Carlile, P. R. (2002). A pragmatic view of knowledge and boundaries: Boundary objects in new product development. Organization science, 13(4), 442-455.
Through field work examining new product development within an automobile parts company, Carlile discusses the problems that arise during knowledge sharing across boundaries. These boundaries are between units where there are different knowledge bases, in this case functional units at the company. He explains how what makes problem solving efficient within functions creates issues in problem solving across boundaries. Which in terms of his pragmatic view of “knowledge in practice” pertain to knowledge as being localized, embedded, and invested within functional units. It is though understanding those characteristics that one can see how three types of knowledge boundaries arise: syntactic, semantic, and pragmatic. Through vignettes he ties the types of knowledge boundaries—syntactic, sematic, and pragmatic—to the types of boundary objects that are needed to traverse those boundaries and finally to the characteristics of successful boundary objects.
First, he begins by describing the types of knowledge boundaries. The syntactic approach to knowledge boundaries refers to different functional units needing to have a common language to communicate. During the explanation he describes how for a long time the dominant approach to addressing boundary spanning in organizational theory and new product development research was to make more information available. However, this approach alone did not address issues of novelty that arise from differences in interpretation for what is needed for the task. As an aside, in the article while he mentions many authors who have studied semantic boundaries, I also thought of the work by Boland and Tenkasi  describing the importance of perspective making and perspective taking as being at the semantic boundary. While the semantic approach focuses on “differences in kind” it does not address issues of dependency which leads to the third type of boundary, pragmatic. At the pragmatic boundary difference, novelty, and dependence are all present and what is needed is transformation of knowledge across boundaries.
By understanding the three boundaries, one can then understand how the characteristics of “knowledge in practice”—localized, embedded, and invested – contribute to each of the knowledge boundaries. Localism refers to the how problems are solved for a given practice. Efficiency in an organization comes from developing knowledge bases that help address common problems functional units face. However, this leads to practices that are not the same across functional units complicating communication across the units, a syntactic boundary. The next characteristic, embedded, relates to how it is difficult to express knowledge outside of practice or the tacitness of knowledge and therefore how people from different knowledge domains will have difficulty communicating knowledge outside their discipline. This is the semantic boundary. Lastly, because knowledge is invested in practice, people want to do things according to what they already know, but faced with a dependency on knowledge from another group and the novelty of the situation, for success they must be willing to transform their existing knowledge, thus the pragmatic boundary.
What is most interesting however is Carlile’s study of boundary objects in new product development and through relating Star’s  four categories of boundary objects to how they can be used at the knowledge boundaries he describes the characteristics of the types boundary objects. Star in studying heterogeneous problems solving enumerated four types of boundary objects: repositories, standardized forms and methods, objects of models, and maps of boundaries. Repositories are stores of information that have common meaning across functional units. Standardized forms and methods provide a shared approach for addressing problems across boundaries. Objects or models are detailed representations that different groups can use during problem solving and maps of boundaries express the dependencies across groups. In his explication of the relationship between knowledge boundaries, boundary objects, and characteristics of boundary objects, Carlile combines objects and methods and maps of boundaries. It is important to understand that the types of boundary objects build upon each other subsuming characteristics of less versatile boundary objects in this order: repositories; standardized forms and methods; and objects, models, and maps.
To understand the characteristics of boundary objects that can be used at the different boundaries, one can look at the purposes of each type of boundary object. Repositories contain information that can be used across groups to solve problems. In this way they address the syntactic boundary and must be adequately representative among units to be useful. Standardized forms and methods help define a shared approach to problem, to be useful they must express differences and dependencies. So, while they represent information much as repositories do to address communication issues, they also enable learning between groups to occur by helping identify differences and dependencies. However, to address those dependencies and the consequences thereof the pragmatic boundary must be addressed, which requires transformation of knowledge. Objects, models, and maps are the only category of boundary object that can satisfy that purpose by supporting representativeness for communication, learning through definition of differences and dependencies, and providing a process for helping different groups mutually transform knowledge.
 Boland, R. J., & Tenkasi, R. V. (1995). Perspective making and perspective taking in communities of knowing. Organization science, 6(4), 350-372.
 Star, S. L. (1989). The Structure of 111 «Structured Solutions: Boundary Objects and Heterogeneous Distributed Problem Solving. Distributed Artifcial Intelligence, 2, 37-54.