This article examines technological interdependence as a way of understanding technology change and learning in the context of Taiwan's quest for industrial upgrading and innovation. A distinctive feature of Taiwan's postwar economic development was decentralized industrialization, which consists of a system of small and medium enterprises (SMEs) clustered in a geographical locale where numerous small firms complement one another in the production process, each specializing in one phase of production. SMEs as a whole formed the foundation of the so-called Taiwan miracle, in which various industries—ranging from shoes, apparel, bicycles, and machine tools in the 1980s to information technology (IT) industries from the 1990s onward—successfully entered the global production networks. The ability to directly participate in the global market distinguishes Taiwan's SMEs from their counterparts in other countries, such as Japan, Korea, the United States, and France. Despite the system of SMEs being the driving force for initial success, it is believed that as the race has intensified, increasing scale of production and increasing expenditures in capital input and research and development (R&D) are the key to staying ahead. Large Korean conglomerates, such as Samsung and Hyundai, are seen as models in the latecomer catch-up literature.

Nevertheless, many SMEs, contrary to the “hollowing out” prediction, have made the transition to high-value-added production and continue to be strong exporters and key players in the world market. The Taiwanese bicycle industry is a case in point. Despite mounting pressure to seek lower-wage production sites, the industry has moved up the value-added ladder and engaged in the global higher-end bicycle trade with advancement in technology and materials. If technological advancement is the key for catching up and staying ahead, what explains the learning capacities of the SME network production system, and where does their learning come from despite relatively low formal R&D expenditures?

I argue that the parts makers are central to technological breakthroughs in a decentralized industrial system, contrary to the conventional view that stresses lead firms as drivers of technological learning. Two concepts are relevant in explaining the technological dynamism of the parts makers. First, the notion of technological interdependence helps to understand incremental and small cumulative improvements/technological advancement by parts makers that have an impact in a wide range of sectors as opposed to the conventional emphasis on radical innovation that breaks away from the past, presumably by leading firms. Second, the emphasis on the decentralized industrial structure is pertinent because it has an impact on the technological diffusion process and the particular paths taken by manufacturers. This article then illustrates the process of technological change and the quest for innovation in the context of Taiwan's decentralized industrialization by studying the adaptation and application processes of entrepreneurs and bicycle manufacturers in commercializing new technologies.

This article contributes to STS studies by highlighting the structural characteristics of the various social groups in affecting the technology development of artifacts and learning of entrepreneurs and manufacturers, following the call by Klein and Kleinman (2002) for bringing structure back into STS studies. In particular, this article introduces industrial dynamics, as expressed in the power relations and capacity differentiations in technology and information access between latecomers (Taiwanese bicycle manufacturers) and their competitors in advanced countries (i.e., the US and Japanese manufacturers) and between assemblers (leading firms) and parts makers in contributing to the adaptation of technologies in bicycle manufacturing. Structural characteristics explain the conditions in which a particular direction of technology became dominant and impacted the outcome.

In what follows, I first summarize the features of decentralized production that are relevant to understanding the technological change and learning I will discuss. I then elaborate on the notion of technological convergence and interdependence in understanding technological change, followed by the case study of the response of Taiwanese bicycle manufacturers to the rise of mountain bikes (MTBs) and breakthroughs by Taiwanese frame makers to illustrate my point. Finally, I explain the foundations of technology learning within the context of a decentralized industrial system.

## Taiwan's Decentralized Industrial System and the Catch-up of Latecomers

The return of the decentralized network of production since the 1970s, when US firms downsized and tapped into resources of global production networks instead of reinvesting and expanding vertically integrated organizations, has generated studies concerning the causes and consequences of such restructuring and its impact on technology innovation (Langlois 2003; Lazonick 2009). This shift in organization, which involved a disintegration of vertical hierarchy (i.e., in-house production) with a decentralized network of production, was often associated with a response to globalization and growing global production. Moreover, the disintegration of vertical production was made possible by the emergence of modularity in technology and industrial standards whereby component makers work on a standardized interchangeable platform that permits autonomous innovation of each component, contrary to the earlier integrated model that depended on a system innovator or a proprietary technology in a closed system. The rise of the global IT production network and the transformation in the semiconductor industry are examples illustrating the dominance of modularity (Sturgeon 2002; Lazonick 2012: 116; Langlois 2003). For instance, in the development of the semiconductor equipment industry, the emergence of cluster-tool equipment in the United States led to the development of customized chip production, which replaced the Japanese fabrication equipment designed for mass production of memory chips (Langlois 2000).

Consequently, a salient feature that has accompanied globalized and modular production is the increasing number of firms focusing on intermediate input instead of final product. These firms are incorporated into the global production network as independent suppliers or subcontractors and thrive on being specialist firms (Whitford and Potter 2007; Andersen and Christensen 2005). Recent trade statistics show that increases in value added and trade are to be found within nonfinished products, as opposed to final products (WTO and IDE-JETRO 2011). A growing number of studies have examined the technological changes generated by these parts makers and specialist firms in the production process rather than treating them as powerless subcontractors supported by large leading firms. Theoretically, Taiwan is an interesting and important case to examine because its integration into the global economy and global production networks was a result of such a shift and an expanding transnational production network (Gereffi 1994). Therefore, the specifics of the Taiwan experience will contribute to the discussion on technological change that accompanied vertical disintegration and the learning by parts makers and specialist firms.

Various industries within Taiwan's decentralized industrial system have the following general characteristics in common:

1. The SME-based production system encompasses an extensive division of labor in which firms complement one another in the production process. They cluster in a geographical locale, or “industrial district” (Piore and Sabel 1984), where numerous firms compete and cooperate in the same industry. Bicycle production, like many other industries in Taiwan, comprises two sectors: an assembling sector and a parts sector. Both the assembling sector and the parts sector involve an extensive system of subcontracting and a high degree of specialization. For example, the various components within a part are subcontracted to small factories that specialize in manufacturing these parts.

2. The SME production network consists of numerous independent parts makers and processing specialists who focus on intermediate input and do not make the final product.

3. The production networks are decentralized in the sense that they are open and nondependent networks in which suppliers and specialist firms are usually not tied to particular assemblers. They can supply several firms within the industry or sell to other industries.

4. Parts makers and specialist firms are incorporated into the global production network and compete directly in the world market rather than being completely dependent on domestic assemblers.

One direct consequence of such decentralized industrialization is that interindustry linkages are high. The open and nondependent network means that information not only travels within the industry but also cascades among industries. Workshops that perform some processing jobs for bicycle assemblers and parts suppliers, such as drilling, lathing, milling, metal surface finishing, and anodizing, are not locked into one particular supplier or one industry—they perform processing jobs for other industries. Moreover, the ability of parts makers to connect to the global market means that their points of access to information are multiple, so they have multiple sources of learning. This means that knowledge and ideas cross industry boundaries and are not contained under a single roof.

As discussed above, modularity has permitted global outsourcing as a standardized interface and has codified information to facilitate geographical expansion of production. The late industrializers are incorporated into the lower end of the production chains by being original equipment manufacturing (OEM) producers who depend on multinational corporations (MNCs) to disseminate information and knowledge transfer in the form of blueprints, technical specs, and technical assistance to develop the capacities of the suppliers to ensure they meet the buyers' requirements—the so-called OEM model (Ernst and Kim 2002). While modularization may provide learning opportunities for the late industrializers, a decentralized industrial structure, as seen in the Taiwanese SME system, is believed to be incapable of moving up the global technology ladder, despite its initial success, because of its low R&D and marketing capability to reap technology rent, an assumption derived from the following two assumptions on the technology learning of the latecomers.

First, the literature of the global production network/global value chains (GVCs) assumes that the learning of latecomers, such as Taiwan, stems from their being incorporated into the globalized production network, meaning linkages to leading MNCs that coordinate the chains and drive technological changes (Humphrey and Schmitz 2002; Gereffi, Humphrey, and Sturgeo 2005). But this position also postulates that most profits are captured by firms that either design the product (product innovators) or use branding marketing and control distribution (Gereffi 1994). Therefore, for latecomers to succeed in the next stage of development, a crucial key is to move up the hierarchical ladder of the GVCs and ultimately become the leading product innovators or control the distribution. Consequently, the indicators of upgrading are based on sectoral transformation and moving up the value chain of a particular product. In turn, technological advancement is assumed to be driven by large firms downstream that make the final product. Second, the other stream of latecomer literature suggests that catch-up of latecomers requires a different strategy: increasing concentration and “scaling up” production to increase return, as suggested by Amsden and Chu (2003). Empirically, the rise of global megasuppliers in various industries (e.g., the electronics and auto industries) is an example of scaling up in response to increasing competition resulting from modularity (Gereffi 2013).

## Technological Convergence, Interdependence, and Learning of the Parts Makers

Despite increasing vertical disintegration of production at the global level, the existing literature assumes learning coming from lead firms and a top-down direction of knowledge dissemination and overlooks the possibility of horizontal learning between firms that participate in different global production networks. Moreover, modularity would suggest a scaling up for latecomers to secure their position in the GVC. Yet, scaling up may be insufficient to explain the technological advancement of parts makers in a decentralized system. The emphasis on changes made by individual component makers independent of others is not sufficient to explain complex technical changes that require overlapping knowledge in a context of network-based production. More and more studies suggest that modularity has created a new set of problems because of the lack of overlapping knowledge that is central to integral design and changes (Helper and Sako 2012: 167; Sabel and Zeiltin 2004).1 Thus, how do we make sense of these technical changes generated by specialist firms and industries that produce no final products, especially among SMEs?

Nathan Rosenberg's (1963) discussion of technological convergence and the associated concept of technical interdependence best illustrate the technological learning and advancement relevant to parts makers and specialist firms. Contrary to understanding the technological changes by focusing on upstream to downstream in the production of a final product, in which techniques and skills are assumed to be tied to particular vertical sequences and a narrow range of industry, Rosenberg shows that throughout the course of American industrialization in the late nineteenth century, a phenomenon that he terms technological convergence was crucial for understanding the technical development of the machine tool industry and the related metal-using sectors. Technical convergence refers throughout the machinery and metal-using sectors to common processes, skills, techniques, and facilities that exist at the intermediate input level (or what he called higher stages of production) and are similar in a wide range of final products/industries (Rosenberg 1963: 423). For instance, all machines performing related operations (e.g., milling, grinding, drilling, or boring) may face similar kinds of technical problems and problems related to the properties of the metals. Thus these problems that occur in the production process are common to the production of a wide range of commodities that are apparently unrelated to the end product's use but are closely related on a technological basis, as can be seen in the technical evolution in firearms, sewing machines, bicycles, and automobiles (Rosenberg 1963).

Technological convergence in turn impacts the development of new techniques and technology diffusion. The breakthrough often starts from solving some particular problem in the intermediate production processes, but the solution then provides free technological inputs (either new skills or new techniques) and cascading effects or applications of the newly learned techniques to other metal-using industries, as illustrated in the case of machine tools (Rosenberg 1963: 426).

The intrinsic characteristics of technological convergence imply that many innovations that are related to manufacturing processes at the intermediate input level by specialist firms are not identifiable at the product level but nonetheless contribute to improved performance of their industry or other industries. Thus, technological change often occurs not at the “main entrance,” measured in the highly visible form of patents, but seeps in through less visible and unobservable rear entrances (Rosenberg 1983: 56). Properties that are often overlooked in understanding the processes of technological change include complementarities, cumulated impacts of small improvements, and interindustry relations; these can account for the potential role of parts makers. The notion of complementarities suggests that the economic and social consequences of a breakthrough in a product or an industry often involve advancements in a number of interlocking mutually reinforcing technologies. Thus, the improvement in a product may depend on the availability of a breakthrough in a complementary technology for a component. For instance, the demand in lightweight and strong materials often awaits appropriate metallurgical inputs (58–61). In turn, technological change often comes from incremental and cumulative small breakthroughs that focus on modifications instead of on a major, spectacular discontinuous innovation.2 Interindustry linkages become important for identifying this kind of technological innovation. The point is that improvement in industry A often came from technological breakthroughs at the intermediate level in industry B yet was not identified by the conventional measures or was observed only in the improvement in industry B (62). This is because many aspects of technological change (e.g., new materials, machines, components, and technical processes) are visible only at the intermediate level and never show up in conventional measures of the final product, as they are mostly in capital goods industries. Moreover, interindustry flows of less visible advancements have yielded wider impacts on product improvement in other industries or social and economic payoffs in the economy (71–72). The technological progress in industry A is highly dependent on the technological development in other sectors, thus making innovation highly interdependent.

The interindustry linkages suggest that emergence of specialist firms and capital goods industries (e.g., machine tools) comes from the demand of users eliciting the simultaneous growth of several industries that were technologically convergent (Rosenberg 1963: 424). This brings interindustry linkages to the forefront in understanding the innovation by parts makers, as specialization of specialist firms would not exist if there were only vertical disintegration and no technological convergence. In turn, technological interdependence and interindustry linkages create an understanding of innovation beyond Marshallian industrial boundaries that enclose the division of labor within a single industry (Rosenberg 1983: 71).

Following this logic, “catch-up” in the latecomer context requires further clarification. Measures such as R&D expenditures and capital investment may not capture what actually goes on in the economies. These breakthroughs, often associated more with the development of R&D and the learning of production processes, may not be recognizable as part of the R&D process or receive any direct expenditure (Rosenberg 1983: 121–22). Therefore, cases studies centering on the technological learning of these specialists are sorely needed. We can now turn to the case study of the Taiwanese bicycle industry.

## Advancements in Bicycle Manufacturing and Processing Technology

The bicycle industry, clustered in central Taiwan, consists of an assembling sector and a parts sector in which numerous firms specialize in one segment of the production process, as is typical of Taiwan's SME decentralized system. Over the past three decades, this industry has upgraded itself from being a third world producer competing on cheap labor to a key player in the global bicycle industry. Since the 1990s, the bicycle industry has begun to move some of its production to lower-wage sites, such as China and Vietnam. In doing so, it has followed in the footsteps of other labor-intensive light industries, but the bicycle manufacturers have managed to shift to high value-added production in Taiwan and have been able to compete in the high-end segment of the bicycle trade. For instance, bicycle exports from Taiwan, which reached 10 million in 1998, dropped to 4 million bicycles in 2003 and rebounded to 5 million in 2008. While the total volume has dropped, the total value of bicycle exports has increased: the average price of the bicycles exported has moved from US$95 in the 1990s to more than US$400 in 2012 (Figs. 1, 2).3 The industry's adaptation experience reveals how SME networks/clusters can remain locally rooted but continue to move up in the GVC—and thus illuminate the governing principles of decentralized production and the sources of capability building and learning.

The conventional view focuses on leading Taiwanese assemblers such as Giant and Merida, which successfully upgraded from OEM to their own brand-name manufacturing, and it emphasizes the importance of learning from OEM in the Taiwanese bicycle industry's ascent up the global value ladder (Cheng and Sato 1998; Chu 1997, 2009). However, it is rather unclear whether one or two leading firms alone could drive the entire process, since the top ten assembling firms constitute approximately 40 percent of production, and there were over ninety assembling firms and over one thousand parts makers according to Taiwan's 2001 and 2006 industry and commerce censuses.

Contrary to the emphasis on leading assembling firms, I contend that the technological prowess of parts makers in the decentralized system has been central both to preventing “hollowing out” and to making a genuine transition to the current high-quality bicycle trade. This is supported by the fact that Taiwan has become a hub for manufacturing high-end bicycle components in the global market, as opposed to a simple assembling production site of imported parts in exchange for processing fees. This indicates that the backward linkages of technology spillover have taken root. Moreover, the fact that the percentage of locally produced parts has moved from 30 percent to 70 percent or more (on average) for a premium bicycle manufactured for the export market suggests improved technological prowess of component makers over time despite competition from countries with lower wages. Throughout my fieldwork, industry representatives indicated that the importance of parts makers had grown hand in hand with that of assemblers. As one interviewee put it, “Parts makers and we [assemblers] are like fish and water—we cannot survive without each other” (HM interview, 2004).4 Those in my interview sample confirmed that the turning point of the industry was the rise of the mountain bike (MTB), which led to the consolidation of Taiwan's bicycle industry in the global competition: “Almost all firms made a lot of profit when MTBs became popular” (AC interview, 2003). This raises the empirical question of how the Taiwanese bicycle manufacturers, as latecomers, managed to establish their dominance in the global bicycle industry.

Following the approach of social construction of technology (SCOT), I reconstruct the history of development of the MTBs by adopting a technological frame centering on the problems and solutions by relevant manufacturers to the rise of MTBs. Technological frame, in Wiebe E. Bijker's formulation (1995: 123), refers to the interactions among relevant social groups evolving around an artifact and consists of the problems, goals, and solutions perceived and executed by various relevant groups in the development of the artifacts. Moreover, as will be illustrated, the analysis shares SCOT's view of interpretative flexibilities, namely, that there are multiple possibilities for technological design and artifacts, instead of a linear view of technological change (Pinch and Bijker 1987; Bijker 1995). This article differs from Bijker's SCOT analysis by highlighting the structural dimensions of the social groups in affecting the technology construction processes. Bijker's agency-centered analysis of interpretative flexibilities places equal weight on and equal access to resources among the relevant social groups. It assumes that closure occurs when interpretative flexibility of an artifact is diminished, implying that the meaning becomes fixed and consensus is reached among the groups (Bijker 1995). Yet the descriptive approach of interpretative flexibilities does not explain how and under what conditions the consensus/closure would occur, how certain groups' interpretations would outweigh others', or how those groups gain such capacity over the others in interpreting technological development. While Bijker in his later analysis (1995: 262–64) invokes semiotic power relations and micropolitics in explaining the outcomes and closures as an attempt to incorporate structure into the analysis, the focus on cognitive power overlooks other forms of structural differentiations among different social groups in contributing to their asymmetrical capacities (either cognitive or economic resources, as well as others) in shaping the artifacts, as Klein and Kleinman forcefully argue (2002: 40).

In congruence with the structural thesis, this article introduces the structural aspects of power relations among different actors situated in the broader social context to explain the outcomes. The case study focuses on the responses of latecomers—the Taiwanese bicycle industry and manufacturers—in relation to their competitors in advanced countries to the emergence of MTBs and their solutions to various problems evolving around MTB frame manufacturing technologies. The case specifies the conditions in which of the powerless SMEs were able to leverage their advantages to bypass control of large MNCs. Special attention is given to the breakthroughs by parts makers and specialists that helped the industry establish itself on the global scene and impelled subsequent technological change. Moreover, the case reveals that interindustry learning and interfirm collaborations have been central to the breakthroughs, contrary to assumptions that modularization of bicycle production would lead to autonomous development by individual firms alone and that improvement of the parts can be made independent of others in the networks (Galvin and Morkel 2001).

In short, the aforementioned concepts of technological interdependence (interindustry learning) and decentralized industrial structure account for the technological capacities of the parts makers, which in turn enabled the latecomers, the Taiwanese bicycle manufacturers, to outcompete their counterparts in advanced countries to become dominant players. In what follows to demonstrate my points, I use the examples of (1) the adaptation of TIG (tungsten inert gas) welding technology to frame making, (2) the adaptation of a new material, aluminum, to frame making, and (3) the application of tube hydroforming to the bicycle industry.

The following account of the history of the MTB and the technological development of the Taiwanese bicycle industry is based on my fieldwork from 2003 to 2014 from site visits and interviews with relevant industry representatives, including owners, engineers, and managers; engineers in government-funded research institutes; and US buyers who outsourced bicycles to East Asia in the 1980s. Not much has been written to date on the subject. Hence I have relied on interviews with industry representatives and key persons who made breakthroughs. To ensure accuracy, these interview data have been cross-checked with other interviewees dealing with the same topic and compared with a variety of secondary sources, such as newspaper articles, trade publications (e.g., the cycle press), and books on the subject.

## Mountain Bikes and Other Transformations since the 1980s

The MTB, which originated on the US West Coast, was a revolutionary invention, in that it led to a new style of bicycle riding, a new production method for bike frames that subverted the existing method, and the creation of new components. Off-road riding became possible and even evolved into a cutting-edge sport. A direct result of the MTB craze was a tremendous increase in both volume and value of MTBs that changed the landscape of the bicycle industry and prompted a reshuffling of competitors within the industry. The Taiwanese bicycle industry was able to ride the wave, capturing most of the market for MTBs, becoming the leading exporter of MTBs, and consolidating its dominance in the industry on the global scene. In contrast, the Japanese bicycle industry, the leading bicycle exporter in the world market in the early 1980s, instead of responding to the changes retreated to its domestic market, leading to a further decline in its global importance.

A crucial factor in the rise of the Taiwanese bicycle industry was its quick response to MTBs with the new TIG frame-welding technology, in which bicycle tubes are welded directly together in a blanket of inert tungsten gas (Figs. 3, 4). This differs from traditional frame making with the lug-brazing technique, in which the bicycle tubes are brazed together inside sleeves or sockets (known as lugs) that connect the tube ends (Figs. 5, 6). Before the emergence of MTBs, the most popular bicycles were racing bikes, whose frame-making technology was based on lug-brazing techniques, which European and Japanese bicycle makers had mastered and in which they specialized. The introduction of TIG welding gave rise to more flexible production and designs for bicycle frames, which had a huge impact on the subsequent development of bicycles and changed the dynamics of the industry worldwide.

In what follows, I discuss bicycle history and the organization of the global bicycle industry to establish a context for the discussion of the relevance of the rise of the MTB and the changes it generated. I then describe the technical changes in bicycle production associated with MTBs, Taiwan's responses to TIG welding, and subsequent breakthroughs in various technologies.

### The Global Bicycle Industry prior to the Rise of MTBs

Prior to the rise of the MTB, high-end bicycles were lightweight road bikes and ten-speed racing bikes made in Europe and Japan. The history of US bicycle manufacturing was one of large, modernized, mass-production plants, which relied upon automatic welding on carbon steel tubes and mass production of heavyweight bicycles (Crown and Coleman 1996; Norcliffe 1997). Automation requires thicker-gauged tubes and results in heavier bicycles.5 As a result, US manufacturers had never acquired the knowledge and technology to manufacture lightweight bicycles, which became crucial in the later development of bicycle technology and trade. In the 1970s, when consumers increasingly preferred lightweight road bikes, the US solution was to import them from Japan. While the Japanese manufactured high-end lightweight road bikes for the US market, their small Taiwanese counterparts were making the popular models of juvenile bikes—namely, high-risers and some lower-end standardized road bikes.

The development of Taiwanese bicycles as an export-oriented industry began in the early 1970s due to a sudden surge in US demand that coincided with Taiwan's shift to export-led industrialization (Crown and Coleman 1996; Hsieh 2011). Taiwanese manufacturers were incorporated into the global bicycle production network as OEM manufacturers of juvenile bikes and, subsequently, of low-end road bikes for the US market. By the late 1970s, Taiwanese bicycle manufacturers were making BMX (bicycle motocross) bikes, a mini version of the earlier MTBs, based on American buyers' specifications. At that time BMX was considered a marginal sport engaged in only by teens and kids in the United States.

None of the Japanese firms was willing to enter the BMX trade, due in part to the notion that bicycles simply did not look anything like the BMX bikes. Consequently, the Japanese had not acquired bicycle welding technology such as electro-welding, which was considered a lower-end technique compared to lug brazing, which was the prevalent method for making high-quality road bikes. Yet, as discussed below, the technology of the Taiwanese latecomers turned out to be an advantage for the shift to making MTBs because the electro-welding technology that they had learned from making the BMX bikes laid the foundation for Taiwan's subsequent breakthrough in MTB welding.

### Changes Generated by MTBs and Responses of Taiwanese Bicycle Manufacturers

MTB riding involves riding off-road and up steep, uneven hills, which requires a strong, lightweight bike. This led to a demand for lighter materials and completely different designs in frames and components than those used in road bikes. When MTBs first came out in the United States, the dominant bicycle producers, the Japanese and European firms, did not aggressively engage in their product development. European bicycle makers' understanding of what constituted a bicycle did not look anything like the MTB. And most Japanese producers assumed that MTBs would gain popularity only in the United States, not in Japan or elsewhere. As it turned out, “this was the wrong fad to ignore. The mountain bike and its urban offspring would prove to be the most important bike trend of the 1980s, accounting for nearly two-thirds of the adult market by decade's end” (Crown and Coleman 1996: 114). The rise of MTBs thus led to a reshuffling of competitors on the global scene as Taiwanese manufacturers rose to be key players. More specifically, the reshuffling was due to the new frame-welding technology for MTBs—namely, TIG welding—that Taiwanese firms had mastered.

By 1980, MTB pioneers in the United States were using thinner-gauge chrome-moly steel (a lighter and higher-grade material than the dominant carbon steel material) to build bikes for mountain riding. They also experimented with TIG welding, which connected the tubes directly, a technique widely used in the aerospace industry. Although the initial MTB prototypes were handcrafted with TIG welding by tinkerers, the first mass-produced MTB (called the Stumpjumper) introduced to the general public in 1982 was constructed with lug brazing; it was designed by a US company called Specialized and mass manufactured by a Japanese company, Araya. Today most MTBs are manufactured by TIG welding, a technique that the Taiwanese actively pursued in bicycle production.

How did TIG welding become the dominant manufacturing method for MTBs? And how did a latecomer (in this context, the Taiwanese manufacturers) adopt TIG welding for MTBs? Most Taiwanese assemblers concur that Taiwanese bicycle manufacturers were the ones who applied TIG technology to the commercialization of bicycle production beyond prototyping, even though the initial idea of using TIG welding to make bicycles came from the United States. A leading assembler noted, “At that time, American buyers said that this needed to be TIG welded. Taiwan did not have TIG technique. But the electro-welding technique was very well developed [for BMX bikes]. Therefore, it was easier to learn” (FC interview, 2003). The former vice president of the US bicycle company Schwinn concurred that TIG became an important consideration for outsourcing bicycles to Taiwan in the 1980s (JT interview, 2004).

Many of those I interviewed credited a man known as Atsu,6 a component maker who manufactured front forks, as the key person in the breakthrough of applying TIG welding technology to bicycle manufacturing. An assembler who claimed to be the first to apply chrome-moly tubes for bicycles and who had worked closely with Atsu in the late 1970s recalled the problems he encountered when dealing with the application of new alloy material:

I was the first one who applied chrome-moly material. But frankly speaking, I was not so certain about how to go about this at that time despite books that would tell you that TIG could go with chrome-moly tubing. I had a lot of defects at that time. . . . Theoretically, thinner gauged chrome-moly tubing needs TIG welding because the thin alloy gauge needs to be welded in a lower temperature. Arc welding [the dominant electro-welding for low-carbon steel] would just overheat chrome-moly tubing and destroy the whole thing. (PL interview, 2011)

A welding technician who participated in TIG welding at that time recalled the experimentation: “TIG welding initially was for industrial use [i.e., for shipbuilding and the aerospace industry]. We [the Taiwanese bicycle industry] were the first ones to apply the technology to a consumer product. At that time, Atsu from 808 [company] imported three machines from Hitachi Japan. The Hitachi technician taught us the basic techniques of TIG welding. . . . But we explored the techniques for welding bicycles” (FL interview, 2004).

TIG welding was thus introduced as a new manufacturing technology to the bicycle industry from its applications in other industries. The assembler, who was a pioneer in making MTBs, emphasized the interindustry learning that this entailed: “The TIG welding equipment/technology was not a new invention. Hitachi, Panasonic, Osaka Denki, and the US Miller [Manufacturing Company] were manufacturers of TIG welding equipment. But what was new here was to apply TIG welding technology to the bicycle industry” (PL interview, emphasis added). Here the breakthrough started from solving problems with the production equipment to meet the welding properties of the materials, which initiated a wave of technological changes in bicycle production. The assembler explained the breakthroughs by Atsu:

Atsu was the first one who applied TIG welding to chrome-moly-framed bikes, meaning mass-produced TIG-welded frames. He had expertise in building equipment and modifying equipment to solve technical problems. Atsu approached the whole thing by building the welding equipment to setting up a TIG welding production line . . . on which each welder focused on welding one segment of the bicycle frame, instead of welding the whole bicycle like in prototyping. (PL interview)

The welding technician highlighted the importance of the modification of the equipment in the adaptation process:

The size of the earlier-generation TIG welding machines was much bigger because they were for industrial purposes. . . . Not many industries in Taiwan were using TIG welding equipment at that time, with the exception of China Shipbuilding (中船). Later on, when the quantities were large enough, we asked them to modify the welding machine specifically for bicycle-making purposes. Now the TIG welding equipment has been modified to be much smaller. (FL interview, 2011)7

Taiwanese bicycle makers quickly investigated the TIG welding technique, building upon their prior experience with various forms of electro-welding. They had already explored different forms of welding techniques, starting with electro-welding, for the earlier steel-made BMX bikes. Subsequently, Taiwan began to apply different materials and different welding techniques to bicycle manufacturing, such as the chrome-moly frame that made for a lightweight bike.

### Competing Frame-Making Technologies: TIG Welding versus Lug Brazing

Most bicycle-frame manufacturers stated that applying TIG welding technology to bicycles was an immense trial-and-error process (AL interview, 2003; MWR interview, 2004). The trials and errors were feasible because of the flexibility and low cost of the new welding method, which suited the system of SMEs, and also because of the Taiwanese firms' prior experience with other forms of welding. With TIG welding, any bicycle tubes can be welded directly, an advantage that lug brazing cannot offer.

Lug brazing, which connects bicycle tubes with a sleeve called a lug, requires tooling for each frame design. The lugs are produced at a fixed angle, and a different mold is needed to produce each specific angle lug (see Fig. 6). Lugs are produced in batches of thousands to reach scale economy. Lug brazing worked well for a conventional standardized diamond-frame design but was inadequate to accommodate all the new changes associated with MTBs.8 MTBs were evolving toward frequent changes in components that would ultimately change the structural design of the frame, and thus the angles needed (see Fig. 4). This required lots of different angle lugs (and molds) for building an MTB frame, even just for a minor change in design. As one interviewee put it, “BMX and MTBs are about creating unlimited freedom, but lugs are constraining the designs” (PL interview). Flexibility was crucial because MTBs, as discussed below, were heading in the direction of customized production, involving many variations in components and frame designs, each with a short product cycle.

In contrast, TIG welding does not require tooling and involves less preparation. Many interviewees mentioned that TIG welding techniques allowed flexible, low-cost, rapid product development that permitted wide-ranging experimentation without breaking the bank. If the experimentation had been done by lug brazing, it would have been very costly because of the need for new tooling and molds just to develop a prototype. The SMEs could not afford to do this. The flexibility provided by TIG welding led to huge experimentation with different designs of bicycle frames and was conducive to customization and flexible production.

The quick adaptation of TIG technology to MTB manufacturing is an example of a breakthrough by latecomers. The anecdotal account of how Atsu came to explore TIG welding for making front forks illustrates the logic of SME cost-down innovations in the context of decentralized production networks. Atsu used to work at Giant, the assembling factory. Then he left Giant and set up his own company, called 808, to make front forks for bicycles. As mentioned, parts suppliers in Taiwan are not bound to one single assembler. Atsu was actively diversifying his outlets by coming up with different designs that involved different specs for different customers (PL interview). When the idea of TIG welding was first introduced, he quickly explored the possibilities because it was a cost-effective way to realize his designs. Following a similar train of thought, other Taiwanese SMEs also quickly embraced the TIG technique.

The flexibility that TIG welding technology generated also reduced barriers for new entries. Interviewees concurred that small factories proliferated with the adaptation of TIG welding. An entrepreneur who was actively involved in the initial experimentation of TIG welding explained:

At that time, the demand was so great, like swarming bees. Even if you got all the orders, you would not be able to digest them all and produce all the bicycles. So you found some friends to work on the orders together. For instance, if my production capacity was 10,000 sets, but the order I received was way beyond my capacity, I needed to find someone else to help in the manufacturing. (FL interview, 2004)

TIG welding technology was conducive to a system of SMEs because people could venture into the business without investing heavily in equipment and tooling for product development. What was required instead was skilled labor—in this case, skilled TIG welders. The decentralized production network favored the rapid diffusion of the new techniques. Taiwan soon became the leader in lightweight MTB manufacturing, and TIG welding was the pervasive technology that was pushed forward by the Taiwanese makers.

### The Significance of TIG Welding: A Shift toward Customization of Designs

TIG welding technology is an example of how firms adopt a new production method that enhances flexibility and reduces production processes. Subsequently, the rise of MTBs and the adoption of TIG welding accelerated customization of designs, generated demand for lighter materials, and gave rise to a flexible production system that could accommodate frequent changes. With the introduction of the new technology, MTBs no longer looked like any standardized diamond-shaped bicycle. The Taiwanese manufacturers constantly emphasized that the flexibility of TIG welding and its ability to accommodate change were key considerations when embracing the new technology. An experienced engineer, specialized in welding, summarized its significance as follows:

What TIG welding meant for the Taiwanese bicycle industry was that we had mastered a core technology that was capable of meeting the demand for variety and changes. Because of TIG welding, we could manufacture bicycles with different designs and different sizes, and every bicycle could be different. . . . At the prime time of the Japanese-made bicycles, they were [made] with lugs and there were not many changes in bicycles. . . . In my view, what we have mastered is an ability of the technology to accommodate various changes and demands for customization. (BRL interview, 2011)

The breakthrough in TIG welding frames is an example of technological convergence and interdependence at work. The incremental and process innovations that stemmed from small accumulated improvements in welding skills, building upon Taiwanese manufacturers' prior experience in other forms of welding, contributed to quick adaptation in TIG welding. The breakthrough started by solving a particular problem regarding TIG welding chrome-moly frames. In this context, the technical changes occurred at the intermediate level, such as modifications and customization of welding machines for bicycle making. In a decentralized industrial system, the breakthroughs came from parts makers/specialist firms. In turn, the improvement in welding technology and adaptation to new materials established a series of improvements in bicycle production, including reduced costs and more efficient welding techniques. Subsequently, the technology cascaded and was embraced by most Taiwanese manufacturers. The adaptation and diffusion of TIG welding bicycle techniques led to a new wave of experimentation in bicycle design, encouraging customization, and thus new production methods, to meet the demand and exploring lighter materials, including aluminum.

## Aluminum Welding in Bicycles: The Advantages of a Decentralized Production System and Joint Collaboration

The rise of MTBs in the 1980s and the application of TIG welding to bicycles opened up new possibilities for experimentation with new designs and new composite materials to fulfill the demand for lighter materials that were sturdy and strong enough for MTB frames. Taiwanese manufacturers have continued to be the front runners in responding to these changes. For instance, they were the first to make aluminum-alloy frames using TIG welding techniques and were responsible for the subsequent breakthrough in different process treatments on various types of aluminum-alloy frames. Again, the technological convergence and interdependence central to these breakthroughs are relevant to understanding the parts makers' capacity for innovation (in terms of introducing and applying new manufacturing methods and materials through the recombination of existing methods). The decentralized industrial structure accounts for a particular technology being adapted and its subsequent rapid diffusion and spillover to other sectors.

By the mid-1980s, both the US company Trek and Taiwanese assemblers were exploring using aluminum alloy for bicycle frames, for aluminum was considered lighter and sturdier than chrome-moly steel. Trek was experimenting with aluminum bikes using aerospace bonding adhesive, whereas Taiwanese makers were heading toward TIG welding. Trek unveiled its aluminum model in 1985, but the frame made with adhesive was not strong enough (Crown and Coleman 1996: 137). The company ultimately converted much of its production to TIG welding. The option of bonding adhesives did not work out because of the limited availability of applicable adhesives and sparse efforts by fewer firms compared to the massive efforts in TIG welding pushed forward by the Taiwanese manufacturers.9

In the meantime, the leading Taiwanese assembler, Giant, had been experimenting with strong-strength carbon fiber as a material for lightweight racing bikes using lug technology.10 Yet it was lightweight TIG-welded aluminum-alloy frames (especially for MTBs) that became the dominant type for bicycles. How did the manufacturing technology associated with aluminum alloy win out? The availability of aluminum materials in Taiwan, where firms could tap the extensive supply networks across industries to develop local sourcing and technology development, played an important part in embracing aluminum for developing lightweight bicycles. Moreover, the organizational structure of decentralized networks in Taiwan was conductive to technology diffusion and adoption.

Here, the breakthrough in aluminum alloy frames came from a specialist firm rather than from leading assemblers by solving production problems at the intermediate input level: the aluminum alloy material. Many interviewees credited the frame-maker specialist A-Plus (a pseudonym) with the breakthrough of the first mass-produced aluminum frame, despite bicycle assemblers having been the first to explore the possibility.

In this case, the frame-maker A-Plus actively explored the knowledge available from various fields. The company first worked with Alcoa (Aluminum Corporation of America), a US aluminum materials supplier (not related to the bicycle industry), to acquire basic knowledge about aluminum alloy, the welding technique involved for aluminum-alloy tubes, and processing knowledge relating to heat treatment. The general manager of A-Plus who was involved in the development of aluminum MTB frames recalled the R&D process and the problems that other assemblers experienced:

When aluminum alloy as a material option first came around the market, most frame makers did not know much about the material or the possible technologies we could use to apply this material. Up to this point, aluminum alloy was used in some small bicycle parts, mostly in parts utilizing casting and forging, like the seat post, handle bar, or brakes. But we did not know much about how to weld aluminum frames. Some assemblers had imported aluminum tubes from Japan and applied welding but had failed. The problem was that the Japanese consultant/supplier of aluminum tubes had advised these assemblers to weld the tubes directly and that there was no need to apply heat treatment. Yet in this situation, without heat treatment, the welded aluminum tubes lacked strength and rigidity and tended to crack on the spot where they were welded. . . . The project was a failure. Nevertheless, those who participated learned the welding techniques.11 (AL interview)

He continued to explain how his company solved this technical problem:

My company then approached aluminum-alloy welding in two directions. We first studied the material and contacted Alcoa in the United States. Its affiliated company, called AlcoTec, is a welding rod supplier and specialized in welding technology. They provided us with knowledge on postweld complex heat-treatment processing and suggested an aluminum alloy, 7005, considered to be the easiest to process using heat-treatment technology. They made suggestions on welding materials, the welding equipment, and the specifications for heat treatment. The welding equipment company Miller is also their affiliated company. Subsequently, the sources of technology for aluminum-alloy frames have shifted from the Japanese to being US based. And now when we talk about aluminum-alloy materials in Taiwan, the code is in English, in contrast to the earlier code in Japanese. (AL interview)

Once this was established, the company began to explore the possibilities for local adaptation, though the initial aluminum material was imported from Alcoa.

The general manager of A-Plus believed the wide local availability of aluminum was a key consideration for them to opt for aluminum alloy materials:

Aluminum alloy is not a new material, but it has a lot of potential, much that has not been explored and realized. There are many aluminum-alloy products in Taiwan, from low-end products such as window frames and doors to various high-tech-related products. It is quite easy to find aluminum material suppliers in Taiwan, and it is easier to get them to shift to supplying other industries. In addition, the welding technique has been quite well developed, and that is our advantage [in terms of choosing a material that TIG welding could be applied to]. On the other hand, chrome-moly material was controlled by the Japanese and needed to be imported, and so was carbon fiber. (BPL interview, 2004)

This local accessibility of materials was important because it meant that suppliers of the material and equipment builders (the capital goods) could be found locally.

Subsequently, the development process of this new aluminum frame involved collaborative experimentation by suppliers working in different industries in Taiwan, illustrating how a specific technology was promoted and diffused among SMEs in a decentralized industrial system. Product development in the context of a decentralized SME-based system is also a matter of convincing others to take on the project. Given the nature of SMEs, the entry barriers for doing so need to be lowered so as to induce other SMEs to venture into the new field. As mentioned, many aluminum-related products were already manufactured in Taiwan, ranging from construction materials to kitchen and sports equipment (e.g., tennis racquets). Therefore, it was relatively easy to locate a variety of aluminum extrusion suppliers locally. The frame maker A-Plus turned to local aluminum material and extrusion suppliers who had experience in lower-grade construction materials, albeit not with the higher-grade materials with the strength and intensity needed for bicycle frames. Although these suppliers were unable to manufacture the seamless tubes needed for aluminum-alloy bicycle frames, what was important was that they were willing to take part in developing them (AL interview). This occurred during a downturn in the construction industry, upon which aluminum suppliers heavily depended, so they were eager to look for alternatives. Moreover, there were many smaller aluminum-extrusion suppliers to work with, making smaller-scale product development possible (AL interview).

The aluminum suppliers also benefitted from this collaboration because they then moved up to become suppliers of the higher grade of aluminum materials that had been used mostly in the aerospace industry but were now being applied to consumer industries such as bicycles, and they even became strong exporters of this material. This cascading upgrading and technology spillover process is an example of technology convergence at work whereby improvement occurred at the intermediate level from materials to equipment and cascaded to other uses. The general manager noted the cascading effects of this product development:

Since the introduction of aluminum bikes, Taiwanese extrusion plants started to produce seamless tubes, and subsequently Taiwan began to build its equipment for manufacturing seamless aluminum tubes and we have a complete local sourcing for aluminum tubes. . . . At that time, the demand was enormous and many extrusion plants that supplied bicycle frames all made a lot of money. They have mastered the technology and become aluminum tube exporters. (AL interview)12

Similar adaptation occurred for working with heat-treatment processing specialists. Prior to the development of aluminum frames, manufacturers had no prior experience in heat-treatment processes for bicycle frames, though there were heat-treatment specialists on smaller bicycle parts (MW interview, 2010; AL interview). Subsequently, the frame maker A-Plus worked with local equipment suppliers and heat-treatment specialists in developing the heat-treatment equipment that this new production method needed for bicycle frames (AL interview).

In the context of a decentralized industry system, R&D requires joint collaboration. The fact that parts makers are not constrained by a single customer means that it is easier for them to draw on resources from various industries for a solution to a given problem. And once the demand for a new material, process, or product is created, others are more willing to become involved with the venture. This explains how Taiwanese manufacturers quickly respond to such opportunities. Once a breakthrough takes place, it involves upgrading a whole supply chain, thus involving a cluster of innovations, as opposed to the new knowledge being kept inside the pioneering firm. While technology convergence may explain how technology cascades, the social structure, here the decentralized industrial system of SMEs, affected the type of technology that was adapted, applied, and extended in Taiwan in a particular time frame, as seen in the competing technology development of frame materials such as carbon fiber versus aluminum alloy.

## The Development of Hydroformed Aluminum Bicycle Tubes

The development and application of technology cascaded across different industries under this decentralized production network. This decentralized industrial system also influenced how engineers from public research institutes approached technology transfer and collaboration with the SMEs. Instead of a linear process of technology transfer to the targeted firm alone, the technology extension services took into consideration local spillover effects, integration, and joint development along the local supply chains. The hydroforming application for fabricating aluminum bicycle tubes is a case in point. Here, the technical changes, including initial breakthroughs and subsequent adoption, started with solving a problem in production technique that introduced changes in the intermediate input level, such as modification of the equipment. Once that was established, technical improvement cascaded and created backward linkages to other industries.

Hydroforming is a specialized die technology that applies high-pressure hydraulic fluid to press a material (e.g., aluminum or steel) into a die. This technology is a cost-effective way to create complex shapes while maintaining the strength and rigidity of the structure. Its other advantage is its savings on tools. Aluminum hydroforming technology has been increasingly used in the automotive industry in response to the growing demand for lighter materials. In the case of bicycles, hydroformed aluminum tubes have provided lighter and stronger bicycle frames with endless possibilities for aesthetically attractive designs (see Figs. 4, 7).

The Metal Industries Research and Development Center (MIRDC; 金屬中心) was one of the first initiators in Taiwan to acquire the manufacturing technology of aluminum hydroforming. A government-funded research institute, MIRDC has been crucial in working with SMEs on technology development and extension of new manufacturing methods to various metal- and processing-related industries in Taiwan. The principal investigator of the aluminum hydroforming project described how they chose the technology and how they came to work with the bicycle industry:

We observed the trend in applying aluminum hydroforming manufacturing technology in the automobile industry in our various overseas research visits. We started to explore the feasibility of introducing this manufacturing method in Taiwan around the year 2000. . . . Our goal was aiming at Taiwan's auto industry. However, it turned out that Taiwan's auto industry lacked the design capacities. . . . We then turned to the bicycle industry and spent a block of time to convince them to work with us on the project to explore the applications of this technology in bicycle tubes. (MPC interview, 2008)

This engineer articulated the complexity of interindustry applications and explained the development process:

Frankly speaking, the complexity of the development and prototyping process [in the bicycle industry] and the technology it had involved were far more difficult than we had anticipated. This technology was initially designed for auto parts that were built inside the vehicle chassis. Yet bicycle parts are directly exposed to the outside environment, and therefore considerations of various processing technologies are different. . . . For one, bicycle parts would require a more precise and lasting surface treatment. Initially, our ability to master that technology [of aluminum hydroforming] was quite poor. To make the matter worse, we did not know much about the bicycles, and we were unable to realize the design that the firm came up with. I must admit that we have learned a great deal from our collaborating firms. (MPC interview)

MIRDC approached the development of the technology by developing a local supply chain, as opposed to a mere top-down technology transfer to one firm. The first thing they tackled was building hydroforming equipment locally for the bicycle industry. The principal investigator described MIRDC's contributions as follows:

I would like to emphasize this particular point: the equipment [i.e., hydroforming press] for this particular manufacturing-process technology was very expensive, mostly from Germany or Japan. It would cost NT$100 million [US$3 million] at least. This costly equipment investment was a major barrier for SMEs should they wish to pursue this technology. So we approached the project by first developing the equipment locally, by tapping into the extensive supply network in Taiwan. . . . Once the equipment could be built locally at half of what it would cost to import from Germany, it would induce more SMEs to explore this technology. Then they could generate value-added by applying this new manufacturing method. . . . In fact, now smaller and locally built equipment could easily be built for the bicycle industry at minimal cost [around NT$10 million, or US$300,000], given that they would not need to use gigantic equipment designed specifically for heavy industries [e.g., a 1,100-ton hydroforming press to manufacture bicycle components vs. a heavy-duty 2,000-ton press]. (MPC interview)

Subsequently, the MIRDC formed a research consortium that consisted of material suppliers, mold-making specialists, processing specialists, equipment builders, bicycle tube makers, and bicycle assemblers. The subsequent breakthrough in commercialized manufacturing of hydroformed bicycle tubes was made by a bicycle tube specialist, not leading assemblers (PMC interview, 2003). Once this was accomplished, the MIRDC took the experience and worked with other industries. The engineer emphasized the necessity for technology diffusion and cross-industry fertilization:

Had we simply focused on transferring the technology to one specific firm, we would not have been able to generate such a cascading effect by inducing others to enter the field. This was especially true because this manufacturing technology could be extended to other consumer industries, such as furniture design, bathroom and kitchen fittings, and so forth. . . . We subsequently took what we had learned from working with the bicycle industry and worked with other consumer product industries. A variety of new designs could be realized with this new technology and benefit from the flexibility this manufacturing method could bring. (MPC interview)

This explains how technology developed and extended through this kind of cross-industry fertilization and learning and diffused rapidly from an initial public-private research alliance.

## Discussion

Contrary to the conventional assumptions that leading firms drive technological changes and that massive R&D expenditure and scaling up are required for a latecomer to catch up, the case study examined here shows how technological convergence and interdependence have facilitated technology learning and diffusion among Taiwanese SMEs in a system of decentralized production where parts makers/specialist firms connected to different supply networks are the center of the story. Process innovations that encourage flexibility and tool saving have been a constant theme among the breakthroughs for parts makers and specialist firms that make no final products. Thus, technological interdependence in a decentralized system provides a structural foundation for understanding SME-based technological learning given their relatively low R&D expenditures.

Moreover, the three examples described above in the bicycle industry—TIG welding technology, aluminum frame material, and hydroformed tubes—show that a key feature of the decentralized production of the SME-based model is that product development is not confined to a single firm; rather, it draws on external economies that integrate and work with different supply chains and networks. This accounts for the cross-industry fertilization among parts makers in different supply chains and explains how ideas cascade from one firm to another. These cross-industry linkages show how technical changes come in clusters of small improvements as opposed to a single radical innovation. This in turn creates a very different understanding of how technological learning and innovation come about: the unit of analysis is no longer the individual firm but the dynamics of a system of firms.

A direct implication is that collective effort is required to develop and promote a new technology or new design for an SME manufacturing system. For instance, in the case of the aluminum-alloy MTB frame, the frame-maker A-Plus acknowledged that, in addition to the presence of local material suppliers, a crucial factor in the breakthrough to the new design was the component maker's ability to develop an oversized headset—a bicycle component that connects the fork to the head tube of a frame (FC interview, 2004; AL interview). The rapid development of the aluminum-alloy bicycle in Taiwan was also due to sufficient interest on the part of firms, plus a relatively supportive infrastructure for realizing the project and then completing the supply chain with technologies ranging from new welding techniques to new designs, and from local material suppliers to government-supported welding testing and verifying facilities (BRL interview).

The need for joint collaboration means that the commitment of a single individual or sporadic effort may not be sufficient to convince others to take part in the new product development. The R&D comes in waves and does not work when a firm is going solo. This is supported by the experience of an interviewee who claimed that he had prematurely experimented with aluminum-alloy welding with imported materials in the early 1980s but was unable to push the development farther because of the lack of skilled aluminum-welding technicians. He was only able to find some welders from the Air Force in central Taiwan and some tinkers who repaired aluminum molds for the tennis racquet industry to do causal work as side jobs during their spare time (MW interview). As a single individual, he was unable either to realize the R&D needed for further development or to convince others to take part in the development, not to mention to develop local sources of materials, given the minimal demand. It was only later on that experimentation with aluminum TIG welding was able to materialize in an environment of competing experimentation and increasing interest in the market so that SMEs could convince other specialists to take part in the product development.

Following this logic, diffusion and dissemination of information are means to leverage technology development. Problem solving in a context of decentralized SMEs means that technology needs to diffuse in order to convince others to join in. The following remarks by interviewees illustrate the SME logic in terms of technology diffusion and leveraging joint R&D: “For instance, if I design a product and I think it is quite good, this is not enough, because I do not have the capacity to capture the market. If it is good, we need others to experiment with this so we can capture the market” (FL interview, 2004). As to why sharing the technology is needed (instead of keeping it under one roof to collect technology rent), an experienced bicycle-industry trader explained, “We are latecomers in the bicycle industry. Thus, if we want to out-compete Japan, the US, and Europe, we need to enlarge our scope and the scale of our products, and we need to develop more people/technicians” (MWR interview, 2003).

The experience of the Taiwanese bicycle manufacturers confirms the recent view that the technology shift from a proprietary closed system to modularity accounts for the rise of component makers and suppliers. Yet, the findings suggest that interfirm/industry collaborations are crucial in understanding the technological breakthroughs among the firms in a decentralized production network, contrary to the view of modularized production in an arm's-length market transaction where autonomous firms tap randomly into external economies.

Finally, the emphasis on a decentralized industrial system contributes to the call for bringing “social structure” back into the studies of STS. This case study of the bicycle industry in Taiwan shows that a decentralized production system affects the direction and choices of the technology being chosen, adapted, and transmitted, as can be seen in the various manufacturing technology advancements of MTBs, where technological convergence is not sufficient to explain the type of technology being chosen when competing technologies exist. Moreover, the flexible production system, enabled by TIG welding technology, influenced the subsequent direction of MTB development in terms of design, materials, and manufacturing methods.

## Acknowledgments

I thank the editors of EASTS and two anonymous reviewers for their insightful comments on earlier drafts of this article. I also acknowledge the valuable inputs by the participants at the Technology and Society workshop held at National Yangming University, Taipei, Taiwan, in September 2013, where I was introduced to and immersed in STS studies. Special thanks go to all my interviewees who shared their wisdom and experience with me. Finally, the earlier stage of the research on which this article was based was supported by the Canadian Social Sciences and Humanities Research Council and Fonds de recherche sur la société et la culture. The later part of the research was supported by the Research Project (99-2410-H-001-002-MY2), Ministry of Science and Technology, Taiwan.

## Notes

1

The debate on the shift to decentralized production and modularity has centered on market (modularity and arm's-length transactions) versus hierarchy (vertically integrated and scale and scope) in explaining technology development and organizational learning in the so-called new economy in the post-Chandlerian era. An alternative position advocates neither a hierarchy nor a market perspective by focusing on interfirm collaborations in understanding decentralized production (Sable and Zeitlin 2004; Whitford and Zeitlin 2004). As will be illustrated, this alternative position is relevant in accounting for the technology learning of the parts makers in Taiwan's decentralized production system as opposed to a scaling up of atomized production by individual firms.

## Notes

2

For instance, Rosenberg (1983: 63–65) uses the example of ship building and machinery to illustrate technological change, which involves a series of piecemeal improvements ranging from engine efficiency to changes in cargo-handling techniques and containerization that reduce turnaround time, improvements in materials, modifications for production techniques for greater convenience, and cost reduction in maintenance and repair.

## Notes

3

This is a startling difference compared with bicycles coming out of China, at an average of FOB (free-on-board) US\$40–50 since the mid-2000s, equivalent to the FOB price of Taiwan's bicycles in the early 1980s.

## Notes

4

Interviewees are indicated with either last names or name codes to ensure anonymity. The year of the interview is given at first cite for each interviewee.

## Notes

5

For instance, a good bicycle manufactured by Schwinn in the 1970s weighed over forty-two pounds, whereas the lightweight bicycle today is about seventeen to eighteen pounds.

## Notes

6

Atsu was the nickname of Mr. Yeh, whom most of my interviewees acknowledged as a somewhat manic technical genius, in some cases referring to him as “the MacGyver of Taiwan.” Mr. Yeh unfortunately passed away from a fire in his house in the mid-1980s. Thus, interviews on the breakthrough in TIG welding had to rely on technical personnel who had worked with him.

## Notes

7

According to the interviewee, the modification of the welding equipment for bicycle use contributed to improvement in production flow. For instance, the easy mobility of the equipment facilitated experimentation in production flows and was conducive to efficiency. Improvements in hands-on operation techniques and size reduced the failure rate and were conducive to diffusion of the techniques (FL interview, 2011).

## Notes

8

An experienced maker of bicycle frames pointed out that, prior to TIG welding, there were not many choices in terms of designs and sizes of bicycles. The bicycle frames were so standardized that it was up to the riders to accommodate the bicycles, rather than vice versa. For example, it was difficult for Asian bicycle racers to find an appropriately sized racing bike. Since the introduction of TIG welding, there are endless possibilities because everything can be TIG welded and custom made (FL interview, 2011).

## Notes

9

The adhesive-bonding technology promoted by Bevil Hogg, a cofounder of Trek, turned out to be a premature move (Crown and Coleman 1996: 138). A former Schwinn engineer noted that the adhesive technology was not viable in the 1980s because the crucial resin was not available for consumer industries at the height of the Cold War (ZC interview, 2011). This is an example of technology complementarity whereby a breakthrough in one process depends on a breakthrough in another process.

## Notes

10

Giant opted for R&D in carbon-fiber material because it thought aluminum would be simply a transitional product, as in the evolution of the development of the tennis racquet. Giant introduced the first carbon-fiber frame to the European market in 1987 as part of its effort to demonstrate its technological capacity to enter the global market using its own brand name. Yet carbon fiber did not become widely adopted until the early 2000s, using technology (bonding adhesive) different from that which Giant and the research consortium with the Industrial Technology Research Institute were pursuing in the late 1980s.

## Notes

11

Aluminum-alloy TIG welding involves a more advanced and precise welding technique. It requires quicker steps to weld since aluminum alloy needs to be welded at a lower temperature—overheated material would cause distortion in the aluminum structure. Aluminum welding uses slightly different materials, specifically for the welding rod (AL interview, 2014; MW interview, 2010; ZC interview, 2011).

## Notes

12

An aluminum supplier confirmed that his company's ability to manufacture seamless tubes had convinced many foreign buyers to approach them (PTC interview, 2004).

## References

References
Amsden
Alice H.
, and
Wan-Wen
Chu
(
2003
).
Beyond Late Development: Taiwan's Upgrading Policies
.
Cambridge, MA
:
MIT Press
.
Andersen
Poul H.
, and
Christensen
Poul R.
(
2005
).
Bridges over Troubled Water: Suppliers as Connective Nodes in Global Supply Networks
.
58
, no.
9
:
1261
73
.
Bijker
Wiebe E.
(
1995
).
Of Bicycles, Bakelites, and Bulbs: Toward a Theory of Sociotechnical Change
.
Cambridge, MA
:
MIT Press
.
Cheng
Lu-Ling
, and
Sato
Yukihito
(
1998
).
The Bicycle Industries in Taiwan and Japan: A Preliminary Study toward Comparison between Taiwanese and Japanese Industrial Development
.
Tokyo
:
Institute of Developing Economies
.
Chu
Wan-Wen
(
1997
).
Causes of Growth: A Study of Taiwan's Bicycle Industry
.
Cambridge Journal of Economics
21
, no.
1
:
55
72
.
Chu
Wan-Wen
(
2009
).
Can Taiwan's Second Movers Upgrade via Branding?
Research Policy
38
, no.
6
:
1054
65
.
Crown
Judith
, and
Coleman
Glenn
(
1996
).
No Hands: The Rise and Fall of the Schwinn Bicycle Company, an American Institution
.
New York
:
Holt
.
Ernst
Dieter
, and
Kim
Linsu
(
2002
).
Global Production Networks, Knowledge Diffusion, and Local Capability Formation
.
Research Policy
31
, no.
8–9
:
1417
29
.
Galvin
Peter
, and
Morkel
André
(
2001
).
Modularity on Industry Structure: The Case of the World the Effect of Product Bicycle Industry
.
Industry and Innovation
8
, no.
1
:
31
47
.
Gereffi
Gary
(
1994
).
The Organization of Buyer-Driven Global Commodity Chains: How U.S. Retailers Shape Overseas Production Networks
. In
Commodity Chains and Global Capitalism
, edited by
Gereffi
Gary
and
Korzeniewicz
Miguel
,
95
122
.
Westport, CT
:
Greenwood
.
Gereffi
Gary
(
2013
).
Global Value Chains in a Post-Washington Consensus World
.
Review of International Political Economy
21
, no.
1
:
9
37
.
Gereffi
Gary
,
Humphrey
John
, and
Sturgeo
Timothy
(
2005
).
The Governance of Global Value Chains
.
Review of International Political Economy
12
:
78
104
.
Helper
Susan
, and
Sako
Mari
(
2012
).
Management Innovation in Supply Chain: Appreciating Chandler in the Twenty-First Century
. In
Management Innovation: Essays in the Spirit of Alfred D. Chandler, Jr
, edited by
Lazonick
William
and
Teece
David J.
,
147
78
.
Oxford
:
Oxford University Press
.
Hsieh
Michelle F.
(
2011
).
Similar Opportunities, Different Responses: Explaining the Divergent Patterns of Development between Taiwan and South Korea
.
International Sociology
26
, no.
3
:
364
91
.
Humphrey
John
, and
Schmitz
Hubert
(
2002
).
How Does Insertion in Global Value Chains Affect Upgrading in Industrial Clusters?
Regional Studies
36
, no.
9
:
1017
27
.
Industrial Development Bureau
(
2007
).
2006 Industrial Development Yearbook of the Republic of China
.
Taipei
:
Industrial Development Bureau, Taiwan Ministry of Economic Affairs
.
Klein
Hans K.
, and
Kleinman
Daniel L.
(
2002
).
The Social Construction of Technology: Structural Considerations
.
Science, Technology and Human Values
27
, no.
1
:
28
52
.
Langlois
Richard N.
(
2000
).
Capabilities and Vertical Disintegration in Process Technology: The Case of Semiconductor Fabrication Equipment
. In
Resources, Technology, and Strategy: Explorations in the Resource-Based Perspective
, edited by
Foss
Nicolai J.
and
Robertson
Paul L.
,
199
206
.
London
:
Routledge
.
Langlois
Richard N.
(
2003
).
The Vanishing Hand: The Changing Dynamics of Industrial Capitalism
.
Industrial and Corporate Change
12
, no.
2
:
351
85
.
Lazonick
William
(
2009
).
Sustainable Prosperity in the New Economy? Business Organization and High-Tech Employment in the United States
.
Kalamazoo, MI
:
W. E. Upjohn Institute for Employment Research
.
Lazonick
William
(
2012
).
The Chandlerian Corporation and the Theory of Innovative Enterprise
. In
Management Innovation: Essays in the Spirit of Alfred D. Chandler, Jr.
edited by
Lazonick
William
and
Teece
David J.
,
89
121
.
Oxford
:
Oxford University Press
.
Norcliffe
Glen
(
1997
).
Popeism and Fordism: Examining the Roots of Mass Production
.
Regional Studies
31
, no.
3
:
267
80
.
Pinch
Trevor J.
, and
Bijker
Wiebe E.
(
1987
).
The Social Construction of Facts and Artifacts: Or How the Sociology of Technology Might Benefit Each Other
. In
The Social Construction of Technological Systems: New Directions in the Sociology and History of Technology
, edited by
Bijker
Wiebe E.
and
Hughes
Thomas P.
,
18
50
.
Cambridge, MA
:
MIT Press
.
Piore
Michael J.
, and
Sabel
Charles F.
(
1984
).
The Second Industrial Divide: Possibilities for Prosperity
.
New York
:
Basic Books
.
Rosenberg
Nathan
(
1963
).
Technological Change in the Machine Tool Industry, 1840–1910
.
Journal of Economic History
23
, no.
4
:
414
43
.
Rosenberg
Nathan
(
1983
).
Inside the Black Box: Technology and Economics
.
Cambridge
:
Cambridge University Press
.
Sabel
Charles F.
, and
Zeitlin
Jonathan
(
2004
).
Neither Modularity nor Relational Contracting: Inter-firm Collaboration in the New Economy
.
Enterprise and Society
5
, no.
3
:
388
403
.
Sturgeon
Timothy J.
(
2002
).
Modular Production Networks: A New American Model of Industrial Organization
.
Industrial and Corporate Change
11
, no.
3
:
451
96
.
Whitford
Josh
, and
Potter
Cuz
(
2007
).
Regional Economies, Open Networks and the Spatial Fragmentation of Production
.
Socio-economic Review
5
, no.
3
:
497
526
.
Whitford
Josh
, and
Zeitlin
Jonathan
(
2004
).
Governing Decentralized Production: Institutions, Public Policy, and the Prospects for Inter-firm Collaboration in US Manufacturing
.
Industry and Innovation
11
, no.
1–2
:
11
44
.
WTO and IDE-JETRO (World Trade Organization and Institute of Developing Economies, Japan External Trade Organization)
(
2011
).