Following Cummings and Bromiley (1996), whose definition of interpersonal trust is most often applied with regard to dispersed teams (Jarvenpaa and Leidner, 1999; Jarvenpaa et al., 2004), we maintain that team members trust their team if they believe that all the team members make an effort to behave in accordance with their commitments, that all the team members are benevolent, and that no team member will take advantage of another, even when given the opportunity to do so.

Empirical studies have shown that interpersonal trust enhances communication and information sharing (Dirks and Ferrin, 2001), particularly in dispersed teams (Jarvenpaa and Leidner, 1999). Knowledge transfer enabled by information sharing allows external and internal knowledge to be combined (Madhavan and Grover, 1998), thus serving as the basis for the creation of new knowledge (Knudsen, 2007). In turn, new knowledge is a major driver of new product development (Brockhoff, 1998).

Since trust leads to the constructive interpretation of partner motives (Uzzi, 1997), mitigating uncertainty about partner behavior (Krishnan et al., 2006), it reduces the potential for relationship or affective conflicts (Zaheer et al., 1998), which is generally regarded as dysfunctional in respect of dispersed teams (Kankanhalli et al., 2006) and also in NPD teams (Dreu and Vianen, 2001). However, by fostering task-related problem-solving effectiveness (Zand, 1972), trust facilitates moderate levels of cognitive conflict, which are assumed to enhance NPD team effectiveness (De Dreu, 2006).

In sum, and consistent with prior work on dispersed (Iacono and Weisband, 1997; Jarvenpaa and Leidner, 1999; Coppola et al., 2004) and NPD teams (Bstieler, 2006; Bstieler and Hemmert, 2008; Bierly III et al., 2009), we surmise that trust positively influences effectiveness in dispersed NPD teams, i.e., the extent to which a team is able to achieve quality goals (Hoegl et al., 2004). Hence, we posit:
Building on trust's general favorability (Atuahene-Gima and Li, 2002), we argue that interpersonal trust's generally positive performance effect may be strengthened under certain conditions and diminished under other conditions. Since collaboration on innovative endeavors is becoming increasingly global (McDonough III et al., 2001; Gassmann and von Zedtwitz, 2003), we focus on global team characteristics as possible moderators and delineate the relevant conditions when interpersonal trust affects NPD teams' effectiveness more strongly.

The central characteristic of global NPD teams is geographic dispersion. Geographic dispersion relates to the team members' physical and temporal distance (Baba et al., 2004), which can vary between the same place and global distribution (Lipnack and Stamps, 2000). Although physical dispersion does not necessarily influence the use of communication technologies – highly dispersed teams can travel and fairly colocated teams mostly communicate via e-mail or telephone (Fiol and O'Connor, 2005) – it is often assumed that more geographically distributed teams are also more electronically dependent (Kirkman et al., 2004). Computer-mediated communication can provide access to relevant expertise beyond the geographical boundaries, thus leading to cultural diversity among team members (Maznevski and Chudoba, 2000). However, team members from different cultural backgrounds could also work together in face-to-face teams. Therefore, national diversity and geographic dispersion often coincide but are not necessarily related (O'Leary and Cummings, 2007). Moreover, the convenient access to expertise that computer-mediated communication offers, supports creative and flexible responses to changing demands by facilitating flexible team membership. The latter enables the team to staff flexibly according to changes in their work assignments (Langfred, 2007). Thus, we argue that geographic dispersion, computer-mediated communication, national diversity, and team membership flexibility are central characteristics of global NPD teams. All these characteristics are considered continuous variables rather than definitional elements. Moreover, although these characteristics are all pertinent to global NPD teams, they are likely to demonstrate unique effects and, therefore, are argued to be independent of one another (Gibson and Gibbs, 2006).

NPD teams become increasingly geographically dispersed, because companies strive to leverage specialized technical and local market knowledge from different sites (McDonough III et al., 2001) to increase innovation effectiveness (Brockhoff, 1998).

We argue that interpersonal trust's positive effects are particularly strong when geographic dispersion is high. With team members relying on one another to fulfill their work obligations (Zaheer et al., 1998), high levels of interpersonal trust lead to a decreased need for monitoring (Langfred, 2004). The positive effects of trust are particularly beneficial for team effectiveness under geographic dispersion, as face-to-face interactions are quite rare (Carson et al., 2003), making the monitoring of task activities and processes more difficult (Gibson and Gibbs, 2006). Moreover, interpersonal trust leads to the acceptance of uncertainty (Lane and Bachmann, 1996). Since global NPD teams are challenged by a decreasing awareness of one another's work status (Dennis, 1996; Tan et al., 1998), the benefits from trusting the other members are likely to increase if dispersion is high. Finally, interpersonal trust is characterized by faith in the partner's benevolence and intentions (Mayer et al., 1995), making each team member more likely to renounce control mechanisms (Fryxell et al., 2002). Thus, the team members assume that each of them is engaged in task accomplishment even if there is no control to verify this (Das and Teng, 1998). This, in turn, leads to interpersonal trust's increasing relevance for dispersed NPD team effectiveness.

These arguments emphasize the critical importance of interpersonal trust under greater geographically dispersion, and point to the latter's positive moderating effect on the relationship between interpersonal trust and effectiveness. Hence, we posit:

Developments in computer-mediated communication systems have made it feasible for NPD teams to work together despite their physical dispersion (Zack and McKenney, 1995). These systems allow digitized information to be entered, stored, processed, distributed, and received (Kahai and Cooper, 1999), as well as allowing computer conferencing systems, electronic and voice mail systems, group decision support systems, and text retrieval systems (Huseman and Miles, 1988). Computer-mediated communication is a continuum and is a question of the relative extent of computer-mediated rather than face-to-face communication (Griffith et al., 2003). A team that operates entirely through e-mail, text exchanges, and teleconferences, never meeting face-to-face, is more electronically dependent than a team whose participants have a monthly face-to-face meeting. (Gibson and Gibbs, 2006)

We argue that in NPD teams with high levels of computer-mediated communication, the relationship between interpersonal trust and team effectiveness is particularly strong. Trust leads team members to consider sources of information and the other team members' motives positively (Mayer et al., 1995). As the integration of the team members' individually unique knowledge is crucial to NPD team effectiveness (Madhavan and Grover, 1998), smooth knowledge exchange processes between the team members are very important in this context. Assuming that the other team members have positive motives, interpersonal trust also diminishes dysfunctional conflict (Jehn and Mannix, 2001). The limited media richness and social presence in electronically dependent teams lead to fewer nonverbal cues (Kahai and Cooper, 1999), less contextual information (Kankanhalli et al., 2006), and, thus, to misunderstandings among team members (Desanctis and Monge, 1999; Mortensen and Hinds, 2001). Moreover, technical expertise is often bound to a specific language, which might differ among the NPD team members (Cramton, 2001), allowing misunderstandings to arise. The positive consideration of other team members' motives thus seems to be particularly essential in settings characterized by high levels of computer-mediated communication.

Finally, trust is reflected in team members' willingness to listen to and absorb others' knowledge without verifying this information (Szulanski et al., 2004). Under conditions of strong computer-mediated communication, this willingness is particularly important because computer-mediated communication largely prevents information verification (Johnson and Lederer, 2005). Thus, the importance of trust increases. Moreover, trust implies a willingness to share one's useful knowledge (Levin et al., 2006), implies that the sources of information are considered trustworthy (Andrews and Delahaye, 2000), and that the amount of information exchanged is likely to increase. Since NPD teams continually depend on the integration of their members' unique knowledge (Espinosa et al., 2007), a willingness to rely on the information given becomes crucial to knowledge-sharing effectiveness. Information verification would increase knowledge exchanges' time and cost.

The positive effects of interpersonal trust are therefore crucial in the presence of strong computer-mediated communication, which is generally characterized by lower levels of information-sharing effectiveness than found in face-to-face teams (Kirkman et al., 2006). Therefore, we propose that if team members' computer-mediated communication increases, interpersonal trust has a stronger impact on team effectiveness:

Dynamic changes in the task, situation, and conditions are a key characteristic of new product development and demand flexibility in the team structure (Langfred, 2007). As such, NPD teams are often characterized by frequent changes in team membership (Gibson and Gibbs, 2006).

Trust's positive relationship with effectiveness is likely to be particularly strong in teams with high levels of team membership changes. Interpersonal trust leads team members to listen to problems, to pass along relevant information (Williams and Anderson, 1991), and explain the work (Van Dyne and LePine, 1998). This positive effect is likely to become even more important in the context of frequent changes in team membership, as team members tend to lack a common vocabulary and their understanding of team processes is limited (Wiersema and Bantel, 1992). These team members are thus faced with uncertainty and anxiety about socially accepted behavior (Harrison et al., 2003). Hence, trust's positive effects tend to increase in teams with flexible membership.

Moreover, interpersonal trust is associated with the willingness to allow deviances from socially accepted behavior (Lewicki and Bunker, 1996). In the belief that team members act with the best intentions and do not deliberately disregard social norms, but simply do not recognize them, deviances are not attributed to actors' lack of goodwill. These benefits become even more important under conditions of flexible membership, as, besides the problem of new members not being acquainted with team norms, shorter tenure also leads to less commitment to established norms and practices (Finkelstein and Hambrick, 1990). Thus, new team members might not feel obliged to adhere to the team norms and could provoke dysfunctional conflict. Therefore, interpersonal trust becomes increasingly relevant for NPD team effectiveness in the context of non-stable team membership. Hence, we posit:

Firms increasingly take advantage of international opportunities to maximize the use of scarce knowledge resources (Cramton, 2001) and to capture comparative cost advantages. Consequently, product development activities' internationalization has shown strong growth (Moenaert et al., 2000; Björn and Bodo, 2008). The more nationally diverse an NPD team is, the more it faces cultural and linguistic differences between its members, as well as differences along broader cultural dimensions, such as individualism and collectivism (Kankanhalli et al., 2006). Culture is a set of shared symbols, norms, and values in a social collectivity (Dongsong et al., 2007), thus serving as a filter for one's perception of the surrounding environment, as guiding behavior, and as social interaction (Chudoba et al., 2005). In their conceptual piece, Nakata and Sivakumar (1996) argue that culture influences NPD team effectiveness. As such, the authors assume, for example, that individualism increases NPD team effectiveness, while power distance is argued to demonstrate negative effects.

In settings with high levels of national diversity, the importance of trust is likely to increase. Trust mitigates uncertainty about partner behavior (Krishnan et al., 2006) by facilitating mutual understanding and giving one another the benefit of the doubt (Zaheer et al., 1998). If one party encounters its partner undertaking unexpected actions, which could be ascribed to both good and bad intentions, a constructive interpretation of the partner's intentions is critical, as the first party is assumed to exhibit goodwill (Krishnan et al., 2006). Since teams with high national diversity demonstrate divergent preferences for social interaction norms (Zhang et al., 2007), this positive effect of interpersonal trust becomes particularly important. Consequently, trust's positive effects may increase in teams with high levels of national diversity. Moreover, trust fosters the predictability of the other party's behavior (Zaheer et al., 1998) by alleviating communication deficiencies. This is particularly important in highly international teams with frequent misunderstandings.

Finally, interpersonal trust increases voluntary deference (Kramer, 1999). A trusting recipient is more likely to accept advice from the other team members and to change his or her behavior. Hence, trust is particularly relevant in nationally diverse teams to overcome the downsides of social categorization and its negative consequences for knowledge exchange. In this context, trust's positive effects are particularly important, as individuals tend to categorize others according to physical cues, such as race, and to develop expectations about one another's behavior (Hogg and Terry, 2000). These cues and expectations then shape their interactions; i.e., the opportunities they assign one another to speak, and the appreciation of one another's contributions (Weisband et al., 1995). Therefore, we propose that with national diversity in NPD teams increasing, interpersonal trust has a stronger impact on team effectiveness because it strengthens the team's ability to cooperate. Accordingly, we posit:

We test our hypotheses on a sample of 80 software development teams from five software development companies' 28 different worldwide labs. Four of the companies are headquartered in Germany and one in the United States. The companies participating in our study organize their R&D in various software laboratories dedicated to the development, testing, and adaptation of basic software programs. While most software labs are proficient at the entire software development value chain, some labs are specialized in certain functionalities or development processes. Each laboratory participating in this study produced a list of projects, which also included the team leaders' names and contact information.

In a second step, the team leaders were provided with additional information about the present study and asked to complete a spreadsheet by providing descriptive details of the team and its team members (e.g., their gender, age, nationality, and the location of each team member's office). In a third step, all the team leaders and team-external managers, as well as randomly chosen team members, were contacted and asked to complete the online questionnaire. The respondents' participation in this study was strictly voluntary. A team was only included in the final data set if the team leader, the team-external manager (i.e., the manager to whom the team leader reported), and at least two team members had completed the questionnaire. The team leaders and managers' response rate was 100%, while that of the team members was 68%.

Each laboratory employed between 20 and 5,500 software developers (median = 600). All software development projects completed 12 months prior to our study's data collection and on which teams of nine or fewer members worked, were included in the study. This selection criterion ensured that the study focused on small integrated work teams, rather than on larger collectives that are often structured into multi-team projects (Hoegl et al., 2004), which often occurs in respect of software development programs (Dubé et al., 2006). Overall, our analyses are based on a total of 392 responses from 80 team leaders, 80 managers, and 232 team members. This sample contains 28% female team leaders and 24% female team members. Teams in our sample have an average of 5.9 members (median = 6, standard deviation = 1.8, min = 3, and max = 9), while the average age of the team members was 37.0 years (median = 37.0, standard deviation = 8.6).

All constructs considered in this investigation refer to the team as the unit of analysis. Accordingly, all measures were specified on the team level. Thus, the respondents were asked to evaluate the team's properties and behaviors as a whole. The questionnaire was administered in German and in English. A pretest was conducted with seven members of a software development team at a consulting and an engineering company. After this pretest, the wording of some items and the online questionnaire's layout were refined for later use in the present study. To ensure content validity and to avoid a possible common source bias, data from various respondents were used to measure the different variables. All items used in this study were drawn from published scales. These items were also discussed with the company representatives before the questionnaire went online. This process resulted in several changes to the wording of items to increase their clarity and interpretability. Table 1 provides the descriptive statistics and correlations in respect of all the variables in our study.

Our results provide substantial support for interpersonal trust's generally beneficial role with respect to NPD team effectiveness. As such, our findings integrate with those of Bstieler (2006) who demonstrates that trust enhances product development effectiveness, as well as with that of Fleming and Waguespack (2007) who argue in favor of trust's benefits for open innovation. Moreover, we tie in with Bstieler and Hemmert (2008) who have demonstrated national culture's moderating effects on the trust-effectiveness relationship in product development teams. However, we extend their research by demonstrating that also other characteristics of internationally dispersed collaboration – geographic dispersionand computer-mediated communication – moderate this relationship.

With regard to dispersed team research, we build on the work by Cummings (2004), Hoegl et al. (2007), and Joshi et al. (2009) who point to dispersion as a moderating condition of the relationship between team processes and team performance. However, we exceed existing moderator conceptualizations by applying a multidimensional dispersion concept to the interpersonal trust effectiveness relationship in NPD teams. As such, geographic dispersion, computer-mediated communication, team membership flexibility, and national diversity serve as conditions to determine when trust becomes particularly important for internationally dispersed NPD teams' effectiveness. Defining all four dimensions as continuous variables, we also tie in with Hinds and Mortensen (2005) who conceptualized dispersion as a matter of degree rather than regarding it as dichotomous.

The study shows that in respect of geographic dispersion, computer-mediated communication, and national diversity, trust's influence on effectiveness is more important. With regard to team membership flexibility, however, our findings did not provide support for trust's predicted stronger effect on team effectiveness. A possible explanation for this deviant finding could be that stability in team membership might not be a core characteristic of global NPD teams. Our data show that flexible membership has nonsignificant correlations with geographic dispersion, computer-mediated communication, and national diversity, while these three characteristics of global NPD teams all correlate significantly with one another. Hence, we speculate that team membership flexibility does not constitute an essential element, and therefore fails to demonstrate the moderator effect which we had derived from relevant prior literature.

With respect to interpersonal trust research, our results indicate that geographic dispersion, computer-mediated communication, and national diversity significantly influence the trust–effectiveness relationship. Hence, the results highlight the importance of identifying the conditions under which trust is particularly beneficial for team effectiveness. As such, our research integrates with Atuahene-Gima and Li's (2002) work on the contextual circumstances under which interpersonal trust unfolds its potential. As the illustrations of the significant interaction effects indicate, more colocated (less dispersed) teams benefit less (or not at all) from trust, while more dispersed teams are largely dependent on it.

Such findings shed new light on the trust–effectiveness relationship in NPD teams, revealing a much more differentiated picture than the often conveyed and universally positive notion that trust is always beneficial. The conceptual and empirical analyses in this paper highlight certain conditions under which trust becomes critical (high geographic dispersion, computer-mediated communication, and national diversity) and other conditions (low computer-mediated communication, and national diversity) where trust is not a driver of NPD team effectiveness.

The findings from this research also have implications for managerial practice. Our study demonstrates interpersonal trust's increasing positive influence on NPD team effectiveness in the context of dispersed collaboration. However, the development and maintenance of high levels of trust is continuously challenged, particularly in dispersed NPD teams (Muethel and Hoegl, 2007), as perceived deviances from team norms, because of different mind sets (Arora et al., 2004), misunderstandings (Kankanhalli et al., 2006), and conflicting motives (Tomlinson et al., 2004), potentially threaten the existence of trust (Elangovan and Shapiro, 1998). The managerial implications therefore focus on trust building in the initial stages of collaboration and on trust repair where trust violations have occurred (Ferrin et al., 2007).

With regard to initial trust development, companies are well advised to invest in face-to-face meetings at the beginning of a project with high levels of geographic dispersion, computer-mediated communication, and national diversity. As our study shows, these conditions determine (at least in part) how critical trust is for team effectiveness, and hence, the degree to which investment in trust development is warranted. Kickoff meetings allow the team to get to know one another personally, to share information, and to openly discuss the goals, processes, and procedures (Cohen and Bailey, 1997). Content clarification (elucidating projects' scopes and requirements, gathering supportive background information, and creating work documents) and process information (outlining work plans and the associated time tables) foster the creation of a common ways of thinking and the development of commonly accepted behavioral norms within the team (Hackman, 1987). These, in turn, form the basis of trust development. This may be reinforced through further interaction during the project, besides focusing everyone's attention on trust-related matters such as trust violation and repair.

When trust is violated, NPD teams need procedures to repair trust to secure team effectiveness. Possible solutions are targeted at either obtaining an apology from the trust violator (Ferrin et al., 2007) or at third party trust facilitators (Mesquita, 2007).

An apology can include the offender accepting responsibility for the trust violation. Furthermore, it should convey the perpetrator's desire to reconcile with his or her fellow team members and to continue the collaboration (Kim et al., 2004). A sincere apology contains explicit statements of remorse; it is characterized by the offender's wish to shape the victim's impressions of his or her intent and motives (Tomlinson et al., 2004). For an apology to restore a trusting relationship, it should be offered soon after the violation, as it signals that the offender recognizes his or her deviant behavior (Lewicki and Bunker, 1996).

Besides the offender offering an apology, third parties could be engaged to facilitate trust reparation (Smith and Blanck, 2002). Trust facilitators employ two sets of trust-building abilities: first, those related to leading NPD teams, and second, those targeting mediation (Mesquita, 2007). Trust facilitators understand the leadership tasks in NPD teams and are able to help team leaders create a trusting environment by, for example, developing and including measures of trustworthiness in effectiveness evaluations (Abrams et al., 2003). Moreover, as mediators, trust facilitators function or intervene in cases of conflict and help the disputants achieve an agreement (Ross and Conlon, 2000). Consequently, trust facilitators mediate the negotiation of rules by disentangling and shielding new, potentially trustful, relationships from tainted existing disputes (Mesquita, 2007). Thus, managers build a trusting environment, request apologies from trust violators, and engage trust facilitators to restore trust relationships.

A few of this study's limitations should also be noted. First, the data used in this research are cross-sectional rather than longitudinal. While our study demonstrates associations between variables, it cannot establish causality. A longitudinal research design with multiple informants would probably further our knowledge in respect of whether the effects found in our study are stable over time, or may be further moderated by timing issues. Second, we consciously focused on software development teams, because this industry's digital output makes collaborating across geographical boundaries more feasible. While we do not assume that our results are specific to software development's task contexts, we encourage further research to test the transferability of our results to other industries.

With regard to further research on trust in global NPD teams, we consider two aspects as particularly interesting. First, our results highlight the importance of trust in dispersed teams, pointing to trust's compensatory effects in respect of control (Inkpen and Currall, 2004) in dispersed environments. Control mechanism incorporate formal rules, such as procedures and policies, for monitoring purposes, as well as informal or social rules, thus emphasizing organizational norms' regulatory power (Costa and Bijlsma-Frankema, 2007). As already noted above, control mechanisms are difficult to implement in dispersed environments due to the lack of face-to-face contact (Carson et al., 2003). However, the generally accepted inverse relation has recently been criticized (Das and Teng, 1998), leading to more fine-grained models (Vlaar et al., 2007). To fully understand more the relationship between trust and control in dispersed environments, future research could examine how they are related in this context.

Second, as underlined by our results, the importance of trust for global NPD team effectiveness hints at the necessity to have high levels of trust from the very beginning of collaboration. As McKnight et al. (1998) have argued, teams might not necessarily start their collaboration with low levels of trust but might demonstrate initial trust. However, Muethel and Hoegl (2007) have shown that in cross-cultural settings, global NPD teams might also be confronted with initial distrust because of negative stereotyping. Hence, further research should investigate initial trust and distrust to offer more insights into how to ensure high levels of trust in the early stages of collaboration.