Abstract

This article reviews the structural characteristics of work organizations that are likely to increase collaboration problems and tests the relationships between collaboration structure and problems using data from a survey of scientists in four fields (experimental biology, mathematics, physics, and sociology). Two groups of problems are identified: problems of coordination and misunderstandings and problems of cultural differences and information security. Greater coordination problems are associated with size, distance, interdependence, and scientific competition. Problems of culture and security are associated with size, distance, scientific competition, and commercialization. Email use is associated with reporting fewer coordination problems, but not fewer problems of culture and security, while neither phone use nor face-to-face meetings significantly reduces problems. We conclude with a discussion of the implications of these findings for designers of collaboration technologies and researchers involved in scientific collaborations.

Introduction

The last two decades have seen a significant increase in scientific collaborations that span fields, institutions, sectors, and countries. The percentage of scientific papers that had two or more authors increased from 48% in 1988 to 62% in 2001 (National Science Board, 2004). These collaborations also increasingly span institutional and national boundaries. For example, the percentage of papers that involve international collaborations increased from 9% in 1983 to 22% in 2001, with mathematics and physics leading the way (see Figure 1). This increase in team science has been driven by a variety of factors, including growing interest in scientific problems that span disciplines (e.g., mapping the human genome or studying global climate change); advances in communication and transportation technologies that make remote collaborations easier to sustain (Walsh & Bayma, 1996a); government policies that encourage collaboration, especially between universities and firms (Walsh & Cohen, 2004); and an increasingly international flow of graduate students and a lowering of political barriers in the post-Cold War era (National Science Board, 2004).

As scientific work becomes increasingly collaborative, scientists are facing the problems that come with organizing a group of workers into a team. While bringing together many scientists from a diversity of locations, disciplines, and institutions provides several important advantages for tackling interesting scientific problems (Cummings & Kiesler, 2005), scientific work teams have to decide on a division of labor, overcome scheduling issues, monitor and coordinate progress, possibly deal with distinct cultures, languages, and worldviews, and ensure that information flows where it should and does not leak to where it should not. When not carefully controlled, these "process costs" create strain in a collaboration and can impede progress (Fox & Faver, 1984).

Collaboration Structure and Collaboration Problems

A very simple collaboration might involve two scientists in the same department each working on an independent replication of an experiment and then comparing the results. As the structural complexity of a collaboration increases, its members are likely to face various difficulties in executing the project. We discuss the effects of group size, diversity, distance, group cohesion, task interdependence, competition, commercialization, and communication media on collaboration problems.

Collaboration Size

Group size is one of the central variables in explaining organizational processes (Blau & Schoenherr, 1971; Pugh, Hickson, & Hinings, 1969). While size is often a proxy for other variables such as heterogeneity or geographic dispersion, it is also a measure of the complexity of the potential division of labor and communication channels (March & Simon, 1958), as well as the difficulty of monitoring the team members (Baker & Faulkner, 1993). We anticipate that larger collaborations will be associated with an increase in reported problems, particularly related to coordination. Also, because of the cultural diversity associated with larger groups, we also expect size to increase problems of culture, trust, and information security. Security becomes particularly problematic as size increases (Baker & Faulkner, 1993).

Group Diversity

Studies of demographic diversity indicate that when individuals from different backgrounds are represented in a group, they are likely to have varying belief structures (Wiersema & Bantel, 1992) resulting in intragroup conflict due to conflicting opinions and interpretations of the task process (Eisenhardt, Kahwajy, & Bourgeois, 1997; Pelled, Eisenhardt, & Xin, 1999). Studies of collectivist organizations show that informal, peer-based coordination and decision-making mechanisms (as is typical for scientific collaborations) work well when the group is small and fairly homogeneous. However, as the group becomes more diverse, it becomes increasingly difficult to count on shared understandings to solve problems as they occur (Rothschild-Whitt, 1979; Sirianni, 1993).

Multi-disciplinary collaborations are likely to face difficulties because fields vary in terms of organization and process (Whitley, 1984). Communication may be further complicated by the failure of participants from different backgrounds to recognize the potential for (or even existence of) misunderstanding. For example, Shrum, Chompalov, and Genuth (2001) found problems in a multi-site collaboration due to different perceptions in each setting of the role of email as a medium for communication within the work group and, perhaps more importantly, the failure of participants to recognize that difference as particularly problematic. However, Cummings and Kiesler (2005) find that discipline diversity is not associated with greater coordination problems, so long as they do not also cross institutions.

As science becomes more global (Figure 1), problems of integrating heterogeneous work groups become more pronounced. For example, Cohen (2000) notes the conflicts between scientists in sub-Saharan Africa and their foreign collaborators over whether using lab equipment for personal business is an appropriate use of collaboration resources. Finally, when collaborations cross institutional spheres, such as the increasingly common university-industry collaborations (National Science Board, 2006), they are ripe for generating misunderstandings, conflicts, and delays.

One example of the diversity of perspectives associated with different fields and institutions is the variation in the degree to which research results are proprietary. Hagstrom (1974) and Walsh and Hong (2003) found substantial differences across fields in how willing scientists are to discuss their research with people outside the collaboration, with experimental biology being especially secretive and mathematics being much more open (Walsh & Bayma, 1996b). Similarly, academic scientists and industrial scientists tend to have different attitudes toward secrecy versus openness. This difference is one source of the controversies around requiring publication delays and non-disclosure policies as a part of industrial funding agreements (Blumenthal, Campbell, Anderson, Causino, & Louis, 1997; Cohen, Florida, & Goe, 1994). Note that it is not that one norm or the other is problematic. Rather, we expect that differences in understandings and expectations will cause problems.

With this in mind, we expect that respondents who are part of multi-disciplinary teams will report experiencing greater problems, both in terms of coordination and in terms of cultural and security issues. Similarly, university-industry collaborations may also be especially susceptible to problems of both types, based largely on differing expectations, as well as on differing demands outside the collaboration.

Physical Proximity

Science is a social process; therefore, contact and communication are vital to successful collaborations (Kraut, Egido, & Gallegher, 1990). Kiesler and Cummings (2002) and Olson, Teasley, Covi, and Olson (2002) highlight the advantages of co-location for collaborative work. Physical proximity provides the opportunity for informal communication, which these researchers find to be extremely important. As Kraut, et al. (1990) note, "physical proximity helps scientists avoid or minimize many of the problems that arise in the process of conducting research—meeting partners, defining problems, planning projects, supervising coworkers and subordinates" (p. 155). Kiesler and Cummings (2002) also point to the social psychological effects of co-location, such as a greater tendency to "like" the other person, and how these can facilitate co-working. Cummings and Kiesler (2005) found that even when researchers represented diverse disciplines, a collaborative project suffered fewer negative effects if the investigators were located at the same university, suggesting that distance is more relevant than diversity.

Furthermore, local collaborators can easily monitor each other (Cummings & Kiesler, 2005). This includes the unobtrusive monitoring that comes from seeing doors open, lights on, and mailboxes emptied, as well as the low-level monitoring that comes from responses to "how is it going?" or similar questions. However, maintaining this progress awareness with remote collaborators is much more difficult, leading potentially to drifting conceptions of where each person is in the project and mistrust of the other's diligence. This is a frequent complaint associated with telework, both by managers and their direct-reports, with managers feeling that teleworkers are not as diligent and remote workers feeling that managers do not recognize their efforts (Salaff, 2002).

In addition to coordination problems, remote collaborations are also likely to suffer problems of culture and information security. These culture and security issues grow out of the different local environments in which the work takes place. While the members of a collaboration share a set of processes and goals vis-à-vis the collaboration, they also are each embedded in a local work setting with its own rules, language, rhythms, histories, and myths (Armstrong & Cole, 2002). For example, Setlock, Fussell, Kiesler, and Weisband (2005) found that geographically and institutionally dispersed collaborations had more problems with coordination and problem resolution, in part due to the conflicting demands from the local work setting (Cummings & Kiesler, 2005). However, Schunn, Crowley, and Okada (2002) found no difference in the frequency of difficulties between local and remote collaborations in cognitive psychology, nor did they find remote collaborations to be less successful.

Thus, we expect that remote collaborations will have more problems of both types, although some prior work suggests that distance has no effect.

Group Cohesion

Group cohesion is also likely to have an impact on collaboration problems. We measure cohesion in terms of tie strength, consistent with social network theory. There is an ongoing debate in social network theory concerning the relative value of strong and weak ties. Some argue that loosely linked networks, e.g., those with relatively many weak ties, are advantageous because they provide bridges across groups, allowing new information—information that is known to one group but not another—to flow (Burt, 1992; Granovetter, 1973; Podolny & Baron, 1997). Others argue that strong ties are essential for information flow. For example, Murray and Poolman (1982) contend that strong ties in the scientific community are essential to the transmission of scientific information. In his study of R&D firms, Hansen (1999) concluded that weak ties are useful for transferring simple knowledge, while strong ties are necessary if the information is complex. To the extent that science is concerned with complex problems and coordinating complex tasks, then strong ties should make communicating about them easier. In addition, group cohesion has been shown to increase the adherence to group norms through sanctioning (Horne, 2001). Thus, to the extent that group cohesion facilitates communication and encourages cooperation, we expect that a collaboration rich in strong ties will report fewer problems with project coordination and misunderstandings. Similarly, groups with many strong ties (high cohesion) should have fewer problems of trust, security, and cultural differences.

Task Interdependence

The organization of work is also an important aspect of collaboration structure (Thompson, 1967). In particular, work structures vary in terms of how interdependent group members' task sets are (Van de Ven, Delbecq, & Koenig, 1976). Collaborative science is conducted with a greater or lesser degree of interdependence, depending upon the task at hand and the scientific field (Whitley, 1984). For example, a multi-city or multi-national survey might require some coordination at the beginning to develop the questions and sampling strategy, but then each team might work independently for months fielding the instrument, analyzing data, and developing initial reports, before the results are re-integrated at the end, a relatively low level of interdependence. The Human Genome Project was loosely coupled in this way. In contrast, some work requires the simultaneous, tightly-coupled joint activity of multiple group members. A space physics research team tracking an upper atmospheric event is one example (Finholt, 2002).

Greater interdependence puts greater communication demands on the group, and results in the greater likelihood of intragroup conflict (Gladstein, 1984; Schmidt & Kochan, 1972; Van de Ven, et al., 1976). Contingency models of organizational structure find that as interdependence increases (from pooled to team), the demands on the organization's communication system increase, requiring increasingly interactive and intensive forms of communication for successful execution of the collaborative work (Lawrence & Lorsch, 1967; Van de Ven, et al., 1976). Alternatively, collaborations can solve the contingency problem by reorganizing the work to uncouple various tasks. For example, Olson and Teasley (1996) found that computer-mediated work groups re-organize their task interdependence to accommodate the leaner communication environment. Similarly, Perlow (1999) found that local collaborations among engineers developed increasingly tight-linked interdependencies that tended to reduce productivity. A field experiment showed that reducing the level of interdependence, while still keeping procedures for sharing information and joint problem solving, increased productivity. Cummings and Kiesler (2005) found that remote collaborations with the goal of tool development (in computer science) had fewer problems than other collaborations, and suggested this was due to the low interdependence of the remote collaborations. Thus, we anticipate that collaborations whose work is characterized by greater degrees of interdependence will report increased problems.

Competitive Pressure and Commercialization

All of the problems discussed above are likely to be exacerbated when competitive pressure increases. Prior work suggests that scientific competition has increased over the last several decades (Walsh & Hong, 2003). The increasing commercialization of science may also increase pressure on collaborations and expose the structural and cultural fault lines (Owen-Smith, 2000). In particular, concerns about use and sharing of information may be foregrounded (Blumenthal, et al., 1997; Walsh, Cho, & Cohen, 2005). Within the U.S. industrial research and development community, there has been an increased emphasis on secrecy (Cohen, Nelson, & Walsh, 2001). Thus, we expect both scientific competition and commercial concerns to increase problems. Commercial concerns should particularly increase worries about information security and, as suggested above, highlight cultural differences.

Communication: Face-to-face and Email

Informal, direct, two-way communication is especially critical when dealing with the complex and uncertain environment that scientists and other knowledge workers frequently face (Blau, 1955; Burns & Stalker, 1961). One important outcome of such communication is not simply the passing of information, but also the creation of a shared culture, which includes developing a shared understanding of the problem and solution set (Olson, 2003). Because of the importance of such informal communication, a remote collaboration faces increased risk of problems, both in coordination and in culture (Cummings & Kiesler, 2005; Setlock, et al., 2005).

Face-to-face meetings play an important role in maintaining the integrity of the collaboration (Nardi & Whittaker, 2002; Olson & Teasley, 1996; Olson, et al., 2002). Such gatherings provide an opportunity for the members to recalibrate to the norms of the group, to dissipate accumulated grievances and re-assert (and be reassured of) the members' commitment to the project and to each other. Of course, such meetings may become an arena for generating conflict. They may also generate shared experiences that can serve as touchstones for later mediated communication. In addition, they provide a set of models of the alters' personalities, communication styles, and assumptions about good-intentions that allow mediated communication to proceed more smoothly and efficiently. Thus, we expect that collaborations will benefit from face-to-face communication. This effect might be especially strong for remote collaborations.

Recent technology, particularly email, has diminished some of the costs associated with communicating over long distances. While intense face-to-face communication has many advantages, email has gained a central place in the communication suite of many work groups, especially scientific collaborations (Walsh & Roselle, 1999).

The use of Internet-related technologies and, in particular, email, has been shown to facilitate scientific collaboration (Hesse, Sproull, Kiesler, & Walsh, 1993; Walsh & Bayma, 1996a; Walsh, Kucker, Maloney, & Gabbay, 2000). Yet prior work suggests some uncertainty as to the expected impact of email on collaboration problems. On the one hand, studies of different types of work groups find that the use of email leads to increased productivity (Cohen, 1996; Hesse, et al., 1993), streamlines communication (Finholt, Sproull, & Kiesler, 1990), allows for richer communication structures with less hierarchical differentiation (Sproull & Kiesler, 1991), and increases member participation (Bickson & Eveland, 1990).

On the other hand, email has also been demonstrated to present problems by replicating and exacerbating problems of time-consuming composition associated with other written communication (Galegher & Kraut, 1990) and by creating disruption when delivery is unfiltered (Ancona & Caldwell, 1990). Also, compared to face-to-face or even phone communication, email communication may produce misunderstandings and extreme reactions and impede reaching consensus (Sproull & Kiesler, 1991). This may be especially true for cross-cultural communications. Trust, which is of particular importance to collaborations for which competition is an issue, is more difficult to establish and to maintain among team members who rely on email to communicate (Bos, Olson, Gergle, Olson, & Wright, 2002), unless efforts have been made to establish a social connection in advance of undertaking a group task (Zheng, Veinott, Box, Olson, & Olson, 2002). In addition, because of the ease with which email can be forwarded (as well as the fact that most email transmissions are not secure), email may be associated with greater problems of information security. Cummings and Kiesler (2005) find that regular phone calls or emails did not improve the productivity of the collaboration, and were unable to substitute for the more direct forms of coordination (such as face-to-face supervision) available to local collaborators.

In addition to face-to-face and email, we will examine the effects of phone communication on collaboration problems. While both email and phone are readily available technologies for communication among scientists (especially in the U.S.), there may be important differences due to the different capabilities of each medium (Walsh & Bayma, 1996a). Briefly, phone communication has the advantage of being synchronous, allowing ready transmission of back-channel information, and of being a richer communication channel (inflection, tone, pace, etc.) (Krauss, Garlock, Bricker, & McMahon, 1977). At the same time, synchronicity has the disadvantage of requiring people to be available for communication at the same time, meaning that it requires at least a first order solution to the coordination problem (the "phone tag" problem). Phone calls have the added problem of being more expensive in most cases (Walsh & Bayma, 1996a).

In contrast, email is asynchronous (a particular benefit if a collaboration crosses many time zones), is generally free to the end user, and allows users time to compose messages and use writing aids (such as a dictionary) before transmission (which may be especially useful when the collaboration crosses cultures and the language of the email is not the native language of the sender/receiver). At the same time, loss of synchronicity and a less rich transmission channel often means a loss of back-channel communication and of various nuances of the message. Krauss, et al. (1977) found that even a short delay of one second led to a serious loss of communication efficacy in "synchronous" communications. Walsh, et al. (2000) found that email is especially useful for coordination activities, but less so for social interaction. Thus, by testing the relative impact of email and phone communication on collaboration problems, we may be able to pinpoint which aspects of the communication system are key to addressing each kind of collaboration difficulty.

Data and Methods

The data for this study come from a survey of scientists in four fields: experimental biology, mathematics, physics, and sociology. We drew a random sample from the professional membership directories for each discipline.¹ This sampling frame provided a representative sample of all types of institutions, scientists who do research and those who do not, and those who do and do not use email, as well as a random sample of collaborations with selection probability proportional to size (Bridges & Villemez, 1986). We sent 889 surveys with stamped return envelopes by postal mail in early March of 1998, and received 399 responses (51% of the eligible sample). In the following analyses, we limit our results to respondents with a Ph.D. or M.D. and who were involved in a collaboration at the time of the survey, leaving an N of 230.

Measures Used in the Analyses

Dependent Variables

The dependent variables were derived from a factor analysis of a 17-item survey question: "To what extent is each of the following a problem in your research group?" Each item was rated on a 5-point scale (from (0) not a problem to (4) a serious problem). The analysis resulted in 15 items loading onto two factors—our two dependent variables. The first factor, problems of coordination, includes items such as "getting others to see your point," "reaching decisions," "division of labor," and "misunderstandings" and explains 30% of the variance (Cronbach's alpha=.89). The second factor, problems of culture and information security, accounts for 16% of the variance, and includes "integrating other cultures," "language problems," "information leaks," and "security of information" (Cronbach's alpha=.77). Table A1 in the Appendix gives the factor scores for each item. It should be noted that "misunderstandings" loads with coordination problems. Thus, many of the effects of divergent local cultures and understandings will be captured by this category of problems, as well as by the culture/security measure.

Independent Variables

Collaboration size. For collaboration size, we asked how many collaborators worked on their current collaborative project. Because of outliers, we capped the number of collaborators at 20. With this restriction, the average number of collaborators is 4.9.

Collaboration diversity. We asked the field or discipline for each individual named as part of the respondent's main research project. We limited this and other questions about specific collaborators to a maximum of seven alters. Based on our question about the size of the research group, 85% had seven or fewer collaborators. To measure the field diversity of a collaboration, we used the following index (Blau, 1977):

Heterogeneity=1-Σp_i²

where p_i is the fraction of the population belonging to category i.
This heterogeneity score ranges from 0 (complete homogeneity) to (asymptotically) 1. In our sample, field heterogeneity has a mean of .35.

Distance. We classified as a remote collaboration those with at least one collaborator located at an institution different from that of the informant. Fifty-five percent of collaborations included at least one remote collaborator.

Strong ties. To determine tie strength, we asked the respondent to report the relation to each of his collaborators (up to seven) and coded these as "1" for a strong (a "close" or "very close") tie, or "0" for a weak ("distant") tie (Burt, 1992; Ibarra, 1998; Podolny & Baron, 1997). Our respondents reported an average of 1.65 strong ties and .63 weak ties.

Interdependence. To measure interdependence, we asked "To what extent do the people in your research group have one-person jobs: That is, in order to get the work out, to what extent do group members independently accomplish their own assigned tasks?" (Van de Ven, et al., 1976). Responses varied on a seven-point scale (from (1) to a great extent to (7) very little). We reverse coded this item to make it a measure of interdependence. The mean level of interdependence for our sample is 2.9.

Competition. We measure competition using Hagstrom's (1974) item: "How concerned are you that you might be anticipated in your current research?" The response scale (after reverse coding) went from 1="I am not at all concerned" to 4="I am very concerned." We recoded 5 ("I have already been anticipated") as missing, since it was not clear how to rank this. The mean level of concern is 2.2.

Commercial orientation. To measure the commercial orientation of the respondents' research, we asked, "Within the past five years, have you applied for a patent based on your research?" Fifteen percent had applied for patents. We also included a measure of university-industry collaboration (which can be viewed as a measure of diversity as well). We coded this as 1 for those who had at least one industry collaborator and at least one university collaborator. Eight percent of our respondents reported a university-industry collaboration.

Communication. Our communication measures are the self-reported number of email messages sent per day and the number of telephone calls made per day.² In order to correct for outliers, we recoded these variables to cap the number of email or phone messages at 10 per day. The mean number of email messages sent per day is 5.1, while phone calls made per day average 3.2.

We also asked our respondents to report the number of times their collaborations meet per year. We grouped the frequency according to the distribution, such that 16 or more meetings per year (i.e., "at least bi-weekly")=4, 7 to 15 ("monthly")=3, 3 to 6 ("quarterly")=2, 1 to 2 ("once or twice a year")=1, and 0 ("never")=0. This grouping gives us a mean level of face-to-face meeting of 1.9 (approximately "quarterly").

We also control for field in our regressions. Appendix Table A2 gives means, standard deviations, and the correlation matrix for our measures.   

Results

To test the main effects of collaboration structure, and controlling for other predictors, we ran ordinary least squares regression of our factor measures of coordination problems and problems of culture/security on our measures of collaboration structure. We then added our measures of communication to see if these modify the effects of the structural variables and if they have direct effects on collaboration problems. Table 1 shows the results for models that include only the structural variables and models that include the measures of communication.

Coordination/Misunderstandings

Because of our relatively small sample size, we designated .10 as our alpha level of significance. Model 1 indicates that as expected, collaboration size, remote collaboration, task interdependence, and scientific competition significantly increase problems. Though strong ties predict a reduction in problems, the effect is not significant. Concern over being anticipated—our measure for scientific competition—is the most significant predictor of problems, which seems to point to the difficulty in coordinating the work when there is greater pressure to finish quickly. However, commercialization (as measured by patent application) has minimal impact. Interestingly, field heterogeneity does not significantly affect problems of coordination, although the effect is positive. Cummings and Kiesler (2005) find a similar result. Similarly, university-industry collaborations are positive but not significant. Thus, we find only limited evidence for our conjectures about the impact of diversity on coordination problems and misunderstandings.

Model 2 adds the measures of communication. Here we see that our communication variables—phone calls, email, and meeting face-to-face—do not lessen the effects of the other significant predictors. The effect of strong ties becomes marginally significant with the addition of the communication measures. In addition, we see that as phone calls made per day increase, so do problems with coordination. Email, in contrast, is associated with significantly fewer problems of this nature.

Culture/Security

Models 3 and 4 of Table 1 give the results of the impact of the same set of variables on problems of culture and security in scientific collaborations. The first equation measures the effects of collaboration structure on problems of culture and security. As anticipated, task interdependence, scientific competition, and commercial orientation are significant predictors of increased problems of culture and security, as is collaboration size. Interestingly, neither field diversity nor having an industry collaborator has an effect here. Therefore, our hypotheses predicting that different cultures across disciplines and organizations will increase these types of problems are not supported.

In Model 4, we see that our communication variables do not lessen the impact of the structural variables. Each of the significant variables in Model 3 maintains its presence in the complete model, indicating that communication cannot be counted on to ease the negative effects of the other predictors. Furthermore, while number of phone calls made per day is associated with a significant increase in problems, email has very little influence, indicating that, in contrast to its effect on coordination/misunderstanding problems, it is of limited use in alleviating problems associated with culture and security (Walsh, et al., 2000).

Conclusion

Our results suggest that scientific collaborations face a variety of problems that cluster into two types: problems of coordination and misunderstandings and problems of culture and information security. We also find that collaboration structure predicts collaboration problems. Remote collaborations are particularly prone to problems of coordination and misunderstandings, although, interestingly, they are not any different in terms of culture or security problems. Scientific competition heightens both kinds of problems, by putting pressure on all the participants and stressing the faults in the structure. Commercial concerns affect culture and security (not surprisingly), but not coordination or misunderstandings. Surprisingly, diversity has very little effect on cultural or security problems and only slight effects on coordination and misunderstandings. While this may be a limitation of our measures, it suggests that structural factors, such as distance, task interdependence and competitive pressure, are more critical than participant demographics (Cummings & Kiesler, 2005).

Task interdependence is also critical. Collaborations may have some design freedom in structuring the division of labor, and thus assigning the work in a way that minimizes task interdependence may be beneficial (Cummings & Kiesler, 2005; Kiesler & Cummings 2002; Olson & Teasley, 1996). However, this must be done without losing some of the important productivity benefits that come from an interdependent division of labor (March & Simon, 1958; Smith, 1976 [1776]). Alternatively, developing communication processes that match the level of interdependence may help alleviate collaboration problems (Lawrence & Lorsch, 1967).

Interdependence is an understudied aspect of scientific collaborations. We know that fields, and projects, vary in terms of how interdependently the tasks are structured (Olson & Teasley, 1996; Perlow, 1999; Whitley, 1984). A more systematic series of studies building on these results may help us specify the factors that influence the necessary degree of interdependence and how this interdependence affects problems in the collaboration, would likely produce significant new knowledge on how to effectively structure knowledge work. If these results are replicated, this has important implications for attacking the problems of collaborations. Concentrating on the structure of the work may be more important than worrying about what mix of backgrounds are including in the work group.

As argued above, many of these structural effects are related to the cultural gaps that structural barriers create and the resulting communication difficulties that these gaps generate. Thus, not only is it more difficult to transmit a message to someone far away, the task is made even more complicated by the need to incorporate a translation scheme in the message in order for the receiver to consume it properly. However, it is size, interdependence, distance and competition pressures that primarily cause problems for the collaboration, not discipline or sector diversity. Thus, it may be the specifics of the local context, rather than training and other characteristics associated with a particular field or sector, that make communication and coordination difficult.

In addition, we find that email may have some utility in overcoming the problems of coordination that collaborations face. However, email may not decrease difficulties of cultural differences or security. Thus, email may be better suited for dealing with the technical rather than the social aspects of collaborations (Walsh, et al., 2000). Phone calls are associated with greater problems of both types. One explanation, of course, is reverse causation. Another possibility is that the one-on-one nature of phone calls (conference calls excepted) may do little to relieve the group-level problems that collaborations face. The more public nature of on-line discussions (with one-to-many emails, copying in of recipients, and discussion lists, for example) may allow groups to overcome some of their issues as they arise (Skovholt & Svennevig, 2006).

We were a bit surprised by the weak effect of face-to-face meetings. While they have some benefits for reducing cultural and security concerns, the effect is not significant. In terms of coordination, the positive, albeit not significant, effect suggests that the coordination costs of setting up such meetings may overwhelm the coordination and mutual understanding gains that the meetings produce, resulting in no net gain, and possibly even a net loss.

Thus, the strengths of email and limitations of phone calls and face-to-face meetings as a way to reduce collaboration problems suggests that asynchronous communication—which easily allows both one-to-one or one-to-many transmission, and which allows easy transmission of longer, text-based messages (documents, agendas, schedules)—may be critical for keeping collaborations on track. In contrast, synchronous communication, although richer in back-channel information, may be neither necessary nor effective, perhaps due to the coordination costs required to set up the conversation (by phone or face-to-face). Other technologies that provide regular information on progress, that allow answering quick questions without necessarily demanding immediate responses, and that allow processing off-line, may be key to facilitating scientific group work; by generalization, they may also be important for other types of knowledge worker teams (Perlow, 1999). An implication of this is that groupware designers might want to focus less on creating a virtual, synchronous tele-presence and more on providing tools that facilitate keeping group members informed about each other's progress and diligence (and, perhaps, rival demands), and sharing research materials, while also not overwhelming them or demanding that schedules be tightly coupled (Olson & Teasley, 1996; Perlow, 1999; Salaff, 2002).

As collaboration technologies develop, these different effects can be explored to reveal what aspects of the technologies may produce divergent outcomes. While email is far from a cutting-edge collaboration technology, its substantial penetration in scientific communities allows for the generation of base-line measurements against which more sophisticated but still experimental technologies might be compared.

This study has important limitations. One is the measurement problem. Information about the collaborations was reported by only one member of each collaboration. We cannot be certain that one individual's perceptions of the group's problems were consistent with the other team members' experiences. Further, the survey provided only a snapshot of the collaboration's history. The problems represented in cross-sectional data may be ephemeral, and may not represent the ebb and flow of conditions over time. Indeed, a longitudinal study of an international gene research collaboration found significant fluctuations in problems over the course of the collaboration (Atkinson, Batchelor, & Parsons, 1998). The data are also a historical snapshot, taken in 1998. With the development of collaboration technology and the growth of remote collaborations, scientists now may be both more skilled at conducting such collaborations and have more tools available to facilitate them, which may change the relationships between collaboration structure and problems. Thus, replications of this research are needed across different technology environments to see how contingent the present results are on the specific technologies used. Finally, there is the problem of causal direction. Some of our observed relations may go in the opposite direction. For example, more phone calls may not be the cause of culture/security problems, but rather an attempt to solve them. More research using experimental or longitudinal data may help address this issue.

Further research on the structure of scientific collaborations, as well as on other knowledge-based work teams, could advance understanding of the mechanisms that cause each type of problem identified in this study, and of the role of different communications strategies and technologies in reducing or exacerbating these difficulties. The results can guide engineers when designing collaboration support technologies toward the characteristics of the work organization that are likely to be problematic, such that new technology might help alleviate some of the difficulties. Similarly, when developing collaborations, scientists and science managers can be aware of what designs are likely to be problematic and therefore can consider alternative ways of organizing, or at least can enter collaborations forewarned about the types of difficulties they are likely to face.

Notes

1. The membership directories were those of the Federation of American Societies of Experimental Biology (Biochemical and Molecular Biology and Cell Biology divisions), American Mathematical Society, American Physical Society, and American Sociological Association. For details of the survey method, see Walsh, et al. (2000).
2. Hesse, et al. (1993) report self-reported email messages per day were correlated .50 (p<.0001) with actual usage data drawn from a three-month period. A parallel analysis using likelihood of using email to contact collaborators produced qualitatively similar results.

References

Ancona, D., & Caldwell, D. (1990). Information technology and work groups. In J. Galegher, R. Kraut, & C. Egido (Eds.), Intellectual Teamwork (pp. 173-190). Hillsdale, NJ: Lawrence Erlbaum Associates.

Armstrong, D., & Cole, P. (2002). Managing distances and differences in geographically distributed work groups. In P. Hinds & S. Kiesler (Eds.), Distributed Work (pp. 167-190). Cambridge, MA: MIT Press.

Atkinson, P., Batchelor, C., & Parsons, E. (1998). Trajectories of collaboration and competition in a medical discovery. Science, Technology & Human Values, 23 (3), 259-284.

Baker, W., & Faulkner, R. (1993) The social organization of a conspiracy. American Sociological Review, 58 (6), 837-860.

Bickson, T., & Eveland, J. (1990). The interplay of work group structures and computer support. In J. Galegher, R. Kraut, & C. Egido (Eds.) Intellectual Teamwork (pp. 245-290). Hillsdale, NJ: Lawrence Erlbaum Associates.

Blau, P. (1955). Dynamics of Bureaucracy. Chicago: Chicago.

Blau, P. (1977). Inequality and Heterogeneity. New York: Free Press.

Blau, P., & Schoenherr, R. (1971). Structure of Organizations. New York: Basic Books.

Blumenthal, D., Campbell, E., Anderson, M., Causino, N., & Louis, K. (1997). Withholding research results in academic life science. JAMA 277 (15), 1224-8.

Bos, N., Olson, J., Gergle, D., Olson, G., & Wright, Z. (2002). Effects of four computer-mediated communications channels on trust development. In Proceedings of CHI 2002. New York: ACM Press.

Bridges, W., & Villemez, W. (1986). Informal hiring and income in the labor market. American Sociological Review, 51 (4), 574-582.

Burns, T., & Stalker, G.M. (1961). Mechanistic and organic systems of management. Management of Innovation. London: Tavistock Publications.

Burt, R. (1992). Structural Holes. Cambridge, MA: Harvard Press.

Cohen, J. (1996). Computer mediated communication and publication productivity among faculty. Internet Research, 6 (23), 41-63.

Cohen, J. (2000). Balancing the collaboration equation. Science, 288 (5474), 2155-2159.

Cohen, W., Florida, R., & Goe, R. (1994). University-industry research centers in the United States. Working Paper, Center for Economic Development, Carnegie Mellon University.

Cohen, W., Nelson, R., & Walsh, J. (2001). Protecting their intellectual assets: Appropriability conditions and why U.S. manufacturing firms patent (or not). http://papers.nber.org/papers/W7552

National Bureau of Economic Research working paper #7552.

Cummings, J., & Kiesler, S. (2005). Collaborative research across disciplines and organizational boundaries. Social Studies of Science, 35 (5), 703-722.

Eisenhardt, K., Kahwajy, J., & Bourgeios, L. (1997). Conflict and strategic choice. California Management Review, 39 (2), 42-62.

Finholt, T. A. (2002). Collaboratories. Annual Review of Information Science and Technology, 36, 74-107.

Finholt, T., Sproull, L., & Kiesler, S. (1990). Communication and performance in ad hoc task groups. In J. Galegher, R. Kraut, & C. Egido (Eds.) Intellectual Teamwork. (pp. 291-325). Hillsdale, NJ: Lawrence Erlbaum Associates.

Fox, M. F., & Faver, C. (1984). Independence and cooperation in research. Journal of Higher Education, 55 (3), 347-359.

Gladstein, D. (1984). A model of task group effectiveness. Administrative Science Quarterly, 29 (4), 499-517.

Galegher, J., & Kraut, R. (1990). Technology for intellectual teamwork: Perspectives on research and design. In J. Galegher, R. Kraut, & C. Egido (Eds.) Intellectual Teamwork (pp. 1-20). Hillsdale, NJ: Lawrence Erlbaum Associates.

Granovetter, M. (1973). The strength of weak ties. American Journal of Scoiology, 78 (6), 1360-1380.

Hagstrom, W. O. (1974). Competition in science. American Sociological Review, 39 (1), 1-18.

Hansen, M. (1999). The search-transfer problem. Administrative Science Quarterly, 44 (1), 82-111.

Hesse, B., Sproull, L., Kiesler, S., & Walsh, J. (1993). Returns to science. Communications of the ACM, 36 (8), 90-101.

Horne, C. (2001). The enforcement of norms. Social Psychology Quarterly, 64 (3), 253-266.

Ibarra, H. (1998). Network assessment exercise: MBA version. Harvard Business School Press Reprint Number 9-4797-002.

Kiesler, S., & Cummings, J. (2002). What do we know about proximity and distance in work groups? A legacy of research. In P. Hinds & S. Kiesler (Eds.) Distributed Work (pp. 57-80). Cambridge, MA: MIT Press.

Krauss, R., Garlock, C., Bricker, P., & McMahon, L. (1977). The role of audible and visible back-channel responses in interpersonal communication. Journal of Personality and Social Psychology, 35 (7), 523-529.

Kraut, R., Egido, C., & Galegher, J. (1990). Communication in scientific research. In J. Galegher, R. Kraut, & C. Egido (Eds.) Intellectual Teamwork (pp. 149-171). Hillsdale, NJ: Lawrence Erlbaum Associates.

Lawrence, P., & Lorsch, J. (1967). Organizations and Environments. Cambridge: Harvard.

March, J. G., & Simon, H. G. (1958). Organizations. New York: Wiley.

Murray, S., & Poolman, R. (1982). Strong ties and scientific literature. Social Networks, 4 (3), 225-232.

Nardi, B. & Whittaker, S. (2002). The place of face-to-face communication in distributed work. In P. Hinds & S. Kiesler (Eds.) Distributed Work (pp. 83-110). Cambridge, MA: MIT Press.

National Science Board [NSB]. (1993). Science and engineering indicators, 1993. Washington: US GPO.

National Science Board [NSB]. (1998). Science and engineering indicators, 1998. Washington: US GPO.

National Science Board [NSB]. (2000). Science and engineering indicators, 2000. Washington: US GPO.

National Science Board [NSB]. (2004). Science and engineering indicators, 2004. Washington: US GPO.

National Science Board [NSB]. (2006). Science and engineering indicators, 2006. Washington: US GPO.

Olson, G. M. (2003, Feburary). Collaboratory life. Paper presented at the American Association for the Advancement of Science. Denver, CO.

Olson, J. S., & Teasley, S. D. (1996). Groupware in the wild. In Proceedings of CSCW 1996 (pp. 419-427). New York: ACM Press. 

Olson, J., Teasley, S., Covi, L., & Olson, V. (2002). The (currently) unique advantages of collocated work. In P. Hinds & S. Kiesler (Eds.) Distributed Work (pp. 113-135). Cambridge, MA: MIT Press.

Owen-Smith, J. (2000). Public Science, Private Science. Unpublished doctoral dissertation, Department of Sociology, University of Arizona.

Pelled, L., Eisenhardt, K., & Xin, K., (1999). Exploring the black box. Administrative Science Quarterly, 44 (1), 1-28.

Perlow, L. (1999). The time famine. Administrative Science Quarterly, 44 (1), 57-81.

Podolny, J., & Baron, J. (1997). Resources and relationships. American Sociological Review, 62 (5), 673-693.

Pugh, D. S., Hickson, D. J., & Hinings, C. R. (1969). An empirical taxonomy of structures of work organizations. Administrative Science Quarterly, 14 (1), 115-126.

Rothschild-Whitt, J. (1979). The collectivist organization. American Sociological Review, 44 (4), 509-527.

Salaff, J. (2002). Where home is the office. In B. Wellman & C. Haythornthwaite (Eds.), The Internet in Everyday Life (pp. 464-495). Oxford: Blackwell.

Schmidt, S., & Kochan, T. (1972). Conflict. Administrative Science Quarterly, 17 (3), 359-370.

Schunn, C., Crowley, K., & Okada, T. (2002). What makes collaborations across a distance succeed? In P. Hinds & S. Kiesler (Eds.), Distributed Work (pp. 407-430). Cambridge, MA: MIT Press.

Setlock, L. D., Fussell, S. R., Kiesler, S., & Weisband, S. (2005). Managing work trajectories in collaborative writing. Unpublished manuscript. Carnegie Mellon University.

Shrum, W., Chompalov, I., & Genuth, J. (2001). Trust, conflict and performance in scientific collaborations. Social Studies of Science, 31 (5), 681-730.

Sirianni, C. (1993). Learning pluralism. In F. Fischer & C. Sirianni (Eds.), Organization & Bureaucracy (pp. 555-570). Philadelphia: Temple.

Skovholt, K., & Svennevig, J. (2006). Email copies in workplace interaction. Journal of Computer-Mediated Communication, 12(1), article 3.  Retrieved December 31, 2006 from http://jcmc.indiana.edu/vol12/issue1/skovholt.html

Smith, A. (1976) [1776]. The Wealth of Nations. Oxford: Clarendon Press.

Sproull, L., & Kiesler, S. (1991). Connections. Cambridge, MA: MIT Press.

Thompson, J. (1967). Organizations in Action. New York: McGraw-Hill.

Van de Ven, A., Delbecq, A., & Koenig, R. (1976). Determinants of coordination modes within organizations. American Sociological Review, 41 (2), 322-338.

Walsh, J., & Bayma, T. (1996a). The virtual college. The Information Society, 12 (4), 343-363.

Walsh, J., & Bayma, T. (1996b). Computer networks and scientific work. Social Studies of Science, 26 (3), 661-703.

Walsh, J., Cho, C., & Cohen, W. (2005). The view from the bench. Science, 309 (5743), 2002-2003.

Walsh, J., & Cohen, W. (2004, August). Does the golden goose travel? American Sociological Association Annual Meeting, San Francisco.

Walsh, J., & Hong, W. (2003). Secrecy is increasing in step with competition. Nature, 422 (6934), 801-802.

Walsh, J., Kucker, S., Maloney, N., & Gabbay, S. (2000). Connecting minds. Journal of the American Society for Information Science, 51 (14), 1295-1305.

Walsh, J., & Roselle, A. (1999). Computer networks and the virtual college. Science, Technology Industry Review, 24, 49-78.

Whitley, R. (1984). The Intellectual and Social Organization of the Sciences. Clarendon: Oxford.

Wiersema, M., & Bantel, K. (1992). Top management team demography and corporate strategic change. Academy of Management Journal, 35 (1), 91-121.

Zheng, J., Veinott, E., Bos, N., Olson, J., & Olson, G. (2002). Trust without touch. In Proceedings of CHI 2002. New York: ACM Press.

Appendix

About the Authors

John P. Walsh is an associate professor of public policy at Georgia Institute of Technology. His current research is on innovation, patenting, and university-industry linkages in Japan and the U.S.
Address: School of Public Policy, Georgia Institute of Technology, 685 Cherry Street, Atlanta, GA 30332-0345, USA

Nancy G. Maloney is an adjunct instructor at South Mountain Community College. Her research interests are in gender inequality in science and the digital divide.
Address: Department of Sociology, University of Illinois at Chicago, 1007 W. Harrison St, Chicago, IL 60607-7140, USA