RESEARCH NOTES AND COMMENTARIES FROM CORE TO PERIPHERY AND BACK: A STUDY ON THE DELIBERATE SHAPING OF KNOWLEDGE FLOWS IN INTERFIRM DYADS AND NETWORKS

We study 892 Italian motorcycle industry projects carried out via 184 different buyer–supplier and supplier-supplier relationships to provide evidence on the knowledge dynamics occurring in dyads and networks and to understand the underexplored but important (perhaps even dominant) leading role that some firms play in the evolution of networks and interfirm learning processes. We develop a multiphase model which, from a multilevel perspective addressing different relational subsets, suggests how firms can best organize to generate and exchange knowledge efficiently. We argue that extant theoretical perspectives can profitably draw on our findings to strengthen their dynamic components and help them explain the widely diffused ‘exploring through partner’ strategies more effectively


INTRODUCTION
Both executives and academics have recently identified knowledge as one of the most important factors contributing to firms' ability to achieve sustainable competitive advantage (Dyer and Singh, 1998;Dyer and Nobeoka, 2000;Zhao, Anand, and Mitchell, 2005). At the same time, a growing body of research has come to consider networks as an appropriate unit of analysis (Powell, Koput, and Smith-Doerr, 1996;Kogut, 2000) and to examine the question of where a firm should draw its boundaries (Parmigiani and Mitchell, 2009) on the understanding that 'where boundaries are Keywords: knowledge exchange and creation; interfirm networks; core and peripheral firms; knowledge-enhancing practices; motorcycle industry *Correspondence to: Andrea Lipparini, Department of Management, University of Bologna, Via Capo di Lucca, 34 -Bologna, Italy. E-mail: andrea.lipparini@unibo.it drawn affects the value placed on knowledge and its usefulness' (Argote, McEvily, and Reagans, 2003: 574). This work has shown how knowledge and other critical resources increasingly span firm boundaries so that the advantage of a single firm is often linked to those of the network of ties in which it is embedded (Dyer and Singh, 1998). In particular, research rooted in innovation literature (Powell et al., 1996;Zhao et al., 2005) and in buyer-supplier relationships (Kotabe, Martin, and Domoto, 2003) has emphasized the increasing importance of networking, alliances and idiosyncratic interfirm relations (Martin, Mitchell, and Swaminathan, 1995), and the advantages of adopting a cooperative mode for knowledge creation (Mesquita, Anand, and Brush, 2008). Strategic literature's focus on knowledge creation and management has been expanding in unprecedented fashion, so that the key knowledge-based questions companies face are not only how to organize themselves to exploit their already developed knowledge or capabilities, but also 'how to organize to efficiently generate knowledge and capability' (Nickerson and Zenger, 2004: 617).
Our multiple case, inductive study (Eisenhardt, 1989) attempts to answer the above question. We present a multiphase model of interfirm knowledge creation and exchange across dyads and networks and discuss the important (perhaps even dominant) role that buying firms play in the evolution of networks and of interfirm learning processes, which has hitherto been underexplored in the strategic literature. We posit that networks become viable organizational forms for sharing and generating valuable knowledge when a set of purposefully implemented knowledge-enhancing practices-which we define as distinctive patterns of action and interaction between core firms-'buyers'-and peripheral firms-'suppliers'-that enable the exchange, recombination, and co-creation of specialized knowledge-allow firms to leverage knowledge located beyond their organizational boundaries. 1 In particular, we focus on a total of 892 projects associated with the design and/or manufacture of components for motorbikes and/or scooters in the Italian two-wheeler industry. These projects involved 17 leading motorcycle manufacturers-core buyer firms who were responsible for setting up and subsequently 'orchestrating' dyads and networks (Dhanaraj and Parkhe, 2006)-and their (peripheral) first-tier suppliers.

CONCEPTUAL BACKGROUND
Although prominent firm perspectives have contributed greatly to explain how firms achieve sustainable competitive advantage, they still overlook the fact that, increasingly, even unique resources span firm boundaries and can be selectively tradable through a network of firms (Kogut and Zander, 1992;Gulati, 1999). Even contributions that have significantly extended resource-based reasoning (Barney, 1991;Peteraf, 1993) and are rooted in knowledge and capability-based views in strategy (Kogut and Zander 1992;Conner and Prahalad, 1 Throughout the article, we make use of the core periphery pattern to convey the particular situation in which core firms, which possess prominence gained through relational skills and their central position in network structures, leverage their position to harness the knowledge dispersed among peripheral firms. 1996; Teece, Pisano, and Shuen, 1997;Felin and Hesterly, 2007) share a main focus on intrafirm transfers of existing knowledge (Hansen 1999;Tsai 2001;Reagans and McEvily, 2003;Levin and Cross, 2004) and are less well equipped to analyze knowledge transfer and creation across fully independent entities (which have become significantly more common over a wide range of industries, from automobiles to semiconductors) (Nickerson and Zenger, 2004;Hansen, Mors, and Løvås, 2005;Mesquita et al., 2008). A growing number of empirical studies rooted in innovation literature (Powell et al., 1996;Zhao et al., 2005), and especially those adopting a relational perspective (Dyer and Singh, 1998), explicitly consider the collaborative network, rather than the single firm, as the locus for learning and innovation. In particular, research on buyer-supplier networks in the automobile industry (Dyer, 1996;Osborn and Hagedoorn, 1997;Argote et al., 2003) has shown how the extensive involvement of suppliers in product and process development has given lead assemblers faster product development cycles, lower costs, better end product quality (Womack, Jones, and Roos, 1990;Dyer and Nobeoka, 2000), and faster problem solving (Takeishi, 2001). The idea that networks are critical for knowledge creation and exchange is corroborated by studies conducted from knowledge transfer perspectives (Tsai, 2001;Reagans and McEvily, 2003;Inkpen and Tsang, 2005). Systematic evidence from this research stream indicates that networks with substantial knowledge transfer mechanisms between manufacturers and suppliers out-innovate relational sets with less well-developed knowledge-sharing routines.
Following Kogut (2000), we see networks as the outcomes of generative rules of knowledge coordination, which constitute capabilities (e.g., speedy response to change, resource orchestration, etc.) generating relational rents, which can be subject to private appropriation. Our study focuses on multiple transactions and on those deliberate and purposeful actions designed to facilitate knowledge generation and mobility across networks' boundaries, which are central to their ability to coinnovate and so co-create value (Zhao et al., 2005). While most studies focus on ties with generically defined content, we disentangle the intertemporal processes of knowledge creation and exchange between organizations and adopt a multilevel perspective that addresses three different subsets Strat. Mgmt. J., 35: 578-595 (2014) DOI: 10.1002/smj of relationships-R&D, operations, and other upand downstream activities. We also investigate learning processes and knowledge-related activities involving both sides of the dyad and consider them as both knowledge transmitters and recipients. Although several studies have included bilateral (buyer and supplier) analyses (e.g., Zander and Kogut, 1995;Takeishi, 2001;Kotabe et al., 2003), they have considered knowledge flows only as unidirectional-one unit providing and one receiving-and such perspectives clearly limit our understanding of cases where firms both provide and receive knowledge. Finally, we show how the effective creation and extraction of network value hinges on the purposeful enactment by core firms of practices designed to lead and facilitate knowledge sharing and deployment.

Research design, setting, and sample construction
Our inductive, multiple case study (Eisenhardt, 1989) examines the Italian powered two-wheeler industry, which we deemed appropriate for several reasons. First, product architecture: motorcycles and scooters 2 are made of dozens of modules and subassemblies whose development and manufacture requires diverse and specific knowhow and expertise (Helfat, Lipparini, and Verona, 2011). Second, the level of product innovation involved is appropriate: manufacturers are continually challenged by the introduction of new materials-plastics, titanium, carbon fibers, light alloys, etc.-into frames and components, as well as the need to meet changing environmental regulations (Helfat et al., 2011). Third, the intense industrial process innovations, which require manufacturers to adopt state-of-the-art equipment and R&D practices and to base their manufacturing activities on leading-edge technologies (Gavetti, 2001). Fourth, the size of Italy's domestic market-the largest in Europe in terms of powered two-wheeler production (nearly 55%), registrations (26.7%), and numbers of motorcycles in use (26%)-and its competitive position as Europe's leading exporter, headed only by Japan and Taiwan in world rankings (ACEM -Association des Constructeurs Européens de Motocycles, 2010). Our study included core motorcycle firms (assemblers) that sold products in Italy (regardless of their country of origin) and peripheral firms (suppliers) localized in Italy that specialized in the design and/or manufacture of industry components. Both of these types of firms work within a network structure of non-equity, longterm cooperative buyer/supplier relationships. We selected core firms on the basis of lists supplied by the Italian Ministry of Transportation (from www.trasporti.gov.it) and the trade association (www.ancma.it), and we used Italian market sales performance figures for 2007, the most recent year for which comprehensive statistics were available. We contacted representatives at the full list of 25 core firms specializing in scooter and/or motorcycle design and/or production, and 17-covering nearly 85 percent of Italian market sales-agreed to participate. In selecting peripheral firms, we first used business publications to identify relevant groups of two-wheeler components and modules, each including multiple, often specialized, subcomponents (thus, 'body' includes frame, suspension, forks, etc.; 'brakes' include drums, cylinders, pads, shoes, etc.). Working from publicly available sources (e.g., motorcycle and spare parts catalogs, national and international exhibition business directories, online bikers' communities, etc.), we identified a list of 43 component suppliers. Checking this list with our full core firm group, we identified 19 suppliers who each had links with at least seven core firms: 13 of them agreed to participate and supplied details about their ties to core firms.

Data collection and analyses
To avoid restrictive theoretical or empirical lenses (Santos and Eisenhardt, 2005), we defined knowledge broadly, as encompassing both easily codifiable and transmittable information and complex and more difficult to codify know-how. The interaction patterns we noted implied the existence of ties between two entities, which we define as voluntary collaborative arrangements of strategic significance between independent organizations, based on written contracts and aimed at sharing tacit and explicit knowledge. Our study employed Research Notes and Commentaries 581 several data sources: quantitative and qualitative data from semi-structured interviews with executives from both tie partners; archival data (press releases, annual reports, corporate documents, etc.); direct observations, including visits to the R&D units, plants, and headquarters of nearly all core firms and every peripheral firm; and e-mails, phone calls, and follow-up interviews. In particular, we conducted 81 face-to-face on-site interviews (totaling nearly 170 hours) over a 28-month period in 2007, 2008, and 2010 with 52 different executives from both core and peripheral firms. Respondents were selected according to several criteria (Yin, 1994): long tenure in the company; direct involvement in knowledge-related decisions; and functional and hierarchical variety. This use of multiple internal sources helped mitigate the potential biases of any individual respondent (Kumar, Stern, and Anderson, 1993). 3 We began our interviews at the core firms. Each consisted of three main parts: background information on the firm and its business strategy; description of key events in the formation and evolution of the firm's ties (if any) with our sample suppliers; and direct questions about knowledge exchange/creation, including details of development projects for new scooter/motorcycle components core firms had with each listed supplier. The first interview round (2007 to 2008) focused on projects from 2004 to 2007, and the second round (2010) focused on 2008 to 2009 projects. We queried the types of knowledge exchanged with each supplier and the main direction of knowledge flows, which might be simple-mostly outbound flows of information and know-how toward suppliers or return flows to core firms-or more advanced-bidirectional and intertwined interactions that might contribute to the cogeneration of knowledge. Lastly, we asked core firms about the practices they had implemented to support such knowledge flows and any that facilitated flows between their suppliers. We used interview techniques such as 'event tracking' to gain accurate information (Eisenhardt, 1989) and repeated the process later with peripheral firms. We triangulated our interview data with observation and archival data to improve its accuracy and completeness, combining contemporary and retrospective data to improve both external and internal validity (Leonard- Barton, 1990).
We selected key peripheral firm informants on the same criteria as for core firms and followed the same basic interview structure so that the responses complemented our core firm data. We asked supplier executives to report about the specific projects identified by the core firms-about the levels of relevant information and know-how they had received from their buyers, the knowledge they had returned to them embedded in components or services, and about bidirectional flows and joint knowledge creation. They also reported on knowledge flows between them and other peripheral firms promoted by a common core firm. In all, we collected data on 892 projects enacted via 184 dyads connecting 17 core firms and 13 peripheral firms: more specifically, on 776 projects carried out via 163 buyer-supplier ties and 116 stemming from 21 supplier-supplier relationships. All these collaborations (58 of which involved racing) were based on written contracts and (as our interview analysis revealed) could be categorized into one of three broad functional areas: R&D, operations, and other up-and downstream activities, which we identified as the industry's focal organizational areas for knowledge exchange and creation. Figure 1 highlights the whole pattern of ties between core and peripheral firms (solid lines), as well as between peripheral firms (dotted lines). Table 1 highlights the main activities carried out by the core firms and the components/groups supplied by peripheral firms.
Our data analyses were structured according to established procedures for theory building from inductive research, working recursively between multiple cases and theory (Eisenhardt, 1989;Yin, 1994). We included all the interviews and archival and observational data in a report of approximately 550 pages (including quotes), beginning with in-depth analyses, synthesizing the data for each project and dyad into individual case histories, and noting knowledge flows, their dominant directions, and the practices underlying them. We then moved to cross-case analyses, which produced consistent patterns and common themes (Eisenhardt and Graebner, 2007): initially we compared varied pairs of cases (e.g., specific knowledge transfer occurrences seen from the core and peripheral firms' perspectives) and added other interfirm knowledge-enhancing practices as patterns emerge to identify more robust Strat

P8 P7
= tie between core firms and peripheral firms = tie between peripheral firms n b = number of ties with core firms n s = number of ties with other peripheral firms causal and temporal relationships (Eisenhardt, 1989). We structured our analysis of knowledgeenhancing practices using a 'temporal bracketing' strategy (Langley, 1999), organizing the data into successive 'phases,' defined so that there is continuity in the activities within each phase but discontinuities at their frontiers (Langley and Truax, 1994). As Langley (1999: 703) notes, 'many temporal processes can be decomposed in this way, at least partly, without presuming any progressive developmental logic.' Our data analysis allowed us to develop a four-phase conceptual model of how core firms deliberately shaped knowledge practices to stimulate interfirm knowledge exchange and creation, at both the dyad and network levels. Our firm informants helped us identify and validate the four phases and the knowledge flows involved.

THE SHAPING OF KNOWLEDGE EXCHANGE AND CREATION IN DYADS AND NETWORK: A FOUR-PHASE MODEL
The interfirm exchange and co-creation of knowledge in the Italian motorcycle industry network we investigated can be portrayed in four phases. In Phase 1, knowledge flows are mainly from core to peripheral firm, the former channeling information and know-how to a number of selected suppliers. In Phase 2, knowledge flows reverse this direction, moving mainly to core firms from peripherals, who return some of the knowledge they originally received in more refined forms and transfers some of their own proprietary knowledge to the core firms. Once knowledge transfer channels have been established, Phase 3 sees bidirectional flows, Strat with core and peripheral firms exchanging knowledge simultaneously, making it difficult to identify a predominant direction. In Phase 4, knowledge flows between peripheral firms that are linked to the same core firm, and learning extends from dyads to the network. These phases do not represent 'stages' in the sense of being a predictable and sequential process and do not imply a progressive life cycle logic as found in many normative change theories. Specifically, they allow us to constitute comparative units of analysis to explore theoretical ideas (Langley, 1999)-and are especially useful in illustrating how the actions in one period affect those undertaken in the next. Figure 2 illustrates the temporal structuring of the different phases, showing how an example core firm (Honda Italia) maintains a portfolio of ties with different peripheral firms. Each core firm will have such a portfolio of ongoing projects with each supplier, which may be at different phases and may relate to R&D, operations, or upstream/downstream activities alone, or to them all. This example case shows Honda's portfolio of eight projects (related to the many versions of its scooters and bikes) in which it collaborates with P12 (a steel alloy body supplier) and their different learning phases. Thus, Project A (the heavyweight CBR1000RR bike entirely designed and manufactured in Italy) entered Phase 1 in Q1, 2003 and Phase 2 in Q2, 2004, and has been in Phase 3 since Q2, 2006. Project B (developing and manufacturing frames for top-selling scooters, e.g., the SH300) spent more than two years in Phase 1, had been through Phase 2 (with inbound flows overlapping the Phase 1 outbound flows by a considerable margin), and entered Phase 3 in Q1, 2009. At the time of the survey, it was expected to enter Phase 4 in Q2, 2012, as the peripheral firm P12 starts collaborations with two other suppliers to develop a frame for next-generation scooters. Project C (developing steel alloy frame for scooters like the Foresight 250) has been in Phase 1 since Q1, 2003: Honda Japan had already developed definitive engineering details, and suppliers were invited to exploit Honda's innovation to add value for both partners. But Phase 4-type information and know-how flows have been taking place between the frame maker and partners supplying braking systems and wheels (three highly interdependent components) since Q2, 2008. At the time of the survey, Project C was expected to enter Phase 2 in Q3, 2012, with inbound knowledge flows from P12 constituting the platform for Honda engineers' future design activities.

Phase 1-Knowledge flows from core firms to peripheral firms
Core firms are channeling knowledge (information and know-how) to peripheral firms in different forms and at different codification levels, leading to spiraling interactions between tacit and explicit knowledge (Nonaka, 1994). Overlapping of knowledge flows practices that shape outbound flows and the main benefits (peripheral) recipients perceive from those flows. With respect to R&D (Table 2, Phase 1), in 301 projects, we found core firms sharing information about development cycles with peripheral firms to ease prototype development and gain time/cost efficiencies. Core firms transferred information and techniques on product concepts and assembly designs to suppliers in nearly onethird of the projects we studied, and they also (to a lesser extent) shared detailed blueprints and offered specific training about product concepts and design-for-assembly logics and techniques. Other core firm practices we observed included: preparing manuals outlining practices and rules for R&D interactions, supporting purchasing costs of R&D tools, and licensing proprietary technologies to peripheral firms. Transferring tacit knowledge effectively requires extensive personal contact and trust and the supplier developing codification processes to transform it into explicit knowledge. At the operations level (Table 3, Phase 1), practices supporting outbound knowledge flows included: sharing detailed production schedules with suppliers, checkup visits to their company and plant, organizing and delivering training about production methods and working practices (such as just-in-time and lean production principles) to align and synchronize manufacturing and assembly processes. Core firms also arranged training on statistical quality control methods and quality performance indicators to assist suppliers with quality assurance and quality certification, as well as on ICT and enterprise resource planning (ERP) platforms, inventory management techniques, and plant management principles. A number of practices covered other upstream and downstream activities (  peripheral firms, thus improving their planning capacities and operational effectiveness, and also gave training on cost breakdowns and on management and strategy tools to help suppliers improve their efficiency, even providing financial support to help them buy tools and equipment. Core firms also established communication initiatives aimed at helping suppliers find new markets to increase their abilities to identify and enter new markets (often related to racing events) to help improve their financial stability: shaping joint visions and seizing market opportunities allowed these partners to develop joint dynamic capabilities.

Phase 2-Knowledge flows from peripheral firms to core firms
Peripheral firms relying simply on markets to access specialized assets and acquire technology are likely to see their capabilities deteriorate (Teece, 1986) and to remain stuck in passive, dependent postures (Rosenberg, 1982). Peripherals that purposefully supply core firms with return knowledge turn their relationships into partnerships, making it harder for core firms to switch to other suppliers. In terms of inbound R&D knowledge (Table 2, Phase 2), our study depicts peripheral firms as providing specialized knowledge and technical information: by sharing proprietary technology and know-how, including alerting their partners about advances in state-of-the-art technologies and proposing design alternatives to allow them to fulfill their customers' unique demands more completely. Other practices included: organizing training sessions to develop design-sharing capabilities, developing rapid prototyping techniques, and using simulation-driven design to give core partners with significant engineering cost benefits and more effective design analysis. At the operations level (Table 3, Phase 2), suppliers shaped their knowledge flows to core firms by investing in highly specialized finishing processes and treatments, a practice we observed in nearly one-fifth of the projects and which benefitted a wide range of activities including product and process feasibility, surfacing, structural analysis, and prototype engineering. Peripheral firms also transferred knowledge embedded in updated production scheduling programs, made considerable investments in developing quality production capabilities, and offered preventive maintenance of core firms' production machinery-flows that were enabled by their possession of dedicated manufacturing tools, CNC machines, CAD-CAM working stations, and flexible manufacturing systems. Such inbound flows also extended to other upstream and downstream activities (Table 4, Phase 2). Peripheral firms provided core firms with detailed price breakdowns and supported the preparation of technical specifications and documents for final users, building core firms' reliability and reputation. Their practices extended to sharing information about complementary sources of components and technology, which may seem to be counterintuitive behavior, but was aimed at reinforcing network potential and effectiveness and supporting core firms' reactions to market opportunities, e.g., by speedy response to changing vehicle emissions regulations.

Phase 3-Simultaneous, bidirectional knowledge flows between core and peripheral firms
In Phase 3, firms reciprocally exchange knowledge and cocreate 'collaborative knowledge' (Simonin, 1997), a prerequisite for virtuous relational learning. In terms of R&D at this phase (Table 2, Phase 3), design engineers on both sides of the dyads recognized the value of working together on projects to share expertise, foster problem solving, and improve the efficiency and quality of product development: the interaction inherent in these knowledge-enhancing practices nourished virtuous cycles of reciprocal learning. In the projects we studied, we observed structured meetings to exchange information and know-how to fix problems rapidly, as well as recurrent codesign and co-engineering activities at component and module levels, joint testing, and the systematic parallel updating of partners' CAD models to reduce design times and costs, allowing both parties to rapidly assess concept, development, and total lead times via interfirm ties. Hosting partners' engineering teams supported simultaneous knowledge exchanges, and our data also revealed a significant number of relational platforms (or 'customer trial centers,' as one core company labeled them) where teams of engineers from core and peripheral firms interacted to solve specific problems. Joint participation in racing also enhanced both parties' knowledge and supported their R&D activities, as many innovations developed for racing bikes were later featured in retail production  (Table 3, Phase 3), monthly meetings were arranged in one-fifth of the projects we studied, allowing them to constantly assess the 'pull' logic, increasing core/periphery alignment and time efficiency and lowering coordination costs. We also noted how parties interacted to jointly assess lean manufacturing principles and techniques and employed value stream mapping to raise their cost and time efficiency and their effectiveness in attaining manufacturing goals.
Other practices supporting bidirectional knowledge flows included: synchronizing purchasing orders and optimizing warehouse stock volumes, jointly implementing integrated enterprise resource planning (ERP), and using 'quality circle' initiatives to cut defects and waste. In the 'other activities' sphere (Table 4, Phase 3), we saw evidence of joint assessment of end user profiles to improve response effectiveness and participation in fundraising projects to increase resource availability. Coevolutionary alignment took the form of the periodic updating of communication and information flows needed to respond quickly to environmental jolts or to competitors' moves.

Phase 4-Knowledge flows between peripheral firms
In the fourth model phase, we observed knowledge flowing between different peripheral firms, whose connection had been promoted and supported by a common core firm. Largely unexplored by strategic literature, this guided evolution of learning processes from dyads to networks yields significant insight into how core firms connect alters to gain informational (Burt, 2000) and other difficult-to-replicate network-specific advantages.
Redeploying knowledge dispersed across collaborative networks requires that core firms possess (or develop) learning and teaching capabilities across organizational boundaries, as this targeted coordination of knowledge mobility is central to a network's ability to co-create value and innovation (Zhao et al., 2005). As for R&D (Table 2, Phase 4), we observed knowledge exchange and creation between peripheral firms being accomplished by: training suppliers on improving team working skills for greater effectiveness across dyads and networks, transferring methods and techniques to stimulate co-design activities among first-tier suppliers, and organizing workshops and seminars on R&D organizing principles at the network level.
Core firms supported such intersupplier knowledge flows by identifying second-tier suppliers with complementary R&D capabilities. At the operations level (Table 3, Phase 4), core firms extended effective supply chain and manufacturing activities across their networks by organizing workshops and coaching sessions, transferring methods and techniques to stimulate coordination among first-tier suppliers, and providing training to diffuse planning tools covering a range of activities from initial orders to delivery. To a lesser extent, we also saw core firms play active roles in helping peripherals manage the initial stages of their interrelationships, such as helping develop joint manufacturing plans. In other areas (Table 4, Phase 4), core firms organized sessions between peripheral firms on market forecasts and identified synergies across different alliances. Core firms' ability to lead by sharing strategy also hinged on them sharing lessons learned, databases, and potential partner profiles to create further opportunities for their suppliers. Core firms ran coaching sessions to reinforce the network's goal alignment and helped suppliers elaborate 'network contracts' to give them unique identities in fund-raising applications. 4

DISCUSSION AND CONCLUSION
We present a structured framework that takes a multilevel perspective to address different relational subsets, depicting knowledge flows at dyad and network levels. We found core firms that took charge of the processes of interfirm learning from firms to dyads, and from dyads to networks, and of knowledge-enhancing practices all needed to nurture the transfer, recombination and creation of specialized knowledge. Such practices are favoring the affirmation of higher-order organizing principles (Grant, 1996), the idiosyncratic learning processes (Kotabe et al., 2003), and the interfirm knowledge-sharing routines (Dyer and Singh, 1998). The network we studied combined the advantages of common identity and language normally associated with hierarchical forms (Kogut and Zander, 1992;Conner and Prahalad, 1996) with the learning incentives that typically occur in dyads and networks. Having multiple