develop sentences by anticipating the reader's perspective. Dean Pruitt and Peter Carnevale (1993) emphasized that the same cognitive processes may be at work in the management of constructive negotiations. Consistent with these suggestions, we observed in Study One that participants in the correctable situation made bilateral offers considering both interests and refraining from egocentric offers and were thus were more likely to reach agreement. However, in Study Two, in which each participant negotiated with the other party but had no experience of direct contact, we could not find evidence of social elaboration, which suggests that in practice, a negotiator would be able to use electronic media more effectively after interacting with the other negotiator face‐to‐face in the first stages of negotiation.

We believe that this aspect of the experiment might have unfortunately decreased the participants' commitment to the negotiation. Considering previous findings that commitment to relationships inclines people toward prosocial or cooperative behaviors (Arriaga and Rusbult 1998; Rusbult et al. 1991), we must deal with its effects on electronic negotiation in future research.

Informational elaboration involves revising messages to clarify them. Especially in the unstable negotiation situations, correctability was indeed found to contribute to the elaboration of messages. In negotiations, the unclear message hinders mutual understanding between the parties and may thereby intensify oppositional tendencies.

The results of Study Two also indicate that, by prompting informational elaboration of messages, correctability may contribute to integrative agreement. Clear messages may help participants to better understand their mutual interests. We assume that informational elaboration contributes constructively to negotiations by means of two mechanisms: by reducing cognitive biases, such as the fixed‐pie perception, and by prompting a better mutual understanding of the issues with which each participant is concerned. In Study One, however, in which each participant negotiated with a party in the next booth, we could find no evidence of informational elaboration. The participants might have felt pressured to make prompt reactions because they were aware that the other party was just on the other side of the door. We interpreted this as a situation where the pressure to respond might have restricted the participants' cognitive effort in informational elaboration of messages.

In Study Two, we found no significant effect of correctability on agreement. No pair reached agreement in the correctable and unexitable situation. In this situation, we believe the participants might have had enough time to correct messages but were not motivated to respond quickly because they perceived that the other party could not exit from this negotiation. It suggests that the effect of correctability can be conditioned by exitability in electronic negotiation.

The present study found that exitability created by the possibility of an alternative negotiator and the chance to correct messages both prompted efficient negotiation processes and more productive agreement outcomes in electronic negotiations. Undertaking future research in electronic negotiation to explore additional factors that regulate exitability as well as to further examine the effect of correctability would be worthwhile.