Limits to Scalability of Shared Mental Models in Large-Scale Software Development


This work was created during a seminar course at the the University of Mannheim during my studies. Do not cite this document, it has not been peer-reviewed or published.

You might use it as a entry point for relevant literature.


  1. Introduction and Motivation
  2. Terminology and Foundations
    1. Shared Mental Models
    2. Large-Scale Software Development
  3. Methodology
  4. Results
    1. Overview of Studies
    2. Findings about Research Questions
  5. Discussion
    1. Main Findings
    2. Implications for Future Research
    3. Limitations
  6. Conclusion
  7. Bibliography

List of Tables 

1  Introduction and Motivation

”U.S., European, or Japanese companies often outsource software development to Eastern Europe, Russia, or the Far East” (Sutherland et al.2009, p. 277). This ongoing trend was already identified by Herbsleb & Mockus (2001a) and is called ”Global Software Development” (GSD). In 2000 already 185 out of the Fortune 500 companies outsourced software development to India (Herbsleb & Mockus2001b). This distributed approach differs from co-locating the software development team to a high degree. Co-located teams experience only low barriers to communicate, whereas communication of distributed teams is much more demanding. The main drivers are increasing software intensity and globalization (Herbsleb & Moitra2001). Limited workforce, development costs or round-the-clock development motivates large-scale software development (Herbsleb & Mockus2001b).

Large-scale software development is done by many teams, which share at least one common goal and have interdependencies with at least one another team (Mathieu et al.2001). This type of organization is called a multi-team system (MTS) (Scheerer et al.2014). Global software development (GSD) is a more granular definition, which concludes the distribution of software development across geographic dispersed sites (Herbsleb & Mockus2001b).

The term ”Shared Mental Models” was introduced by Cannon-Bowers et al. (1993) to describe the shared understanding within a team about the task and the team (Espinosa et al.2001). Many researchers have built upon this definition and conducted empirical research (Levesque et al.2001Lim & Klein2006Mathieu et al.2000Scozzi et al.2008). Mathieu et al. (2000) for example showed that knowledge about the team positively correlates to the team performance whereas knowledge about the task does not. But there is a lack of research on shared mental models in context of multi-team systems and software development. There is currently one paper empirically examining shared mental models in large-scale, distributed software development (Espinosa et al.2002b).

Shared mental models allow an explanation of how teams build and share a common understanding. Many researchers also derived that performance can be explained by suitable shared mental models (Carpenter et al.2008). This is also valid for software development (Espinosa et al.2002b). But in the context of large-scale and also global software development there is only little research, which aims to investigate in shared mental models. The objective of this review is to identify general literature to explain the fundamentals and the literature which explicitly addresses this issue.

It is important to further investigate in this topic because it is actually used and has a high impact on the world of business. But this is not very well researched (Espinosa et al.2002b). According to practitioner even linear scalability can be reached (Sutherland et al.2009). However, there is no empirical evidence proofing this. Relevant factors need to be identified and correlated.

2  Terminology and Foundations

2.1  Shared Mental Models

Shared mental models (SMM) were defined to describe and explain the behaviour of teams (Jonker et al.2011). Cannon-Bowers et al. (1993) defined SMM as knowledge structure hold by each team member enabling them to have a common and accurate understanding of their task and environment. This enables the prediction of team members’ needs and actions as well as the required coordination actions between team members (Jonker et al.2011Salas et al.2005).

Cannon-Bowers et al. (1993); Klimoski & Mohammed (1994) and also other researchers suggest that there are several mental models to be shared among a team. Mathieu et al. (2000) define two major domains, which mental models can be reduced to: task and team. Task-related model subsumes Cannon-Bowers et al. (1993)’s equipment and task model and team-related subsumes Cannon-Bowers et al. (1993)’s team interaction and team model.

Shared mental models have three attributes: Sharedness, similarity and accuracy. Sharedness describes how much of the knowledge is shared among team members. Similarity is the extent to which mental models overlap among team members and accuracy is the degree to which the SMM conform to the real world. (Mohammed et al.2010)

The term ”team mental model” (TMM) is also often used. But the definition differs among authors. Lim & Klein (2006) use TMM as a synonym for SMM. Langan-Fox et al. (2004) and Klimoski & Mohammed (1994) focus more on team members as collective instead of dyads. In this paper the terms are used as synonyms. SMM are titled as one of the ”Big Five in Teamwork” (Salas et al.2005).

Scalability in the context of shared mental models has not been defined yet. So far shared mental models have been mostly researched in the domain of intra-team coordination. There is a lack of research in inter-team coordination (Standifer & Bluedorn2006). The objective of this paper is to address how shared mental models behave not only within one team, but within teams of teams. And by doing so, one goal is to find the extent to which shared mental models can be distributed among teams.

2.2  Large-Scale Software Development

Large-scale software development is a collaboration between many individuals, who unify knowledge from diverse areas in order to develop a complex software (Espinosa et al.2001). Although there is empirical evidence that cross-site teams take more time for the same work as co-located teams, geographic dispersed setups are more and more common (Herbsleb & Mockus2001a). As projects at this scale need to integrate a lot of diverse knowledge, it is not always possible to locate all teams or individuals at one place (Espinosa et al.2001).

There are three main drivers for distributed software development recognized in literature. The progress in communication technologies makes it easier to communicate across geographic boundaries (Espinosa et al.2007a). Additionally, the dispersion of workforce enables companies to cover new markets, which are distant to old structures, as well as to be near to clients (Espinosa et al.2007a). Finally the geographic dispersion leverage companies to access to skilled labour, which would not be accessible otherwise (Espinosa et al.2002b).

But there are also disadvantages of dispersion. If the software development is done across different sites, it is harder to manage the process, which often leads to an increased development time (Espinosa et al.2007a). Coordination in software development is substantial because the development process contains complex dependencies. But it is difficult for distributed teams to coordinate well (Espinosa et al.2007a).

Organisational Setup

A larger number of individuals is usually not managed in one team but rather by many teams. Two or more teams that share at least one common goal and need to interact with at least one another team are called a multi-team system within the organizational psychology domain (Mathieu et al.2001Scheerer et al.2014). MTS expand the team-of-teams concept (Mathieu et al.2001).

Process Framework

The process framework of the respective organization has to be taken into account. Therefore, it is necessary to differentiate between the waterfall and the scrum model. Waterfall projects aim to define a project as exact as possible before any development begins and takes most of the responsibility from the developers. In contrast, in scrum projects there is only little requirements engineering in the forehand and also the responsibility is transferred back to the developers. (Scheerer et al.2014)

3  Methodology

The objective of this review is to address and answer the following research questions.

  • RQ1: Do and if yes, how do shared mental models impact large-scale software development with multiple teams? If yes, are there limitations?
  • RQ2: How can shared mental models be scaled up in large-scale software development?
  • RQ3: How can shared mental models be used to validate, monitor and explain the performance of teams in large-scale software development?

The literature was found by applying methods described by Webster & Watson (2002). As supposed, there is not much research on this topic. Additional search was performed to find interdisciplinary literature. Only literature which was written in English and available online was taken into account.

The approach was very similar to the one Webster & Watson (2002) proposes. First, the table of contents of the specified journals were studied to identify high quality literature and also to create a categorized list of keywords. The journals MIS Quarterly, Journal of Management Information Systems, European Journal of Information Systems, Organization Science and Journal of Applied Psychology were looked up.

Secondly, the citations of the found literature were reviewed to determine other articles, which should be taken into account. Webster & Watson (2002) calls this step ”Go backward”.

Thirdly, Google Scholar was used to determine other relevant literature, which cites the previous found literature. In the notation of Webster & Watson (2002) this is ”Go forward”.

Because the topic is not very well researched, the approach was extended. Additional literature was searched on Google Scholar. Table 3.1shows the identified relevant keywords. The search was separated in two steps. In order to get some general knowledge, the terms of each category were searched separately. Afterwards, to span between the domains the keywords of category ”SMM” were combined with the keywords of the category ”GSD” as well as with the keywords of the category ”MTS”. The keywords were combined via the ”AND”-operator. An example for a separate search is ”Global Software Development”, ”Shared mental model AND Global Software Development” is an example for the combination of keywords. Due to the inexhaustible nature of the web, the review was only conducted on the first 50 results of each search. For example, ”Shared mental model AND Global Software Development” results in 181000 hits from which 1000 can be browsed.

Table 3.1: Categorization of Keywords
SMMShared mental model, mental model, team mental model, shared task understanding
GSDGlobal Software Development, Global software engineering, cross continent development, dispersed teams, distributed development, geographically distributed software development, large-scale software development, large-scale software engineering, geographically distributed software development
MTSmulti-team system, across teams, distributed team, virtual team

A paper was only added to the relevant literature if it got accepted after the inclusion decision, which was done in the following sequential order: title, abstract, introduction and conclusion, full review. At each steps, the relevance was checked. A paper needed to address at least one of the general categories, ideally two or all.

A paper was downloaded after the abstract was found to be relevant. 82 papers were downloaded and 42 were used in this review. Espinosa, Herbsleb, Cannon-Bowers, Mathieu and Salas were identified as the authors, which are leaders of the research in context of this paper.

4  Results

4.1  Overview of Studies

Table 4.1shows the paper sorted by year. The selection was found via the process explained in Chapter 3. The two papers in 1993 and 1994 are those building the concept for shared mental models most of the others rely on. 2001 and 2003 mark the peak of research. In these years research was especially done in the area of global software development due to increasing awareness of globalization and internet getting massively available. After that, the research in this area decreases until the years from 2005 to 2008. In these years research was mostly focused on shared mental models and their application to teams.

Table 4.1: Paper by Year of Publication
YearPapers (Absolut)Papers (relative)
1993-199424,8 %
1999-200037,1 %
2001-20021228,6 %
2003-200424,8 %
2005-2006614,3 %
2007-2008716,7 %
2009-201037,1 %
2011-201237,1 %
2013-201437,1 %
201511,4 %

Table 4.2shows the identified papers categorized into shared mental models, global software development and shared mental models in the context of global software development. Additionally, the papers are separated in type of paper and research method. The number of relevant papers reflects the extent of research, which has been conducted in the specific area.

Shared mental models uniquely have literature reviews in this summary. This is attributed to the age of the theory as well as the general applicability in many domains and reflects maturity to some extent. Additionally, the overall sum of literature proves a high extent of research in this domain.

Contrary, global software development does not have a literature review. Neither does it have a quantitative research, which could be attributed to the general problem that is not easy to simulate in quantitative research in such a setting.

The higher number of SMM in the context of Global Software Development can be traced back to Espinosa, which authored 5 out of the 9 publications. This leads to the conclusion that there is a low extent of maturity and research is needed to gain more empirical knowledge.

Table 4.2: Paper Categorization
CategoryType of PublicationSum
Shared Mental ModelLiterature-Review223
Global Software DevelopmentLiterature-Review08
Shared Mental Model in the Context
of Global Software Development

4.2  Findings about Research Questions

The findings of the study will be presented in three parts. First some general aspects of the literature regarding shared mental models will be replicated. Afterwards literature studying global software development is reviewed and in the end the research of shared mental models in the context of global software development will be disclosed.

Shared Mental Models

SMM are most often investigated in the context of a single team (Standifer & Bluedorn2006). Although the focus of this work is on multi-team SMM, this section will handle SMM inside a team because the mechanisms and different variables need to be understood.

SMM highly influences the performance of a team (Cannon-Bowers et al.1993). Most researchers focus on teams in environments like command and control (Badke-Schaub et al.2007) where synchronous tasks are necessary (Espinosa2001). Teams with higher-quality and higher-sharedness SMM have better team processes and performance (Mathieu et al.2005). Lim & Klein (2006) showed that team performance is positive correlated with SMM similarity and SMM accuracy. But Badke-Schaub et al. (2007); Espinosa (2001) also showed, that SMM can be detrimental if the SMM are inaccurate. In contrast, Mathieu et al. (2000) did not find a correlation between SMM similarity and team performance. Langan-Fox et al. (2004); van den Bossche et al. (2011) both concludes that task-related shared mental models improve the performance of teams.

Levesque et al. (2001) and Mathieu et al. (2000) have researched the development of SMM over time. Mathieu et al. (2000) concludes that the mental models in a team convergence over time and thereby increase performance. Levesque et al. (2001) comes to the conclusion that especially in temporary task-focused teams the SMM divergences over time because team members specialise and do no need to interact comprehensively anymore.

Global Software Development

Besides SMM, a lot of research has been conducted in the field of global software development. Geographical dispersion, time separation and cultural differences are identified as main drivers for problems (Baffin et al.2001Espinosa et al.2006). Baffin et al. (2001); Espinosa et al. (2007a); Herbsleb & Mockus (2001a,b); Scheerer et al. (2014) found out, that coordination & communication are extremely challenging at large-scale and disrupted at global scale. Multi-Sites are introducing conflicts and it is hard to involve the right people (Herbsleb & Mockus2001a,b). According to Baffin et al. (2001) geographic dispersed teams experience a loss of ”teamness”. Herbsleb & Mockus (2001a,b) concluded cross-site work takes longer compared to co-located work if the circumstances (equal size and complexity of project) are the same.

The literature also presents recommendations to deal with these problems. Herbsleb & Mockus (2001b) propose to create a ”virtual-site”, which means the establishment of common tools, practices and processes. This extends the suggestions of Herbsleb & Mockus (2001a,b); Sutherland et al. (2009) for a adequate infrastructure like phones, development environment, shared calendars, instant messaging, high quality video conferencing and a tool to find the responsible person or knowledge holder. Baffin et al. (2001); Herbsleb & Mockus (2001b) recommend de-coupling of work as much as possible to let the teams work as independent as possible. According to Scheerer et al. (2014); Sutherland et al. (2009) the scrum approach should not only be used in small-scale setups but also in large-scale setups (Scrum-of-Scrums) because of a much higher productivity compared to the waterfall process model. To deal with geographical dispersion and time separation, Baffin et al. (2001); Sutherland et al. (2009) deployed so called ”proxies” or ”liaisons” where a team member acts as a pivot to connect the teams although they are separated. Espinosa et al. (2001) proposes teambuilding measures at the beginning so that team members get to know each other. This could be done as face-to-face meeting, which was found out to be helpful if team members are from different cultures (Espinosa et al.2006Sutherland et al.2009). This is also relevant in the context of the work of Graham et al. (2004) who investigated in social network distance. Their research concluded that social network distance is as important as physical distance.

Shared Mental Models in the Context of Global Software Development

Espinosa (2001); Espinosa et al. (2002b) determined that SMM have a positive effect on asynchronous and distributed teams. But the geographical distance has negative effects, which are partly proved (Espinosa et al.2002b). Partly because these effects go often hand in hand with other factors like time separation, cultural differences and language (Espinosa et al.2001).

Especially ”frequent, rich, spontaneous communication leads to more shared knowledge” (Espinosa et al.2002b, p. 430). An obvious occurrence of rich and spontaneous is face-to-face, for example during lunch (Espinosa et al.2002a). Distributed teams can not develop shared mental models as easily as co-located teams because they lack this informal and ad-hoc communication through their separation from their collaborators by time, place or even both (Bass2006Espinosa et al.20062001).

Espinosa et al. (2002b2007a) showed judgement that the coordination of tasks across sites is more difficult and takes more time than within a single site. This relates to the fact that coordination is negatively correlated to software development time (Espinosa et al.2002b).

If there is a lack of familiarity with the other sites’ context and colleagues, geographical distance has negative effects. Prior knowledge can mitigate these negative effects of distance and improve the quality of communication and the relationships between team members (Espinosa2001Espinosa et al.2001). Although Espevik et al. (2011) did not research in the context of software development their findings are relevant. They demonstrated in a study that ”familiar teams outperformed unfamiliar teams on all outcome measures” (Espevik et al.2011, p. 631) because of a better developed SMM. Familiar teams are aware of the informational needs of their colleagues. Thereby, they push more information during high workload-phases to their colleagues and focus on the relevant information.

If there is no prior knowledge, these negative effects can be offset with ”knowledge-based mechanisms such as shared knowledge about team members and presence awareness” (Espinosa et al.2007a, pp. 158). According to Espinosa (2001) team-related SMM is less important in asynchronous teams compared to real-time teams. But Espinosa (2001) was not conducted under the aspect of a distributed team. Espinosa et al. (2007a) proved the benefit of team familiarity in distributed software teams. Furthermore, they concluded that the team-related SMM is more important for cross-site working teams whereas the task-related SMM is more important for members within a site. The team-related SMM becomes more important as larger or more distributed the organisation is (Espinosa et al.2007b). In contrast, task-related SMM becomes less important if the tasks are more structurally complex (Espinosa et al.2007b).

Bass (2006) researched how GSD projects can be monitored via SMM. He asserted that artifact driven monitoring does not ensure that a project stays on track. He also proposes that agile development is well suited for creating a common mental model within a team (Bass2006). Yu & Petter (2014) pursued a similar direction. They examined the contribution of artifacts from agile development to shared mental models.

Levesque et al. (2001) have made some points on the divergence of shared mental models in software development teams. Divergence is functional. This means team members do not need to know the exact implementation of others’ work. They should think of it as a black box with a clear defined interface. Putting the pieces together the system works if each member has met the specification. This fitting can be ensured in two ways, (a) ”team members have a shared mental model of the sub-product deliverables and how they fit together into a coherent whole” or (b) ”have a chief programmer or project manager who understands how the pieces work together” (Levesque et al.2001, p. 141).

5  Discussion

5.1  Main Findings

RQ1: Do and if yes, how do shared mental models impact large-scale software development with multiple teams? If yes, are there limitations?

There is no empirical evidence showing that shared mental models impact multiple teams. The problem is the unit of analysis. Authors like Espinosa have attributed to large-scale software development but in the terms of a large team and not multiple teams. Therefore, in the following it will be supposed that large-scale software development does not have to be conducted within a MTS necessarily.

Shared mental models do not only influence co-located teams (Cannon-Bowers et al.1993) but also asynchronous and distributed teams (Espinosa et al.2002a). The effect does not have to be necessarily positive, it can also be detrimental if the SMM are inaccurate (Espinosa et al.2001).

But the effects on co-located and distributed teams differ. Large-scale software development has three main barriers: geographical dispersion, time separation and cultural differences. These barriers make coordination and communication challenging. They also change the way how SMM are built and how teams can benefit of them.

The most obvious observation is that co-located teams can do the same work in less time than distributed teams keeping the variables (size of teams, complexity of project) the same. This can be explained, that in a co-located setting the building of a SMM does not nearly need as much as attention as in a distributed setting because it happens implicitly as a side effect, for example through informal talks at the coffee machine. In a distributed setting building a common understanding among team members, which can only communicate asynchronously and not by face-to-face, requires a lot of effort.

An educated guess is that shared mental models also impact large-scale software development with multiple teams. Additionally to the previously described barriers and possible problems, the team barrier is added. But nevertheless it is important to communicate and coordinate over teams and that is what shared mental models are thought for. A practical example to bridge the gap is the usage of proxies or liaisons, which specify the same concept with different names. A proxy is someone who for example attends the standup meetings of different connected teams to answer and ask questions. Thus, the shared mental models of a team gets synchronized.

An important and general statement on SMM can be formulated by using evidence provided by Espinosa et al. (2007a): The larger or more distributed the organisation becomes, the more important gets team-related SMM. Complex projects get partitioned and de-coupled into parts that are handled by teams. Each team again splits the parts over its members. Therefore it is important to know who is responsible for which part and who have knowledge on a domain. The importance of task-related SMM declines but does not fully disappear because team members need to know the interfaces but they do not need knowledge on how a component works.

RQ2: How can shared mental models be scaled up in large-scale software development?

The methods and tools enabling shared mental models to scale up are structured by initial phase, organizational setup and general infrastructure.

Initial phase. In large-scale software development it will mostly happen that the developers do not know each other. This is a disadvantage, as they need to get used to skills and weaknesses of each. Due to this Bass (2006) proposes to keep teams as persistent as possible. If this is not possible, face-to-face meetings in the beginning are suggested to remove barriers creating a successful team (Espinosa et al.2001). This is in line with the research on social network distance (Graham et al.2004). But there is a lack in research how to deal with situations where a face-to-face meeting is not possible.

Organizational Setup. In big and complex projects the work must be decoupled (Herbsleb & Mockus2001b). The dependencies between teams in distributed settings need to be minimized. This results in a decrease of effort in coordination and communication between teams. Additional research proposes to install a ”proxy” or ”liaison” if development is performed in a distributed manner (Baffin et al.2001Sutherland et al.2009). A proxy is a team member of each team; together with other proxies they synchronize and therefore share SMM among teams. This is especially useful as an organization grows.

General Infrastructure. An adequate infrastructure helps to build and scale shared mental models. For example, a minimum requirement for distributed work is a sufficient internet connection. This may sound trivial and not relevant for SMM but without a sufficient internet connection coordination and communication are hardly possible. Infrastructure also contains tools, which need to be established among the whole organisation. For example instant messaging and high quality video chats can be a simple replacement of face-to-face chats. Tools to discover who has which experience aims at supporting required team-related SMM and shared calendars ease the coordination.

RQ3: How can shared mental models be used to validate, monitor and explain the performance of teams in large-scale software development?

Some authors expect SMM to converge over time. Levesque et al. (2001) suggests the opposite. Especially teams which are temporary established for developing software, the team members specialise over time and thereby their SMM will diverge. Observing the problem more detailed, leads to the conclusion that SMM in large-scale software development do diverge and converge simultaneously over time.

For example, task-related shared mental models have a lower importance in a distributed setting but are nevertheless relevant for knowledge on interfaces. This common understanding should increase over time because the interfaces need to fit in the end. In contrary, team-related SMM might be relevant in the beginning to create an architecture and start working. But as everyone specialises and focus more and more on his or her individual task, this SMM might decrease.

This is related to the research question because well-balanced shared mental models could lead to an increase in performance. Obviously, it would not harm if the SMM over the teams would be perfectly developed and accurate. But for what costs? Businesses will ask for the best or most reasonable performance with the least costs. To do so, an ongoing monitoring of the development teams’ mental models is proposed (Bass2006). A possibility is to survey team members with online questionnaires twice a month, evaluate the answers and use the answers to execute measures correcting identified weaknesses.

5.2  Implications for Future Research

This work leaves questions, which should be empirically answered. There is a lot of space for potential research that can be conducted.

Levesque et al. (2001) has found divergence of shared mental models among temporal limited projects in software development teams. Assuming that team-related shared mental models are more important in a distributed setting, this leads to the question to what extent the development of SMM is necessary and adequate.

SMM are developed through interaction between team members. If face-to-face contact is limited, mostly asynchronous, technological communication tools mediate this interaction. The different tools and approaches need to get reviewed empirically. Questions would be how they differ in efficiency, if they are only applicable in special environments and what type of SMM do they influence.

This leads to the fundamental question if team members do need to have face-to-face contact or if it can be totally replaced. Espinosa et al. (2007a) found that geographic dispersion has negative effects on SMM. But they also conclude that some effects can be offset with special mechanisms. A question every manager is interested would be: ”Can face-to-face meetings in software development be totally replaced? If yes, how? If not, to what extent can it be replaced and to what extent is it necessary?”. An educated guess is that social network distance could outplay geographic distance.

For ongoing projects the monitoring and prediction via SMM is interesting. Shared mental models reflect the knowledge within each team member of the project regarding to team and task. If it is wrong or not well developed enough, it is likely that the project will fail. By developing measures to detect incorrect and misleading shared mental models projects may succeed.

Originally, this paper also wanted to address shared mental models in large-scale software development within multiple teams. This can not be done as there is no empirical work handling this topic. Therefore, it is proposed that interested researcher step into. They could investigate how and if the measures of single teams can be scaled up to multiple teams. Also the question how and which shared mental models do need to scaled up is relevant. And finally, one could research the previously discussed role of a proxy and its effect on shared mental models in a multi-team system.

Ultimately, a long-term goal should be a handbook for practitioner. It should derive practical recommendations from proven research for people in businesses by considering different setups and circumstances. Also, it could propose technology to mediate the development of SMM.

5.3  Limitations

This literature review does not suggest being a exhaustive, comprehensive, cumulative review. It aims to reflect the current state of research in this topic. Due to its nature, this review is limited in several dimensions.

Search. The restricted time frame only allows to search a selected number of databases. Only looking up 50 results per search on Google Scholar is an obvious but necessary limitation to deal with the huge amount of literature found in it.

Methodology. A systematic literature review (SLR) also has potential limitations. As no primary research is conducted, it relies on previously published articles and especially their correctness. Because of the same reason, no evidence for statements can be made. Another factor is the reliance on the appropriateness of the search method (Webster) and criteria (e.g. only English literature which is available online).

Results. The quality and informative value of a SLR highly depends on the literature, which was found. This is probably the biggest problem and also chance of this SLR. There is no research in the context of software development in multiple teams. There is some literature in the context of global software development. But this research was mostly conducted by Espinosa et al. As this may be seen as a disadvantage, there is plenty of room for future research.

6  Conclusion

Global software development heavily influences the performance of software development. Shared mental models are a way to describe influences and to explain the outcomes. Unfortunately, shared mental models in this context are not very well researched. Going one step further, the effect of a multi-team system in software development on shared mental models is not researched at all.

Nevertheless, some statements of existing literature can be derived to support these settings. For example, the larger or more distributed an organization is, the less relevant are task-related shared mental models and the more important are team-related shared mental models. If this knowledge is combined with the concept of de-coupling, this could be the fundamental of an approach to deal with multiple teams in software development.

This work should motivate researchers to step into and perform new research. Therefore, a lot of possible fields are mentioned within this work. It seems to be inexplicable why such a topic is at low level of research. Empirical findings will have a big impact on the world of businesses as they would substantially benefit. High skilled labour is expensive. If it can be utilized more and if outsourcing or geographical dispersion is less difficult, companies could significantly increase their performance.


  1. Badke-Schaub, P., Neumann, A., Lauche, K., & Mohammed, S. (2007). Mental models in design teams: a valid approach to performance in design collaboration? CoDesign, 3(1), 5–20.
  2. Baffin, R. D., Crocker, R., Kreidler, J., & Subramanian, K. (2001). Leveraging resources in global software development. IEEE Software, 18(2), 70–77.
  3. Bass, M. (2006). Monitoring GSD Projects via Shared Mental Models : A Suggested Approach. GSD ’06 Proceedings of the 2006 international workshop on Global software development for the practitioner, (pp. 34–37).
  4. Cannon-Bowers, J. A., Salas, E., & Converse, S. (1993). Shared mental models in expert team decision-making. In Individual and Group Decision Making: Current Issues (pp. 221–246). N. John Castellan.
  5. Carpenter, S., Delugach, H. S., Etzkorn, L. H., Utley, D. R., Farrington, P. a., & Fortune, J. L. (2008). Studying team shared mental models. Systems Engineering, 6541, 41–48.
  6. Cramton, C. D. (2001). The Mutual Knowledge Problem and Its Consequences for Dispersed Collaboration. Organization Science, 12(3), 346–371.
  7. Espevik, R., Johnsen, B. r. H., & Eid, J. (2011). Communication and Performance in Co-Located and Distributed Teams: An Issue of Shared Mental Models of Team Members? Military Psychology, 23(6), 616–638.
  8. Espinosa, J. A. (2001). Shared Mental Models: Accuracy and Visual Representation. AMCIS 2001 Proceedings, (pp. 2102–2107).
  9. Espinosa, J. A., Carley, K. M., Kraut, R., Lerch, J., & Fussell, S. R. (2002a). The Effect of Task Knowledge Similarity and Distribution on Asynchronous Team Coordination and Performance: Empircal Evidence from Decision Teams. Cognitive Research (CORE) Conference, (9812123).
  10. Espinosa, J. A., DeLone, W., & Lee, G. (2006). Global boundaries, task processes and IS project success: a field study. Information Technology & People, 19(4), 345–370.
  11. Espinosa, J. A., Kraut, R., Lerch, J., Slaughter, S., Herbsleb, J. D., & Mockus, A. (2001). Shared Mental Models and Coordination in Large-Scale, Distributed Software Development. ICIS 2001 Proceedings.
  12. Espinosa, J. A., Kraut, R., Slaughter, S., & Lerch, J. (2002b). Shared Mental Models , Familiarity , and Coordination : A Multi-Method Study of Distributed Software Teams. In ICIS 2002 Proceedings.
  13. Espinosa, J. A., Slaughter, S., Kraut, R., & Herbsleb, J. D. (2007a). Team Knowledge and Coordination in Distributed Software Geographically Development. Journal of Management Information System, 24(1), 135–169.
  14. Espinosa, J. A., Slaughter, S. a., Kraut, R. E., & Herbsleb, J. D. (2007b). Familiarity, Complexity, and Team Performance in Geographically Distributed Software Development. Organization Science, 18(4), 613–630.
  15. Graham, J., Schneider, M., Bauer, a., Bessiere, K., & Gonzalez, C. (2004). Shared Mental Models in Military Command and Control Organizations: Effect of Social Network Distance. Proceedings of the Human Factors and Ergonomics Society Annual Meeting, 48(3), 509–512.
  16. Herbsleb, J. D. & Mockus, A. (2001a). An empirical study of global software development: distance and speed. International Conference on Software Engineering, 23, 81–90.
  17. Herbsleb, J. D. & Mockus, A. (2001b). Challenges of global software development. Software Metrics Symposium, 7, 182–184.
  18. Herbsleb, J. D. & Moitra, D. (2001). Global software development. IEEE software, 18(2), 16–20.
  19. Johnson, T. E., Lee, Y., Lee, M., O’Connor, D. L., Khalil, M. K., & Huang, X. (2007). Measuring Sharedness of Team-Related Knowledge: Design and Validation of a Shared Mental Model Instrument.
  20. Jonker, C., Van Riemsdijk, M. B., & Vermeulen, B. (2011). Shared Mental Models. COIN 2010 International Workshops, (Section 2), 132–151.
  21. Klimoski, R. & Mohammed, S. (1994). Team Mental Model: Construct or Metaphor?
  22. Langan-Fox, J., Anglim, J., & Wilson, J. R. (2004). Mental Models, Team Mental Models, and Performance: Process, Development, and Future Directions. Human Factors and Ergonomics in Manufacturing, 14(4), 331–352.
  23. Levesque, L. L., Wilson, J. M., & Wholey, D. R. (2001). Cognitive divergence and shared mental models in software development project teams. Journal of Organizational Behavior, 22(Shared Cognition), 135–144.
  24. Lim, B. C. & Klein, K. J. (2006). Team mental models and team performance: A field study of the effects of team mental model similarity and accuracy. Journal of Organizational Behavior, 27, 403–418.
  25. Marks, M. A., Zaccaro, S. J., & Mathieu, J. E. (2000). Performance implications of leader briefings and team-interaction training for team adaptation to novel environments. The Journal of applied psychology, 85(6), 971–986.
  26. Mathieu, J. E., Heffner, T. S., Goodwin, G. F., Cannon-Bowers, J. a., & Salas, E. (2005). Scaling the quality of teammates’ mental models: Equifinality and normative comparisons. Journal of Organizational Behavior, 26(1), 37–56.
  27. Mathieu, J. E., Heffner, T. S., Goodwin, G. F., Hobson, K., Ivory, K., Trip, M., & Windefelder, N. (2000). The Influence of Shared Mental Models on Team Process and Performance. Journal of Applied Psychology, 85(2), 273–283.
  28. Mathieu, J. E., Marks, M. A., & Zaccaro, S. (2001). Multiteam Systems. In Handbook of Industrial, Work & Organizational Psychology: Volume 2. SAGE Publications Ltd.
  29. Mathieu, J. E., Maynard, M. T., Rapp, T., & Gilson, L. (2008). Team Effectiveness 1997-2007: A Review of Recent Advancements and a Glimpse Into the Future. Journal of Management, 34(3), 410–476.
  30. Maynard, M. T. & Gilson, L. (2013). The Role of Shared Mental Model Development in Understanding Virtual Team Effectiveness. Group & Organization Management, 39(1), 3–32.
  31. McComb, S. A. (2007). Mental Model Convergence: The Shift from being an Individual to being a Team Member. Research in Multi-Level Issues, 6, 482.
  32. Mohammed, S., Ferzandi, L., & Hamilton, K. (2010). Metaphor No More: A 15-Year Review of the Team Mental Model Construct.
  33. Resick, C. J., Dickson, M. W., Mitchelson, J. K., Allison, L. K., & Clark, M. A. (2010). Team composition, cognition, and effectiveness: Examining mental model similarity and accuracy.
  34. Salas, E., Sims, D. E., & Burke, C. S. (2005). Is there a ”Big Five” in Teamwork? Small Group Research, 36, 555–599.
  35. Scheerer, A., Hildenbrand, T., & Kude, T. (2014). Coordination in large-scale agile software development: A multiteam systems perspective. Proceedings of the Annual Hawaii International Conference on System Sciences, (pp. 4780–4788).
  36. Scozzi, B., Crowston, K., Eseryel, U. Y., & Li, Q. (2008). Shared Mental Models among Open Source Software Developers. Hawaii International Conference on System Sciences, (pp. 1–10).
  37. Stagl, K. C., Salas, E., Rosen, M. A., Priest, H. A., Burke, C. S., Goodwin, G. F., Johnston, J. H., Stagl, K. C., Salas, E., & Rosen, M. A. (2015). Distributed Team Performance: A Multi-Level Review of Distribution, Demography, and Decision Making. Research in Multi-Level Issues, 6, 11–58.
  38. Standifer, R. & Bluedorn, A. (2006). Alliance management teams and entrainment: Sharing temporal mental models. Human Relations, 59(7), 903–927.
  39. Stout, R. J., Cannon-Bowers, J. A., Salas, E., & Milanovich, D. M. (1999). Planning, Shared Mental Models, and Coordinated Performance: An Empirical Link Is Established. Human Factors: The Journal of the Human Factors and Ergonomics Society, 41(1), 61–71.
  40. Sutherland, J., Schoonheim, G., Kumar, N., Pandey, V., & Vishal, S. (2009). Fully Distributed Scrum Linear Scalability of Production between San Francisco and India. Agile Conference, (pp. 277–282).
  41. van den Bossche, P., Gijselaers, W., Segers, M., Woltjer, G., & Kirschner, P. (2011). Team learning: Building shared mental models. Instructional Science, 39(3), 283–301.
  42. Webster, J. & Watson, R. T. (2002). Analyzing the Past to Prepare for the Future: Writing a Literature Review. MIS Quarterly, 26(2), xiii – xxiii.
  43. Yu, X. & Petter, S. (2014). Understanding agile software development practices using shared mental models theory. Information and Software Technology, 56(8), 911–921.

Leave a Comment

comments powered by Disqus