When Data Sharing Becomes Institutionalized: Best Practices in Local Government GI Relationships

When Data Sharing Becomes Institutionalized: Best Practices in Local Government GI Relationships

(Version 10/17/05)

David Tulloch and Francis Harvey

Introduction

As agencies function within the context of an increasingly digitally interconnected world, the ability for information to flow from a source to many other unanticipated individuals has been accelerated, especially through networks of other interested individuals (Watts 2003).  The response to this phenomenon of many people involved with local government GIS continues to be apprehension and resistance resulting in data sharing networks that exhibit a variety of patterns.

Whether as a means of data dissemination or acquisition, data sharing has become an essential element of local government GIS activities.  Despite the prominence of this activity and its centrality to the day-to-day function of many local government systems, decisions about these activities are rarely made following a thorough consideration of the organizational and political complexity.  This paper examines the techniques used in a variety of communities and extracts from those a summary of best practices for data sharing in local government.  These best practices are offered as guidelines borne out by the outcomes experienced under a variety of circumstances, but as any best practices they remain admittedly suggestive more than absolute.

The importance of existing political, institutional, professional, and legislative relationships is emphasized in this paper. These relationships largely determine the geographic information activities at local government agencies. Sharing and coordination are to a large extent informal activities that correspond to these relationships. Formal arrangements are taken only at the final stages of establishing sharing or cooperation agreements, basically they manifest themselves at the conclusion to satisfy internal procedures and/or legal requirements. This understanding poses an interesting conundrum for developing local government participation in the NSDI.

Given the importance of informal agreements the primary best practicesí question arises: How do local governments successfully share geospatial data and coordinate geographic information activities?  Improved understanding of the issues, potentials, and problems that local governments face is key to establishing programs and policy support to address local government needs and develop better opportunities for them to participate in geographic information infrastructures connected to any SDI.

BACKGROUND

Data sharing is much more than a simple technique for accessing GI, it is an unique form of institutional interrelationship that is necessary for unleashing many of the most important benefits of public GIS applications (Tulloch and Epstein 2002; General Accounting Office 2003).   Perhaps the grandest potential public GIS applications would be a SDI which can only be accomplished with significant advances in data sharing practices and a fundamental improvement in the understanding of the ways in which these practices are employed.

Many of the larger GIS initiatives (both formal and informal) are based on the notion of many data producers working collectively to develop large volumes of data combined to describe relatively large areas with an expectation that those data will be widely available to potential users.  Included among these initiatives are a National Spatial Data Infrastructure (NSDI), a Global Spatial Data Infrastructure (GSDI), INSPIRE, and the Digital Earth initiative.  Each of these would require a large volume of existing data and substantial accessibility to the data.

The nature of data access in the US  can be quite complex with a variety of state and federal laws governing public organizations, and constantly changing markets governing commercial distribution.  Initially, access to data was allowed largely under ad hoc arrangements, but as GIS institutions continue to grow more organizations are developing policies and practices that formalize the ways in which their geospatial data can be used and disseminated.  The different forms of access have come to be generically called ìdata sharingî.

ìSpatial data sharing is defined as the (normally) electronic transfer of spatial data/information between two or more organizational units where there is independence between the holder of the data and the prospective user.  The transfer may be in the form of periodic bulk transfers, routine daily transfers, or on-line access driven by individual transactions.  The participants may be separate organizations or may be departments within the same organization.  For our purposes, the distinguishing characteristic of spatial data sharing is that there be an armís-length exchange or transfer.î  (Calkins and Weatherbe 1995, 66)

The arguments for data sharing are many, but a particularly strong one is that it allows data, otherwise isolated and underutilized, to be used repeatedly for many purposes, thus increasing their value without increasing their cost (Maackey 1982).

"The value and social utility of geographic information comes from its use.  Sharing of geographic information is important because the more it is shared, the more it is used, and the greater becomes societyís ability to evaluate and address the wide range of pressing problems to which such information may be appliedî  (Onsrud and Rushton 1995, xiv)

Despite the apparent values of data sharing, there are many hurdles that still serve as barriers to sharing from Craigís (1995) broadly presented ìinstitutional inertiaî to Kevanyís (1995) detailed list of 30 geographic information factors.  Most recently, Nedovic-Budic, Pinto and Warnecke (2004) have examined the mechanisms and means through which data sharing currently occurs with a particular eye to the motivations behind sharing.

The identification of best practices has become a common approach in business and public administration.  Many have traced best practices to the popular press book, In Search of Excellence (Peters and Waterman 1982).  However, the approach has grown into tool for studying a variety situations and identifying successful ways of addressing those situations.  Best practices reporting has become a particularly useful tool for capturing options available to managers or decision-makers, and identifying appropriate applications for these practices.

METHODOLOGY

Since much of the methodology of this paper has been previously described (Harvey and Tulloch 2003, 2005) we will only provide an abbreviated version for background.  The project used two primary methods of data collection:

a)                  Interview-based local government data sharing case studies

b)                  Multi-modal local government data sharing surveys.

The methodology is divided into three distinct phases. In phase one we examined the experiences of data sharing participants with local sharing and coordination arrangements. Objectives of this phase were the determination of strategies and key issues in demonstration projects that have worked and find out what strategies failed and the problems that were resolved and continue to have impacts.

In phase two, interviews were conducted at a variety of local governments to determine how they share geospatial data and coordinate geographic information activities. These interviews were held with numerous local government agencies.  Interviews throughout the entire data sharing network provided substantive insights into the contexts and practices of sharing and coordination in situ. The local governments whose GI sharing practices were studied were selected with great care.  We included large and small local governments representing a diverse geography and worked to ensure that legislative and political differences at the regional and state level could be identified for comparison.

The in-depth interview case studies were held in 6 areas around the US.  These in-depth interviews of multiple participants within a single data sharing network provided great detail complemented by the more tightly-focused polling of local governments in the surveys.   For each series of case study interviews, the authors identified a significant spatial data sharing network that relied heavily on local government data.  Several of the case studies were chosen based on their use as FGDC demonstration projects.  All of the examples focused on local government data that was shared in a variety of ways with different users or organizations.  The interview processes generally began with an extended interview with one of the networksí most central GIS coordinator and then moved on to interviews with other data providers and data users within that network. 

RESEARCH FINDINGS

This section identifies specific repeated lessons from the surveys, interviews and case studies. We identify those outcomes most broadly relevant to local government GIS data practices.  The findings described in this section are derived directly from the data collection processes described in the methodology.  Some are based on experiences many different organizations, while others are derived from the experiences of the most successful data sharers.   

These findings reveal some recurring keys to success in communities where data sharing has thrived, while also illustrating some complications that could confuse a causal observer.  Based on these findings and related results of this research project (Tulloch and Harvey 2003, 2005), the authors have developed a list of data sharing keys and contradictions that reflect data sharing best practices.

As is frequently pointed out by critics of ìbest practicesî as a form of knowledge management (e.g., Quinn 2001), no one set of best practices can exist for a complex, context-dependent, network-based activity like data sharing.  But this list provides a quick summary of some better practices and important contradictions that emerge from the project described in this paper.

 

 

To designate the manner in which each finding contributes to a best practice described later in the paper, the text in this section is marked with notations indicating which key or contradiction (e.g., [K2] or [C4]) the finding most closely supports. The first part of the section describes different models for understanding local government data sharing arrangements. After an evaluation of underlying issues impacting these models, the second part presents a set of best practices.

Different Structures Require Different Responses

The context within a local government for data sharing should tie into the broad issues within which an SDI, or a larger network of data users, is situated. Different types of arrangements respond to the existing context. Certainly, for a study of best practices, this requires understanding the activities of local government agencies.  We have identified some distinct data sharing structures that form the context within which sharing often occurs (Harvey and Tulloch 2005).  Understanding the contextual structure helps clarify concerns and responses described by local government agencies [K1].  For example, a centralized ìhub and spokeî model that relies on a strong coordinator (Figure 1) might create situations for which formal arrangements and fully compliant metadata are less important.  But unexpected changes in the model, particularly the sudden departure of the central figure, can create institutional chaos.

Figure 1 ñ A conceptual diagram of the centralized network (or GIS Dictator) model of data sharing.  This can be a very attractive model when the central figure is has strong expertise and is generous (a benevolent dictator).  Less considerate or skilled coordinators can quickly turn this model into a plan for disaster.

Figure 2 ñ The sudden departure of a strong centralized coordinator (or GIS dictator) can leave other users in a difficult situation.  They may even be unable to access data that was produced by a third party or decipher cryptic metadata that was considered important when the source could be consulted.

We can contrast the centralized model with that of either the Federation-by-accord or the Federation-by-mandate (Harvey and Tulloch 2005).  These models for federated data sharing arrangements among local governments (Figure 2) reflect either the negotiated agreements that led to the formalization of data sharing arrangements or were stipulated by a government agency. Federated data sharing arrangements are the most common and not all are thoroughly formalized, but are in a state of flux corresponding to changes in institutional, personal, and political arrangements (Figure 3). While this may seem to be a negative point from a policy point of view, practically this flux makes these arrangements enormously flexible and resilient to changes. The few stipulated arrangements exhibit both negative and positive characteristics, depending on the degree of funding for the activities and their role as a ìseedî for additional data sharing arrangements.

Figure 3 ñ A conceptual diagram of the federated network model of data sharing.  As it grows more complex, the federated network can create many redundancies which become very important in difficult times.

The Hidden Costs of Cost Recovery: The Ripple Effect (Or the Cost Recovery Cascade)

An important issue in all data sharing is financing and conflicts related to cost recovery strategies. Much has been written about the nature of the debate within the GIS literature regarding the appropriateness (or inappropriateness) of policies promoting cost recovery techniques which effectively charge a fee for access to public data (Epstein, Onsrud). 

This research isolated a pattern for data sharing arrangements involving prohibitory cost recovery policies, not previously described in the literature, caused by a cost recovery policy.  This pattern, which weíll call the Ripple Effect, was found, for example, in a community where one of the most important datasets (parcels) was protected by a thoroughly developed and rigorously enforced policy that charged a significant fee for access and limited redistribution.  The policy included specific language that restricted the ways that these data could be incorporated into other forms of analysis that might eventually be released publicly.  As it played out, this policy had three different levels of impacts, many of which were not immediately obvious, but all of which demonstrate the ways that such a policy stifles activities throughout a GIS community [K3].

The primary (and most discussed) impact of a cost recovery policy is to limit access.  Only those individuals, agencies, and companies that can afford the data can access and use it.  As described in the URISA Journalís Special Issues on Access and Participatory Approaches (Onsrud and Craglia 2003), this raises a number of questions about democratic participation in public decisions using those data as well as concerns about asking members of the public to pay twice for the data.

The secondary impact was to limit or openly prevent the distribution and development of other geospatial data.  In this example, a licensed user of the data used the parcel data as an important input for the development of its land use/cover datasets.  However, the boundaries of the land use/cover were determined to reveal too many details of the licensed parcel data, and were thus could only be distributed to other organizations possessing the parcel data license.  A creative solution was found by which the agency produced a degraded raster version of the data which could be released, but it meant that the public was receiving different information than was actually used for public decisions by the agency. 

There also exists a tertiary level of impact with significance for the GIS industry.  Many of the surrounding municipalities have found it necessary to buy into the license for the previously mentioned data (either for the parcels, or for the spin-off datasets).  Those municipalities require bidders on a variety of GIS and engineering projects to acquire a data license so that they can use the municipal datasets.  As a result, a local firm with an existing license (affordable when working on multiple licensed data projects across the region) has an upper hand over firms from out of town when competing for a local contract.  Over time, this policy could impede competition and create an ad hoc good-old-boys network that reinforces the policy.

Concern about the ripple effect comes in the combined impact of the hidden ripples (Figure 4).  It probably wouldnít surprise most to hear that access restrictions in a heavy centralized model could impact GIS users throughout the entire community.  But, the example described above comes from a federated network where one actor significantly impacts many others within the network. 

Figure 4 ñ The Ripple Effect of Limited Access illustrates how a simple decision to limit access through the use of user fees can have far reaching effect.

Costs of Enforcement

A repeated theme surfacing in interviews was the difficulty in constructing and enforcing a policy that could appropriately limit the redistribution of public geospatial data.  Discussions with agencies engaged in either highly restrictive policies or price schemes, very often lead to descriptions of complicated processes relying significantly on the work of a legal staff, diligent efforts to sniff out data poachers violating license agreements, or processes through which only known and trusted entities could access public data.  In some cases the fees charged didnít seem sufficient to cover the costs of establishing and enforcing the restrictions, in others the enforcement was minimized to the status of being a major distraction [K3, K7]. 

The Benefits of Sharing

Another recurring theme was the value that many communities saw in sharing data.  One interviewee was particularly ecstatic in stressing how important it was to simply serving the community.  However, others saw significant political gains resulting from their open data sharing policies.  The ability to leverage the ìFreeî data that they shared was seen as a politically expedient reason to continue sharing their data [C1].  At times this was done with a clear expectation of getting something in return, whether that was data from other organizations or simply a general recognition of the data development and dissemination effort which might lead to larger budget allocations. 

The political expediency of data sharing is dependent upon working to insure that the agency of origination receives full and proper credit [C2] and careful and explicit communication about the costs of data development [C3].  One of the easiest ways to get credit is to make sure that data is accompanied by metadata that provides that information.  At a time of budget cuts, a well-informed user community can be a much more important asset than few thousand dollars in data access fees.  By sharing public geospatial data, interviewed data developers are able to impact other, sometimes critically important, technology applications [C5].  Many systems managers were very aware of the importance of their user community and worked to promote those applications in other departments (e.g., a highly visible sheriffís department) more than their own data.

A different approach used by some communities has been to share their data as a way to promote their GIS development staff as a GIS Service Center.  Charging a reasonable fee for services allows the agency to develop a revenue stream that is somewhat independent of budget cycles.  It also allows the agency to develop exceptional examples of how their data can be applied [C4].

Formal/Informal Relationships

The most successful data sharing networks clearly relied on a combination of formal and informal relationships to facilitate the largest distribution of spatial data.  The problem with informal relationships is that sometimes they can be all about ìwho you know,î which involves social concepts including trust (Harvey 2003) [K6].  A frequent complaint associated with formalized networks and sharing arrangements was the difficulty in negotiating and maintaining the arrangements and the rigidity that these arrangements often imposed upon day-to-day functions.  Much of the conflict between formal and informal is also related to the conflicts between administrative and political short-term and long-term conflicts and between personal and institutional goals.

Data Sharing and Metadata

Whether formal or informal, the development and exchange of metadata represented one of the most important practices in determining the success of large scale data sharing activities.  While a variety of practices existed with respect to metadata, it was clear that networks with significant amounts of data sharing relied heavily on a well developed system of documenting data.  Some smaller data sharing networks function without any notable form of metadata, relying instead on verbal communication and interpersonal relationshipsóstrengthening the authority and power of persons controlling information about the data available for sharing.  Larger organizations or networks rely much more heavily on the institutionalization of processes insuring the development of metadata -- whether formal, informal, or FGDC compliant ñ because of the inability to provide support to each individual and organization that tries to use their data [K4].  The other interesting characteristic related to the relationship between data sharing and metadata is that organizations that took the effort to produced relatively detailed metadata almost always had geospatial data of some value.  [K5]

GIS as a Social Phenomenon

Ultimately data sharing networks are made up of people and the relationships between those people.  Any system relying on people and relationships is subject to the complexities of social coordination and the influences guiding or motivating involved individuals. One of the most difficult factors to address or change may be the fundamental beliefs or attitudes of individuals with authority over decisions regarding data sharing [K2].  This research encountered a very wide variety of attitudes regarding data sharing ranging from individuals who felt that free and open data sharing was a moral imperative, to others who viewed it as a major burden of little personal importance, to others who felt obliged to capitalize on their data any way possible (particularly by charging significant access fees).  While some of these perspectives can be explained by a particular context, they are also impacted by the personal attitudes of the individuals involved.

CONCLUSIONS

Best Practice 1: Education

The cases studies made clear the importance of education as the highest priority for data sharing best practices that will make any major SDI initiative a success. Workshops or other educational meetings are one avenue, but written materials for local and regional governments would serve as an invaluable resource. An "SDI Guide to GI Sharing and Coordination" obviously could complement workshops, but more importantly a guidebook would have its own legs and reach people and places that workshops barely scratch.  In particular, the guide should address "data themes" that are most impacted and recommend that they focus on trying to "loosen them up" a little somehow (locally-specific themes like parcels and zoning that would be hardest to develop top-down, but sharing policies often prevent these from being stitched together). Mini-grants to generate contributions would be helpful, this is something that national organizations could coordinate.  Attention to financial and copyright issues are critical. A stronger connection to geospatial activities and resources in national government is also needed.

In terms of guidance for SDI development in the US, the 1997 "Framework Guide" can provide some preliminary ideas.

Research Issues

Regarding research, two separate but related areas should be prioritized. First, a new--and improved--Framework Survey is urgently needed. We have too little information about what is occurring nationwide. Are people involved in building framework layers? Are they participating in the development of the US NSDI? What strategies are people employing locally and regionally? Second, although our research turned up some valuable insights into the processes of data sharing and coordination, we were not able to spend enough time to actually find out how formal approaches interact with existing formal and informal approaches. In the US, the NSDI isn't being built on a blank slate; most SDIs arenít. Every government agency has existing mandates that necessitated data production, sharing, or coordination in the past. We could hypothesize that the largest hindrance for the NSDI at the local level arise from missed chances to piggyback NSDI development on existing government policies and programs. The policy level issues are many times abstract. The rubber hits the road in local government and we still don't understand how GI is practically interwoven in mandated and legislated activities.

Connected to this point, is an equally urgent need to assess the financing models for GI in local governments. We have larger studies that elaborate the capitalization approach for building a national infrastructure, but what about the current financing models? Related questions about costs and copyright also are important.

Organizations and Coalitions

The formal development of a coalition of organizations could help drive home the importance of data sharing for local governments and SDI development. URISA s Summit of Partnership and Collaboration (http://www.urisa.org/FedSummit/Summit.htm) demonstrates that there are a variety of organizations willing to rally around this topic area.  The National Association of County Officials (NACO) and the League of Cities also have GIS related activities that reach local governments. Obviously, other state and national groups should be involved.

Last, small, but symbolically significant, awards for local governments could help boost awareness and make NSDI participation a more prestigious attribute.

Moving On

Best practices of data sharing and coordination involve many aspects of administrative and political activities. The best practices for any particular locality at any particular time are contingent on a number of factors and characteristics. Of all the practices we have identified in this research, a key practice seems to lie in the approach towards data and colleagues. If data sharing and coordination is just about the data, it will be very difficult at best, and may likely not work at all, nor for any length of time. Data sharing and coordination are best understood as part of other activities. Some of these activities require interaction, many others are assisted and promoted by data sharing. Establishing and supporting a social network among colleagues, citizens, and elected officials that supports their interests seems to be critical in all cases.

Like many institutional and social interactions, the critically important process of data sharing will continue mildly confounding but can be advanced by an awareness of the issues relating to data sharing by all involved parties.  The issues described in this paper, like context and attitudes, contribute in different ways placing an unexpected burden on many GI science professionals.  Technology helps overcome some barriers, but also creates new barriers.  Similarly, social networks can be of great assistance, but can also undermine individual efforts.  Perhaps one the greatest long-term steps towards improving these conditions is a more thorough integration of ideas about data sharing, like the keys and contradictions, into educational curricula.

REFERENCES

Azad, B., and L. L. Wiggins. 1995. "Dynamics of Inter-Organizational Geographic Data Sharing: A Conceptual Framework for Research," pages 22-43 in Onsrud and Rushton 1995.

Calkins, H. W., and R. Weatherbe, 1995, ìTaxonomy of spatial Data Sharing,î Chapter 4 in Sharing Geographic Information (Onsrud, H. J. and G. Rushton, Eds.). New Brunswick, NJ: Center for Urban Policy Research, Rutgers, The State University of New Jersey.

Craig, W., and D. Johnson. 1997. ìMaximizing GIS Benefits To Society,î Geo Info Systems. 7 (March) 3: 14-18.

Craig, William, 1995 ìWhy We Canít Share Data: Institutional Inertia,î Chapter 6 in Sharing Geographic Information (Onsrud, H. J. and G. Rushton, Eds.). New Brunswick, NJ: Center for Urban Policy Research, Rutgers, The State University of New Jersey.

Deuker, K.J. and D. Kjerne.  1989.  ìMultipurpose Cadastre Terms and Definitions.î Proceedings of the American Society for Photography and Remote Sensing and American Congress on Surveying and Mapping. Falls Church, VA., p. 12.

Federal Geographic Data Committee, 1995.  Development of a National digital Geospatial Data Framework.  Washington, DC: Federal Geographic Data Committee.

Federal Geographic Data Committee, 1997.  Framework Introduction and Guide.  Washington, DC: Federal Geographic Data Committee.

General Accounting Office, U.S. Congress. House Subcommittee on Technology, Information Policy, Intergovernmental Relations and the Census, Committee on Government Reform (2003). Geographic Information Systems: Challenges to Effective Data Sharing. Testimony by Linda D. Koontz, Director of Information Management Issues, U.S. General Accounting Office to. U. S. Congress, House Subcommittee on Technology, Information Policy, Intergovernmental Relations and the Census, Committee on Government Reform, Washington DC, June 10, 2003. GAO-03-874T.

Harvey, F. and D. Tulloch, 2004, ìHow Do Local Governments Share and Coordinate Geographic Information? Issues in the United States,î 10th EC-GI & GIS Workshop -- ESDI: The State of the Art, Warsaw, Poland, June 23-25, 2004.

Harvey, F., & Tulloch, D. (2003). Building the NSDI at the Base: Establishing Best Sharing and Coordination Practices among Local Governments (Report). Minneapolis, MN and New Brunswick, NJ: University of Minnesota. Retrieved June 2004, from http://www.tc.umn.edu/fharvey/research/fw-comp.pdf.

Harvey, F., 2003, "Developing geographic information infrastructures for local government: The role of trust." Canadian Geographer 47(1): 28-37.

Harvey, F., and D. Tulloch, 2005. ìLocal government data sharing: Evaluating the foundations of spatial data sharing infrastructures,î International Journal of Geographical Information Systems. (Accepted pending minor changes)

Kevany, Michael, 1995 ìA Proposed Structure for Observing Data Sharing,î Chapter 6 in Sharing Geographic Information (Onsrud, H. J. and G. Rushton, Eds.). New Brunswick, NJ: Center for Urban Policy Research, Rutgers, The State University of New Jersey.

Mackaay, E. 1982. Economics of Information and Law. Boston: Kluwer-Nijhoff Publishing.

Mapping Science Committee, 1993. Toward a Coordinated Spatial Data Infrastructure for the Nation. Washington, D. C.: National Academy Press.

Mapping Science Committee, 1997. The Future of Spatial Data and Society: Summary of a Workshop.  Washington, D.C.: National Academy Press.

Mapping Science Committee, 2001. National spatial Data Infrastructure for the Nation. Washington, D. C.: National Academy Press.

Masser, I. 1998. Governments and Geographic Information. London: Taylor and Francis.

National Academy of Public Administration. 1998. Geographic Information for the 21st Century: Building a Strategy for the Nation. A Report by a Panel of the National Academy of Public Administration for the Bureau of Land Management, Forest Service, United States Geological Survey, and National Ocean Service. Washington, D.C: National Academy of Public Administration.

Nedovic-Budic, Z., J. Pinto, and L. Warnecke, 2004.  ìGIS Database Development and Exchange: Interaction Mechanisms and Motivations,î URISA Journal, 16 (1): 15-29.

Onsrud, H. J. and G. Rushton, 1995, ìSharing Geographic Information: An Introduction,î Chapter 4 in Sharing Geographic Information (Onsrud, H. J. and G. Rushton, Eds.). New Brunswick, NJ: Center for Urban Policy Research, Rutgers, The State University of New Jersey.

Onsrud, H. J. and M. Craglia, 2003, Special Issues on Access and Participatory Approaches (APA) Number 1 and 2, URISA Journal 15 (1 & 2).

Patton, M. Q., 2001, ìEvaluation, Knowledge Management, Best Practices, and high Quality Lessons Learned,î American Journal of Evaluation, 22 (3):239-336.

Peters, T. J., and R. H. Waterman, Jr., In Search of Excellence: lessons from Americaís Best-Run Companies, New York: Warner Books, 1982.

Tosta, N., 1999. ìNSDI was supposed to be a verb: A personal perspective on progress in the evolution of the U.S. National Spatial data Infrastructure,î Chapter 2 in B. Gittings (Ed.) Integrating Information Infrastructures with GI Technology: Innovations in GIS 6 Philadelphia; Taylor and Francis.

Tulloch, D. L., and E. Epstein. 2002. "Benefits Of Community MPLIS: Efficiency, Effectiveness, And Equity," Transactions in Geographic Information Systems, 6 (2): 195-212.

Watts, Duncan. 2003. Six Degrees, Norton & Company, New York.