The Competitiveness of Nations

in a Global Knowledge-Based Economy

April 2003

AAP Homepage

Wendy Faulkner

Conceptualizing Knowledge Used in Innovation:

A Second Look at the Science-Technology Distinction and Industrial Innovation

Science, Technology, & Human Values

Volume 19, Issue 4

Autumn 1994, 425-458.




The Science-Technology Distinction

The Early Debate: Blurring the Boundaries

Historical Studies of Technological Knowledge

The Distinctiveness of Technological Knowledge

Knowledge Used in Industrial Innovation

Knowledge from Internal and External Sources: The Evidence

Why Companies Use Internal and External Sources of Knowledge

Categorizations of Knowledge Used in Innovation

Broad Distinctions in the Character of Knowledge Used in Innovation

Detailed Categories of Knowledge Used in Innovation

Some Comments on the Composite Typology





This article reviews empirical and conceptual material from two distinct research traditions: on the science-technology relation and on industrial innovation.  It aims both to shed new light on an old debate - the distinction between scientific and technological knowledge - and to refine our conceptualizations of the knowledge used by companies in the course of research and development leading to innovation.  On the basis of three empirical studies, a composite categorization of different types of knowledge used in innovation is proposed, as part of a broader framework encompassing two further taxonomic dimensions.  It is hoped that this typology and framework might provide useful research tools in furthering our understanding of the knowledge transfers and transformations that occur in the course of innovation.  It could also prove useful for organizations and groups facing difficult strategic choices about technology.



This article joins other recent attempts to conceptualize technological knowledge and expertise [1] used in the course of research and development (R&D) [2] and design activities leading to innovation.  It draws together conceptual and empirical threads from two distinct research traditions in science and technology studies: on the science-technology distinction and on industrial innovation.  My primary objective is to propose a composite categorization of different types of knowledge used in innovation, which, it is hoped, will prove useful as a research tool.  As a secondary aim, I also consider

AUTHOR’S NOTE: An earlier version of this article was presented and discussed at the University of Edinburgh Programme on Information and Communication Technologies (PICT) workshop “Exploring Expertise” held in November 1992, and I am grateful to colleagues who provided feedback on that occasion.  I would also like to thank David Edge, Keith Pavitt, Jacqueline Senker, and Andrew Webster for their time and effort in reading and commenting on earlier drafts of the article.


whether we can distinguish between scientific and technological types of knowledge, and to what extent each is used in the course of industrial innovation.

The term knowledge is used here in its broadest sense, to encompass what we call knowledge, expertise, skills, and information.  Of course, the processes by which scientific and technological knowledge is created and deemed legitimate are very political in nature - witness the institutional authority associated with medical knowledge, for example.  My main concern in this article is more narrowly focused, on the cognitive or epistemological features of the knowledge used in innovation.  These features are never simply “internal,” however.  They are intimately related to questions of who possesses particular knowledge and how easy it is for particular groups to access and make use of this knowledge.  Indeed, my hope is that a more sophisticated characterization of technical knowledge - one that identifies specific and substantive differences of type - will further our understanding of what happens to knowledge as it moves between, and is developed by, different groups in the course of innovation.

My interest in this project arose out of a particular concern with the flows of knowledge between public sector research (PSR) organizations and industry, which was the subject of a recent study I conducted with Jacqueline Senker and Lea Velho. [3]  This study sought to identify the particular types of knowledge obtained from academic and government laboratories by firms in three different technological fields, in order to understand variations in the extent and nature of industry-PSR research linkage in different fields.  The findings of this study are reported fully elsewhere (Senker and Faulkner 1992; Faulkner, Senker, and Velho 1974).  However, I present some of our findings here, together with some findings from a similar study by Gibbons and Johnston (1974), to examine the range of knowledge types used in the course of innovation, and to illustrate the general applicability of a research approach that utilizes detailed categorizations of knowledge.

The first section of the article reviews literature on the science-technology relation, including work in the history of technology that highlights a number of important cognitive and epistemological features of technological knowledge.  The second section reviews literature on knowledge used in industrial innovation, focusing on the contribution of internal and external sources of knowledge used in innovation, and on broad differences in the types of knowledge obtained.  The following section compares various attempts to categorize technological knowledge and proposes a composite typology of knowledge types.  Possible research and policy uses of this typology are outlined in the conclusions.

426 Index


The Science-Technology Distinction

The science-technology relation was the subject of recurring debate in the field of science studies, more or less from its inception until the early 1980s.  During the 1970s interest in the science-technology relation was largely superceded by a concern with technology per se, in particular, a concern to characterize more adequately the nature of technological knowledge (Staudenmaier 1985).  Ironically, whereas the early literature tended to stress the increasing proximity and overlap of science and technology, the latter work has tended to highlight the distinctiveness of technology.


The Early Debate: Blurring the Boundaries

The early literature challenged the prevalent linear model of the science-technology relation.  Within this model, science is the “springhead” of innovation, as if scientific discovery necessarily implies technological invention, whereas technology involved the rather humdrum, responsive activity of applying science (see Barnes and Edge 1982 for a review).  Critics of this model pointed to the numerous occasions - not least, the advent of the steam engine - when technology “led” science (e.g., Layton 1988).  Even where the reverse appeared to hold, detailed case studies revealed that the linearity was illusory.  For instance, although the scientific quantum theory of semiconduction was a necessary precondition for the invention of the transistor, the theory itself did not suggest the technology: rather, the transistor arose primarily out of the development of rectifier technology within the fields of radar and radio (Gibbons and Johnson 1982).  The linear model was thus shown to be fundamentally flawed in its perception of both science and technology.

The alternative, “two stream” model, championed by Derek de Solla Price, accords better with the historical evidence.  On the basis of extensive citation analysis of science and technology publications, de Solla Price (1965) concluded that science tends to build on old science and technology on old technology, but there is a weak and reciprocal interaction between the two.  He argued that maximum interaction occurs during the period of training when budding scientists and technologists read the archival literature of their respective endeavors, “packed down” in textbooks.  Accordingly, the education cycle accounts for a time lag (of approximately ten years) in the translation of new science into new technology, and vice versa (de Solla Price 1965).

Later, de Solla Price revised his views somewhat.  He argued that “basic and applied research are linked inseparably to technology by the crafts and


techniques of the experimentalist and inventor,” and proposed the term instrumentalities “to carry a general connotation of a laboratory method for doing something to nature or to the data in hand” (de Solla Price 1984).  A pertinent example is Rosalind Franklin’s ability to make good X-ray diffraction pictures from very small samples of poorly crystallizable material: without this instrumentality, Watson and Crick would not have been able to “see” the double helix.  De Solla Price cited numerous instances when the advent of such instrumentalities has simultaneously opened up major new opportunities for scientific investigation and technological innovation. [4]  At such times at least, he conceded, any time lag in the interaction between science and technology may be very short indeed.

More recent bibliometric analyses reveal just how short the time lag can be.  In some technological fields (for example, biotechnology) the citation behavior of patent applicants (and examiners), in terms of the frequency of references to basic research publications and the time distribution of these citations, is very similar to that of researchers in neighboring scientific fields (Carpenter, Cooper, and Narin 1980; Carpenter, Narmn, and Woolfe 1981; Narin and Noma 1985).  Thus some technologies are strongly science related.  Far from relying on archival literature, technologists keep up with the “research front” literature in science. [5]  In a study of solid-state technology publications between 1955 and 1975, Marvin Lieberman found that the frequency and age of scientific citations followed a pattern of “overlapping waves” that, he concluded, were associated with “the continual birth of new science-related technologies within the solid state field” (Lieberman 1978).

Although this pattern explains the generally high level of coupling between science and technology evident in many advanced fields, it also suggests that such coupling is greatest during the early stages of development of a technological field.  This theme is evident in economic theories of technological development, even though these theories do not explicitly model the science-technology relation (Dosi 1982; Granberg and Stankiewicz 1978).  Empirical evidence that industrial innovation and growth tend to be “knowledge-led” during the early development of an industry is provided by Vivien Walsh’s (1984) detailed examination of time trends in scientific publication, patenting, and output in the chemical industry.

The emergence of the research-based chemical and electronics industries (Freeman 1982, chaps. 1-3), and of the more science-related technologies identified above, indicates that science and technology have become increasingly intimate endeavors during this century.  This trend has its historical roots in the establishment of R&D laboratories in industry and of specialist science and engineering departments in universities.  Ironically, these developments involved an institutional separation of science and technology.  Hendrick

428 Index

Bode (1965) argued that science was able to overlap with, and contribute to, technology in the areas of theory, experimental technique, and specialized knowledge precisely because it had earlier achieved its own momentum.  At a cognitive level, the institutional changes signaled “a more profound and subtle sort of science... It meant that by digging deeply enough we could expect to turn up new phenomena and new relationships not readily predictable from ordinary experience” (Bode 1965, pt. 2).  So, modern scientific inquiry demands not only lengthy training in the relevant specialisms but also the use of sophisticated equipment and instrumentation (Bud and Cozzens 1992).

Taken together, the case studies and bibliometric evidence lead to two conclusions about the science-technology relation.  First, it is a strongly interactive relationship between two semiautonomous activities, with instrumentalities as an important area of overlap.  Second, science and technology are now particularly intimate activities, at least in some new fields and at times of major change.  This latter conclusion obliges us to accept a blurring of the boundaries between science and technology as these terms are conventionally understood.  Otto Mayr sums up the issue:

[although] a practically usable criterion for making sharp and neat distinctions between science and technology simply does not exist... the two words “science” and “technology” are useful precisely because they serve as vague umbrella terms that roughly and impressionistically suggest areas of meaning without precisely defining their limits. (Mayr 1982, 159)

We are left, it seems, with only nuances of meaning to distinguish science and technology.  Indeed, for some, these boundaries are all but obliterated - witness Bruno Latour’s use of the term technoscience (Latour 1987).  It seems pertinent, therefore, to review what others have to say about what makes technological knowledge distinct.


Historical Studies of Technological Knowledge

The demise of the science-technology debate coincided with the shift of attention within the field of science studies toward the subject of technology.  “Revisionism” in the understanding of the science-technology distinction has come from scholars of technology, in particular, historians of technology.  They have provided a strong empirical basis for the critique of the linear model, and much evidence about the nineteenth-century changes through which technology is held to have become more closely science related.

Edwin Layton’s (1974) article “Technology as Knowledge” effectively launched a project to elucidate the nature of technological knowledge.  This


project has focused largely on the emergence of engineering as a distinct academic discipline (e.g., Layton 1976; Channell 1982; Constant 1984a), and we should be careful not to conflate engineering and technology.  Walter Vincenti (1991, 6) argues that, whereas technology properly includes draftspeople, shopfloor workers, and so on, engineers have a special relationship to technology.  In line with the commonly accepted definitions of engineering, he identifies engineers as those who organize, design, construct, and operate artifacts that transform the physical world to meet a recognized need.

Studies of engineering have examined various aspects of the development of engineering curricula in universities, including the adoption of scientific methods and principles, and the balance of theoretical and practical training in different countries.  David Channell (1982) concluded that “engineering science” occupies an intermediate position between “pure” engineering and science, by translating knowledge and techniques from one to the other.  He thus rejects the assumption (made, for example, by Musson and Robinson 1969) that engineering simply “lifted” the scientific methodology.  Layton (1976) and Channell (1982) both stressed that the closer organization and interplay of science and engineering were ideological as well as practical expedients; engineers were keen also to acquire some of the image and status that nineteenth-century science enjoyed.  Indeed, in a pertinent but rather facetious critique of Layton’s and Channell’s work, Fores (1988) argued that the transformation associated with the emergence of modern engineering has been grossly overstated, precisely because of the “totemic” qualities of the label “science.” [6]

Layton’s project has been considerably furthered by the engineer-turned-historian Walter Vincenti.  In What Engineers Know and How They Know It, Vincenti starts from the now traditional view within the history of technology that engineering is not a derivative of science but an autonomous body of knowledge that interacts with it (Vincenti 1991, 1).  He seeks to develop an epistemology of engineering based on a series of detailed empirical studies of the growth and maturation of aeronautical engineering between 1900 and 1950.  At the beginning of this period, aircraft design was largely “cut and try”; by the end, there existed a substantial body of underpinning theory, experimental techniques, and data.  As a result of these developments, Vincenti argues, the level of uncertainty surrounding design declined dramatically because “acts of [communicable] skill” increasingly reduced reliance on “acts of insight,” and aeronautical engineering is now a more or less systematic and cumulative body of knowledge (Vincenti, 1991, 168).

A number of important conclusions emerge from Vincenti’s work.  First, an extended learning process is often necessary if the requirements of the

430 Index

user are to be understood and integrated into design specifications.  For example, it took twenty-five years of close interaction between engineers and pilots before flying quality specifications could be drawn up.  The “hands-on” experience of pilots had to be translated into codified design parameters.  Second, much of the practical experience and knowledge required was local and tacit in character: shopfloor experience with different flush riveting techniques, for example, could not be easily codified or communicated between firms.  Third, the generation of data was vital to improved analytical capability and this required the emergence of vicarious testing techniques to reduce reliance on the always costly and sometimes disastrous field trials.  A key aeronautical example here is the use of wind tunnels for simulation tests.  Fourth, certain theoretical tools and ways of thinking (for example, experimental parametric variation) were crucial to this development.


The Distinctiveness of Technological Knowledge

Historical work on engineering has highlighted rather than blurred the distinctions between science and technology.  This cannot be entirely due to historians having a retrospective orientation, because scholars of modern technology reach similar conclusions (e.g., Sørenson and Levold 1992).  Drawing on both the historical and contemporary literatures, we can point to three closely related areas in which technology is distinguished from science: (1) in its purpose or orientation, (2) in its sociotechnical organization, and (3) in its cognitive and epistemological features.

With respect to orientation, there are familiar nuances of difference: technology is about controlling nature through the production of artifacts, and science is about understanding nature through the production of knowledge.  Mayr (1982, 159) argued that, although this distinction does indeed separate science and technology, “it is valid only on the level of semantics.  If we analyze actual historical events, we find that the motives behind actions are usually mixed and complex.”  Vincenti nevertheless privileged the distinction: “However phrased, the essential difference is one between intellectual understanding and practical utility” (Vincenti 1991, 254).  In a similar vein, Layton (1988) described engineering, medicine, and agriculture as “technological sciences” because they involve the “science of the artificial” in contrast to the “basic sciences” of the natural.

The practical-artifactual orientation of technology has important consequences for its organization, both as a body of knowledge and as a social activity.  With respect to the latter, Edward Constant (1984b) and others have stressed the hierarchical features of the development of complex technologi


cal systems.  Once a project specification has been drawn up and an overall concept design selected, the task of detailed design is “decomposed” according to major components of the system, specific problems and subproblems, and specialist disciplines.  The technological project is thus highly structured, but in a way that allows maximum interaction and coordination between specific groups.  Development of the design is coordinated and iterative, and the end product succeeds in integrating all of the necessary knowledge.

Scholars of modern technology, working in a constructivist framework, also stress the socioinstitutional complexity of technology, but they argue that this extends far beyond the laboratory or individual company.  The recent article of Sørensen and Levold (1992, 19) is particularly helpful here.  They synthesize John Law’s concept of heterogeneous technology with Constant’s hierarchy, arguing that the heterogeneity of technology is evident both in the terrain of the technoscientific - that is, the decomposition into problems requiring expertise from various groups - and in the sociotechnical - that is, “how they [problems] are analyzed and integrated” (Sørensen and Levold 1992, 19).  Sørensen and Levold conclude that technology is far more complicated than science on both counts; “technology is usually surrounded by a larger number of powerful political and economic and political actors than is science... [thus]... science involves less of the social, and the social terrain on which scientists manoeuvre is much simpler than that of engineers” (Sørensen and Levold 1992, 16). [7]

Science can be distinguished from engineering in terms of five distinct cognitive and epistemological features.  First, because of its practical-artifactual orientation, the central activity in technology is design.  In practice, design only sometimes demands the generation of new knowledge.  Although design always enters into R&D, much of it is quite distinct from R&D both institutionally and cognitively (Walsh et al. 1992).  According to Sørensen and Levold:

Technology... is far more than what can reasonably be subsumed under the concept of engineering science.  Development of technology still involves activities better described by the metaphor of art than of science.  Practical intuition and a developed “engineering gaze” are frequently more important than calculation and analysis. (1992, 19-22).

A second and related distinction is Sørensen and Levold’s point that problem solving in technology is a more heterogeneous activity than it is in science. [8]  Most science is also more homogeneous than technology in terms of disciplinarity, expertise, and social groupings, with the result that knowledge claims in science are generally far less heterogeneous than are innovations in technology.

432 Index

A third distinction, which has been widely recognized since Polyani’s (1966) book on the subject, is the vital importance of local and tacit knowledge in technological innovation (Senker 1993; Winter 1987).  This is demonstrated, for example, in Vincenti’s (1991, chap. 5) study of flush riveting.  James Fleck (1988) showed that the development of complex information technology (IT) systems demands extensive knowledge of the contingencies operating in user organizations.  The picture that emerges is in stark contrast to the presumed universality of scientific knowledge.  Of course, the work of Harry Collins (1974) shows that tacit knowledge is also important in scientific experiments.  Sørensen and Levold argue, nevertheless, that its significance is far greater in technology because replication of reported experiments is not widespread in practice, [9] and because failure to replicate in science merely raises questions about a knowledge claim, whereas failure of an artifact can have disastrous social and economic consequences.

The final two areas of distinction are more problematic.  They are the role of theory and the character of methodology in technology.  Another of the nuances of meaning commonly attached to science and technology is that the former is more theory based and the latter more empirical.  As the debate between Fores and Layton and Channell indicates, this assumption is highly contestable.  In any case, it would be wrong to assume that all theory is necessarily scientific.  Vincenti argues that the theoretical tools in engineering lie on a science-engineering spectrum.  At the “essentially scientific” end are the purely mathematical tools, followed by mathematically structured theoretical knowledge about the physical world.  Such theories generally originate in science and attract scientific interest for their explanatory powers.  However, they generally need to be reformulated or “recast” in order to make them applicable to technological problems.  The “essentially engineering” end includes theory “based on scientific principles but motivated by and limited to a technologically important class of phenomena or even a specific device” (Vincenti 1991, 214).  Interest in such theory depends entirely on the utility of the artifact to which it relates.  At the far end of the spectrum, Vincenti identifies phenomenological theory, based primarily on ad hoc assumptions (presumably derived from trial and error practice) and only marginally on scientific principles.  The explanatory power of such theory is limited, although its practical utility is high.

In the area of methodology, most scholars see little to distinguish science and technology.  Sørensen and Levold (1992), for example, noted that there is much methodological variety within both science and technology.  Constant nonetheless argues that modern technology may be distinguished from both craft technology and science by the application of a methodology based on “bold total-system conjecture and rigourous testing to large-scale, complex,


multi-level systems” (Constant 1984a).  Vincenti similarly identifies what he calls “characteristically engineering methodology” suggestive of something akin to scientific methodology (1991, chap. 5).  He cites the use of experimental parametric variation and scale models to test aircraft propellers - activities that took place independently of physical theory and that provided vital data for design and analysis where no useful theory existed to predict performance. [10]  Like Channell (1988), Vincenti (1991, 168) concluded that although elements of engineering methodology appear scientific, engineering methodology as a whole did not emerge within science.

We should remember that there are important respects in which science and technology are still generally held to be similar.  Both conform to the same natural “laws.”  Both are cumulative and diffuse largely through the same mechanisms: education, publications, and informal communication.  And both are organized around professional communities with marked disciplinary autonomy.  Nonetheless, the studies cited here point to some quite significant distinctions between these two activities and associated bodies of knowledge - distinctions that hinge on the practical and artifactual orientation of technology, and include a number of socioinstitutional and epistemological differences that flow from this orientation.  I would argue that this work does indeed oblige us to rethink earlier conclusions about the apparently vanishing boundaries between science and technology.  Further support for this position comes from scholarship on knowledge used in industrial innovation, to which we now turn.


Knowledge Used in Industrial Innovation

Our concern here is with knowledge actually used in the process of innovation.  To the extent that innovation relates to artifacts and not understanding per se, it is decidedly technological.  However, I am not assuming that the knowledge used in innovation is exclusively technological, hence the formulation adopted here.  Indeed, it is clear that scientific knowledge also contributes to innovation.  This section summarizes literature that seeks to calibrate and explain the relative contribution of internal and external sources to the knowledge used in innovation.  The studies reported here are based on quite different research questions and methodologies from those described above.  They are contemporary rather than historical studies, with a specific focus on the innovating firm rather than on science or technology more broadly.  In spite of this, both literatures point to similar conclusions about the types of knowledge used in innovation.

434 Index

Knowledge from Internal and External Sources: The Evidence

The early concern in science studies with the science-technology relation was paralleled in the new field of science policy by a concern to assess the contribution of public “science” to technological innovation.  This concern underpinned the celebrated but now widely discredited retrospective studies such as TRACES (Illinois Institute of Technology Research Institute 1968; Batelle Columbus Laboratory 1973) and Project Hindsight (Sherwin and Isenson 1967). [12]  It also informed many of the early innovation studies of the 1970s.  In contrast to the retrospective studies, this research focused on the knowledge actually used in the course of innovation, rather than knowledge that in some abstract and convoluted way might have contributed to it.  The innovation studies sought to identify the main institutional source of the original idea for the innovation under investigation and of the major technical inputs to subsequent problem solving (see Rothwell 1977 for a review).

Averaging across industries, there is remarkable convergence in the results of these studies.  Around two-thirds of the knowledge used by companies in the course of innovation derives from their own in-house R&D effort and expertise; the remaining third comes from external sources.  The largest single external source of scientific and technological contributions to innovation is other industrial companies, especially users or suppliers but also competitors.  The contribution of academic and government laboratories varies across sectors from 5 percent to 20 percent (Rothwell 1977).

The early innovation studies revealed that the translation of new knowledge into new artifacts is an extremely complex process; that the relationship between academic and industrial research is neither obvious nor direct; and that innovation demands knowledge from a range of internal and external sources.  Most significantly, success in industrial innovation rests on the effective organizational “coupling” of technical and market opportunities and intelligence (Rothwell et al. 1974; Freeman 1982, chap. 5).  Thus management capability is required in a range of areas, not simply in research.  The challenge posed by innovation now tends to be seen more as one of organization than of intellect, and this has become the central preoccupation of the innovation literature.  Perhaps as a result, there has been little concern to examine further the cognitive features of industrial innovation.

An important exception was the study by Gibbons and Johnston (1974) of thirty award-winning innovations.  It sought to assess the particular contribution of “public science” to innovation by asking industrial R&D staff to identify all of the scientific and technological “information” used by them in


Table 1. Content of Scientific and Technological Input (STI)

to Innovation by Source






Content of Knowledge

Info Units



Companies a

PSR b **

Theories, laws, and general principles

8% (69)

52% (36)

16% (11)

32% (22)

Properties of materials, and components

32% (270)

74% (200)

16% (42)

10% (28)

Design-based information, Operating principles

24% (205)

81% (165)

15% (30)

5% (10)

Test procedures and techniques

10% (78)

80% (62)

12% (9)

9% (7)

Knowledge of knowledge

26% (217)

57% (124)

30% (66)

12% (27)


100% (839)

70% (587)

19% (158)

12% (94)

Source: Gibbons and Johnson 1974.

a. “Other companies” includes here the trade and technical literature, plus contacts with organizations such as British Standards and Research Associations.

b. PSR = public sector research, which is described by Gibbons and Johnston as or “public science” and defined as scientific journals, books, and so on, as well as personal contacts in government and academic laboratories.

* percentage of information units; numbers in parentheses.

** PSR is referred to as “public science” by Gibbons and Johnston, and defined as Public Sector Research.


the course of new product development.  This yielded 887 units of information that were then analyzed in terms of the content of the information, the sources from which that information was obtained, and its impact on problem-solving activity.  Table 1 summarizes their data on the content of information obtained from different sources. [13]

The results in Table 1 can be compared with those in Table 2, which analyze the “impact” of knowledge from different sources on different areas of companies’ innovative activities, grouped under six broad headings. [14]  These data were generated in our recent study based on 44 interviews with R&D staff in 23 firms, covering three fields of technology.  This study investigated the knowledge flows or scientific and technological inputs (STI) associated with industry-PSR linkage activity.  Following a similar approach to that of Gibbons and Johnston, it examined the type, source, and impact of

436 Index

Table 2. Impact of Scientific and Technological Input (STI) on Innovative Activities








Other Companies


NR b

Future innovations

57% (52.0)

33% (30.0)

10% (9.0)


Search activity





Scouting for new applications

45% (13.5)

27% (8.0)

28% (8.5)


Scanning research frontier

30% (8.5)

18% (5.0)

52% (14.5)



37% (22.0)

22% (13.0)

40% (23.0)


Ongoing R&DC





Underpinning knowledge

40% (12.5)

2% (0.5)

58% (18.0)


Routine problem solving

88% (28.0)

9% (3.0)

3% (1.0)



65% (40.5)

6% (3.5)

30% (19.0)







Research equipment

18% (5.5)

52% (15.5)

30% (9.0)


R&DC procedures

47% (14.5)

24% (7.5)

29% (9.0)


Skills in experimentation and testing

55% (16.5)

17% (5.0)

28% (8.5)



40% (36.5)

31% (28.0)

29% (26.5)



51% (23.5)

46% (21.0)

3% (1.5)


Technical backup

73% (14.5)

10% (2.0)

18% (3.5)


Overall total

51% (189.0)

27% (97.5)

22% (82.5)


Source: Faulkner, Senker and Velho (1994).

a. PSR Public Sector research.

b. NR = nonresponses; this includes a small number of responses that gave equal weight to all three sources.

c. R&D = research and development.

* percentage of responses; numbers in parentheses.

** PSR is referred to as “public science” by Gibbons and Johnston (1974), and defined as Public Sector research

STI to innovation. [15]  The major difference is that we did not attempt to quantify the types of knowledge used in innovation.

The two studies have a crucial common feature: both start from an analysis of firms’ total knowledge requirements (or use) as the most appropriate basis for assessing the particular contribution of PSR.  This makes the findings in Tables 1 and 2 interesting for two reasons.  First, they reveal that companies obtain different types of knowledge from different sources.  Second, they provide a quite detailed picture of the full range of knowledge types utilized in the course of R&D leading to innovation.


The dominant contribution of internal sources to knowledge used in innovation is confirmed by the data in both tables. [16]  Researchers we interviewed reported almost unanimously that what they called tacit skills, acquired largely on the job, make a greater overall contribution to innovation than does formal knowledge, acquired from literature and education.  Further questioning revealed that tacit knowledge is also obtained from other companies and from PSR (Senker and Faulkner 1993).  Significantly, though, industrial R&D activities demand a synthesis of these diverse contributions from both internal and external sources.

Table 1 shows that internal sources make a particularly high contribution to design and to test procedures and techniques, and contribute substantially to properties of materials and components.  Similarly, in Table 2, internal sources dominate in routine problem solving (as well as technical backup) and contribute substantially to skills in experimentation and testing.  Thus the type of knowledge obtained from internal sources is primarily associated with the core activities of R&D and design.  Gibbons and Johnston found that half of all knowledge from internal sources is collectively generated, as a result of in-house activities (mostly experimentation and analysis), and half is personal in the sense that it is already known to the individual researcher, as a result of previous education and work experience. [17]

Instrumentalities are an important aspect of R&D.  Table 2 shows that the impact of internal sources in this area is slightly less than average.  This, in part, reflects the fact that other companies make a major contribution to research equipment (and a relatively high contribution to production and knowledge of knowledge).  A second factor is the slightly greater-than-average contribution of PSR to the procedures and skills used in R&D (although this is not particularly evident in Table 1).  Other evidence from our study highlights the practical help with experimentation provided by contacts in PSR, which supports de Solla Price’s contention that instrumentalities are an important area of overlap between academic and industrial research.

Table 1 indicates that internal sources contribute less to theory than to any other category of knowledge, although still more than external sources.  Conversely, theory is the only category in which PSR makes a greater-than-average contribution: one-third of all theory-related knowledge comes from this source.  This is consistent with our finding that PSR contributes most significantly in two areas: scanning the research frontier and underpinning knowledge. [18]  In Table 1, design is the category to which PSR contributes least, whereas Table 2 indicates that PSR can contribute materially to product design and development (at least in the more design-based fields) and to scouting for new applications.  Nonetheless, our finding that PSR has only a very minor impact on future innovation supports the conclusion of Gibbons

438 Index

and Johnston that “public science” is generally “not the springhead of innovative ideas” (1974).

It is not possible to separate scientific and technological knowledge within the various categories used in these two studies.  However, the data presented here do shed light on the relative contribution of each to innovation, which is relevant to the science-technology debate outlined earlier.  First, it is clear that scientific knowledge (as defined earlier) is a part of the theory, knowledge of properties, and methodology used in the R&D that leads to innovation.  Second, at least some of this scientific knowledge comes from internal sources (and, sometimes, other companies): there is a considerable in-house contribution in the areas of theory, underpinning knowledge, scanning the research frontier, and instrumentalities.  Third, PSR also contributes some technological knowledge through its input to research equipment and, in some fields, to engineering research and design.

Biotechnology is an interesting case in point because the boundaries between science and technology appear particularly indistinct in this field.  Yet our study and my own earlier research (Faulkner 1986, 1989) reveal that even here the nuances of distinction between scientific and technological knowledge still have some meaning.  Significantly, industrial researchers in this field - unlike those in advanced ceramics and parallel computing -  unambivalently call themselves scientists.

The various threads of evidence presented here lead to three broad conclusions.  First, with regard to the overall knowledge used in innovation, the dominant contribution of internal sources is confirmed; this knowledge is primarily associated with design and R&D activities.  Second, the more significant contribution identified as coming from PSR is in research rather than in design and development; PSR contributes new knowledge, theoretical knowledge, and knowledge related to research techniques, which is consistent with the idea that the balance of frontier research occurs within PSR.  Third, these studies clearly show that industrial research is by no means exclusively concerned with technological knowledge, any more than public sector research solely constitutes “public science.”


Why Companies Use Internal and External Sources of Knowledge

Keith Pavitt (1984) noted that by privately funding R&D activities, companies add to the total stock of knowledge as well as drawing on knowledge that is publicly available.  Three factors explain the dominant role of internal knowledge.  First, firms need to appropriate technology related to specific artifacts, normally by patenting or trade secrecy, in order to extract a reasonable rent from them.  When a product is a radical improvement on the


existing ones, obtaining proprietary advantage may even secure the innovator monopoly profits for a while.  Appropriating external technology is often necessary but problematic.  The experience of technology transfer reveals that ownership of intellectual property alone is inadequate, because additional tacit knowledge and skills are generally needed in order to effect the transfer. [19]  And in this, as in other forms of external knowledge acquisition, an understated but significant irony is that companies must have some related in-house expertise if they are to make sense of and to fully exploit external knowledge (Gambardella 1992).

This relates to the second reason for the dominant role of internal sources of knowledge in innovation, namely the cumulative nature of technological development.  This is a recurring theme in economic histories of technology, and one I would stress. [20]  At the level of technological fields, the cumulative nature of development is reflected in “path dependence” whereby one development appears to suggest the next.  Path dependence is captured in the now common use of the terms technological trajectories and paradigms (Dosi 1982). [21]  Knowledge acquisition and generation are also strongly cumulative at the level of the firm.  It is easier for companies to build on existing capability than to start afresh, and organizational learning is necessary to build up capability.  Learning is particularly crucial in relation to difficult-to-acquire tacit and skill-based knowledge, which may explain why tacit knowledge is often identified as being derived primarily from in-house capability and efforts.

The third reason for the dominance of internal sources of knowledge is the role of specific, as opposed to general, knowledge in innovation.  Pavitt (1984), for example, explains the overriding importance of internally generated knowledge by the specificity of the knowledge inputs necessary for product differentiation in the marketplace and for appropriability.  Giovanni Dosi (1988) relates the concept to the breadth of in-house R&D, arguing that low product specificity in R&D enables companies to achieve synergy across product areas, whereas high firm specificity is more likely to secure appropriability.  These comments on specific knowledge in product innovation resonate with Fleck’s (1988) concern with contingent knowledge in complex process innovation.

Specificity pertains primarily to design and development work, some three quarters of industrial R&D expenditure.  By contrast, industrial research is likely to have a broader remit; it may be characterized as a search activity, undertaken to identify new opportunities and to resolve attendant problems.  Nathan Rosenberg (1992) stressed that companies are simply unable to know fully in advance what they should be searching for, or to pursue all possible alternatives in their search efforts.  Richard Nelson (1982) conceptualizes

440 Index

knowledge as “capability for efficient search” and argues that corporate expenditure on basic research “enhances the productivity of applied research and development” by helping companies to establish where they should be looking.  This is likely to be especially crucial at times of technological discontinuity or paradigm shift.  Nevertheless, companies tend to underinvest in basic research because of the uncertainty surrounding its outcome and the difficulty of appropriating any benefits (Nelson 1959).

Both the knowledge contribution of PSR to innovation and the “division of labor” between industrial R&D and PSR are broadly explicable within this framework (Rosenberg 1990; Pavitt 1991).  In effect, government funding of basic research underwrites the long-term interests of industry by conducting an open-ended and speculative search operation on its behalf.

The contribution of other companies is most easily explained in terms of the importance to innovation of knowledge flow among companies in the supply chain.  Innovations of all types may demand knowledge from suppliers of materials or components incorporated into the final product (von Hippel 1988).  Moreover, success in innovation depends crucially on the quality of knowledge flow about user needs (Rothwell 1977).  The relationship is of course particularly strong with specialist users of complex technologies (Fleck 1988, 1993).  Such considerations should not blind us to the importance of knowledge flow between competitors, however.  Nelson (1982) argued that, in addition to companies’ interests in securing proprietary advantage by keeping certain knowledge private, companies have a collective interest in keeping much technological knowledge in the public domain.  Without this knowledge, no companies on their own would be innovative.  Of course, the patent system and other publications are formal mechanisms by which technological knowledge is shared.  Our field experience suggests that, although industrial staff are careful not to disclose commercially sensitive information, some types of technological knowledge (for example, knowledge of instrumentalities) are quite extensively shared through informal interaction. [22]

The question why some knowledge is generated internally and other knowledge is obtained from external sources is very significant for economists of technology.  A useful conceptual framework for addressing such issues has come from the field of evolutionary economics, in particular from the recent work of Stanley Metcalfe and Michael Gibbons (1989) who write about the organizational knowledge base that companies must “articulate” in order to produce a given set of artifacts.  This framework seeks to explain the dual occurrence of continuity and change in technological innovation.  Path dependence and cumulative knowledge acquisition within the firm explain why companies articulate knowledge most effectively in fields that are


familiar to them, and why they find it difficult to extend their existing knowledge base into new areas of innovative activity.  Metcalfe and Gibbons suggest that external sources of knowledge will be especially important in cases of radical innovation because movement into new fields is strongly constrained, not only by the existing capability of a firm but also by technological paradigms.  According to Dosi (1982, 155), technological paradigms have a “powerful exclusion effect” and so limit the ability of firms to “see” knowledge (including technological options) that is available outside.  Firms’ external search activity and research linkages are important means to overcome these constraints.


Categorizations of Knowledge Used in Innovation

Given that companies use a range of knowledge types in the course of R&D leading to innovation, how should we best conceptualize this epistemological variety?  Five different attempts to categorize knowledge used in innovation are summarized in Tables 3 and 4.  As indicated, these categorizations differ in terms of the disciplinary perspective of their authors and the purposes for which they were devised.  They also fall into two distinct groups in terms of the level of conceptualization attempted.  The contributions in Table 3 concern what we might call broad distinctions in the character of knowledge used in innovation, whereas those in Table 4 represent more specific categories of knowledge.  The latter categorizations provide the basis of the typology proposed here, and the broader distinctions provide a useful context for this typology.


Broad Distinctions in the Character of Knowledge Used in Innovation

Table 3 presents a synthesis of the contributions of Sidney Winter (1987) and Giovanni Dosi (1988), both of whom attempt to understand knowledge used in innovation from within the framework of evolutionary economics.  Although not directly empirically derived, both contributions build on earlier case studies.  The contribution of James Fleck and Margaret Tierney (1991) is based on a detailed study of the development and management of expertise in the course of strategic innovations in financial services that they analyze from a “social shaping of technology” perspective.

Elsewhere, Fleck (1992) proposed a useful framework in which he characterizes technological expertise as involving a tripartite relationship along the three axes of knowledge, power, and tradability, addressed respectively


Table 3: Broad Categorizations of Knowledge Used in Innovation


Fleck & Tierney (1991)

Winter (1987); also Dosi (1988)


Social shaping of technology

Evolutionary economics


To conceptualize knowledge in terms of sociocognitive structures that relate the content of knowledge to how it is distributed among individuals and organizations

To distinguish features of technological knowledge that impinge on the ease of technology transfer between firms




Contingent knowledge

Tacit knowledge

Informal knowledge

Formal knowledge







Elements of a system-independent



by the disciplines of epistemology, politics, and economics.  He concluded that, although some work successfully addresses two of these three axes (one or the other side of the triad), no existing approach integrates all three.  The categorization proposed by Fleck and Tierney (1991, 12) is an attempt to move in this direction.  It explicitly links the social context within which expertise is generated and utilized with its cognitive character:  “Viewing knowledge in terms of components of socio-cognitive structures provides a means for relating the content of knowledge to its specific embodiment” (Fleck and Tierney 1991).

Fleck and Tierney’s sociocognitive structures have two major dimensions: (1) the components of knowledge (identified in Table 3), and (2) the distribution of knowledge among different carrier groups or individuals.  Issues of power and tradability clearly enter the latter dimension, because these issues determine the monetary value and status attributed to particular competence or knowledge.  Labor process factors, for example, shape both the construction of skills and expertise, and internal and external labor markets for specific expertise.  Similarly, the extent to which knowledge can be appropriated impinges crucially on the success of the innovating organization and on the wider diffusion of this knowledge.

As noted earlier, the tradability and ease of transfer of knowledge between companies are central concerns for economists of technology.  Winter (1987) and Dosi (1988) separately suggested continua that characterize technologi-


Table 4: Detailed Categorizations of Knowledge Used in Innovation


Vincenti (1991)

Gibbons and Johnston (1974)  Faulkner, Senker, and Velho (1994)


History of technology

Innovation studies


To develop an epistemology of engineering, in particular to relate categories of knowledge to knowledge-generating activities

To establish the extent and character of knowledge flows from public sector research into industrial innovation by investigating the full range of knowledge inputs to innovation


Categories of Knowledge

Fundamental design concepts
Criteria and specifications
Theoretical tools
Quantitative data
Practical considerations
Design instrumentalities


Knowledge-generating activities

Transfer from science
Theoretical engineering research
Experimental engineering research
Design practice
Direct trials

Content of information                          Broad Knowledge Types

Theories, laws, general principles          Knowledge of particular fields

Properties, composition,                         Technical information

characteristics of materials and              Skills

components                                           Knowledge related to artifacts

Operating principles or rules

Required specifications, technical            Impact on company activities

limitations                                                New product ideas

Design-based information                       Articulation of user needs

Test procedures and techniques           Feedback on existing products

Existence of equipment or materials       Scouting for new applications with particular properties                                Scanning the research frontier

Existence of specialist facilities or          Underpinning knowledge

services                                                 Routine problem solving

Location of information                           New research equipment

New R&D procedures

Skills in experimentation and testing

New process technology

New production methods

Technical backup

444 Index

cal knowledge and together have a strong bearing on cross-sector variety in appropriability regimes and technology transfer.  These have been amalgamated in Table 3.  (The features listed on the right-hand side are likely to be associated with ease of transfer.)

These continua echo a number of elements in the Fleck and Tierney categorization.  The concepts of specific and contingent knowledge refer to the important role of local knowledge discussed earlier, and they may be contrasted with Fleck and Tierney’s category of the more taken-for-granted metalevel knowledge that is universal. [23]  In addition, their distinction between formal and informal knowledge appears to be subsumed in Winter’s distinction between tacit and articulated knowledge.  He elaborates on this aspect by distinguishing between what is nonarticulated and nonarticulable, and between what is teachable and nonteachable.  These distinctions usefully focus on skills that may be taught (by example) although they are not articulable, and on articulable knowledge that sometimes gets “lost” because it is never articulated. [24]  The category “observability in use” describes how easily the underlying knowledge embodied in a product is revealed in practice.  This is likely to depend both on the capability of the observer and on the willingness of the producer to cooperate and share relevant tacit knowledge.  Finally, Winter suggests that the more complex and system dependent the particular technological artifact and knowledge, the less accessible it will be.  This resonates with Fleck’s (1988, 1993) earlier work on configurational technologies in which he classifies technologies in relation to the complexity of the knowledge associated with them.

In principle, it should be possible to combine the elements of the two categorizations in Table 3.  The broad character of knowledge could be related to the wider social and economic factors that influence where knowledge is located and what knowledge gets transferred between individuals and groups.  My purpose here is not to attempt the synthesis Fleck envisions but to elaborate primarily on the more narrowly cognitive or epistemological aspects of his triad.  However, it should be borne in mind that the two other axes of expertise identified by Fleck - power and tradability - cut across the more detailed categorizations of knowledge that we explore below.


Detailed Categories of Knowledge Used in Innovation

Table 4 lists three sets of categorizations constructed on the basis of the studies by Vincenti, Gibbons and Johnston, and ourselves.  Differences in emphasis among these categorizations reflect largely differences in the empirical studies from which they were derived.  Vincenti’s study was based on case studies of “normal” design in one field of engineering over an


extended period ending in 1950; Gibbons and Johnston’s categorization was based on individual successful innovations from a range of sectors in the 1970s; and our own categorization was based on R&D into promising new technologies in the 1980s and 1990s that have not yet yielded significant innovations.  Although they all address technological processes, these studies highlight different aspects and different periods of recent history.  The historical research methods of Vincenti are distinct from Gibbons and Johnston’s and our interview-based methodology.  Nevertheless, all three categorizations have been derived from open-ended and detailed empirical inquiries into the totality of technical knowledge utilized in the course of innovation.  The categories have then been imposed on these data by the researchers. [25]  Both Vincenti and I have striven to use labels and categories identifiable to, if not identified by, practitioners (and so meaningful to them) - except for the term instrumentalities.

Table 5 draws together what I see as the main elements identified in Table 4.  It presents a composite typology that groups fifteen different types of knowledge used in industrial innovation, under five headings.  These categories should be fairly self-explanatory, but their main features are briefly outlined below with reference, where appropriate, to the three studies from which they were derived.

Knowledge relating to the natural world.  This includes theories and knowledge of the properties of materials, two categories that are generally easy to identify.  The domains of science and technology are both present.  Theory, in the sense described by Vincenti, includes the theoretical tools (such a parametric variation) used in engineering experimentation, whereas the category “properties” encompasses properties of artifacts and of nature.

Knowledge related to design practice.  Design-related knowledge, most evident in Vincenti’s categories and least in ours, [26] constitutes a vital aspect of technological innovation.  Typically, design and development activities follow four stages (Walsh et al. 1992, chap. 7).  First, design criteria are drawn up on the basis of the dual requirements of the companies and the potential users.  Second, more detailed specifications are then produced on the basis of technical considerations (feasibility, etc.).  Third, alternative concept designs are considered and one is eventually selected.  Finally, the detailed design of the product is elaborated.

The various types of design concepts listed in Table 5 come from Vincenti.  “Fundamental operating principles” are the principles that make a particular artifact work: for example, the fixed-wing aircraft that flies because of Cayley’s

446 Index

Table 5. Composite Typology of Knowledge Used in Innovation

Related to natural world

Scientific and engineering theory

“laws” of nature; theoretical tools

Properties of materials

Natural and artificial materials

Related to design practice

Design criteria and specifications

Understanding of user requirements

Demands of company and technology

Specifications of components

Design concepts

Fundamental operating principles

Normal configurations

Creative ideas

Design instrumentalities a

Design competence a

General design competence

Competence in specific product area

Practical experience

Related to experimental R&D b

Experimental and test procedures

Research instrumentalitiesa

Ability to utilize experimental techniques and equipment

Ability to interpret test and experimental results

Research competencea

General research competence

Competence in particular specialism

Experimental and test data

Related to final product

New product ideas

Operating performance

Performance of components or materials

Pilot production, field trials, and so on

User experience

Production competencea

Design requirements for manufacture

Competence in pilot production/scale-up

Related to knowledge

Knowledge of knowledge

Location of particular knowledge

Availability of equipment, materials, specialist facilities, or services

a. Indicates knowledge that is heavily skill based.

b. Research and Development

principle that one can “make a surface support a given weight by the application of power to the resistance of air” (cited in Vincenti 1991, chap.


7).  “Normal configurations” are the arrangements and shapes commonly taken to be the best embodiments of operating principles; they represent the framework of “normal” design.  Such knowledge is intrinsically technological rather than scientific and is often taken for granted, having been absorbed from earlier engineering.  The role of creative ideas in both concept design and detailed design is everywhere acknowledged, even if “creative ideas” are a little awkward as a category of knowledge.  Good ideas rarely emerge in a vacuum.

The category design instrumentalities adopted by Vincenti encompasses structured procedures (such as decomposing a problem into subproblems), ways of doing things and thinking (for example, the use of analogy and “what would happen if?” approaches), and judgmental skills (for example, the ability to balance conflicting design requirements) (Vincenti 1991, chap. 7).  The role of skills or competence in all aspects of design - general and specific - is self-evident (although there is arguably considerable overlap between design skills and design instrumentalities).  Practical considerations, in Vincenti’s use, imply knowledge drawing more on experience than skill (1991, chap. 7).  They include the vital elements of user experience of operation, shopfloor experience of construction or production, and the “rules of thumb” from previous design experience. [27]

Knowledge related to experimental R&D.  Experimental and test procedures are accepted ways of setting up experiments and tests. [28]  Research instrumentalities, following de Solla Price (1984), are knowledge and skills related to the techniques and artifacts used in the course of experimental R&D.  As with the example of Rosalind Franklin’s skill at X-ray crystallography, instrumentalities include the ability to use research instruments effectively.  Senker’s work on advanced engineering ceramics demonstrates that the ability to interpret test results obtained from sophisticated equipment is also crucial (Senker and Faulkner 1992).

The nature of general and specific research competence is self-evident, as is the category of experimental and test data.  The latter is perhaps the most tangible knowledge output of R&D.  Vincenti’s work stressed the importance of data to engineering capability and showed that quantitative data may be either theoretically or empirically derived and may be either descriptive or prescriptive (Vincenti 1991, chap. 7).  Our study and Gibbons and Johnston’s reveal that data often relate to both properties of materials and to operating performance (see below).

Knowledge related to the final product.  We have found that new product ideas rarely emerge in a single step from a single source but rather “coalesce”

448 Index

over a period of time.  Knowledge about the operating performance of the product is clearly vital and is obtained variously through pilot production, direct trials, and user experience.  Knowledge about the performance of components and materials is obtained from suppliers and users, and from experience.  Production competence contributes to early concept design (ideally), as well as to later detailed design through pilot production.

Knowledge related to knowledge. The category “knowledge of knowledge” comes from Gibbons and Johnston’s study.  It captures the facility to find out things that are necessary to new product development but that are not known to those immediately involved.  Our study confirmed that, in the course of search activity and problem solving, external contacts are widely used to locate facilities, literature, and other contacts in particular specialisms.


Some Comments on the Composite Typology

Because of the quantitative methodology employed by Gibbons and Johnston, their data (in Table 2) give some indication of the relative importance of different knowledge types.  Thus information relating to design accounts for one quarter of knowledge used in innovation, as does knowledge of knowledge.  Knowledge of the properties of materials and components together account for one-third (unfortunately, we do not know the distribution between materials and components).  And knowledge of test procedures and theories together account for nearly one-fifth.

There is a fair degree of overlap and fluidity between the fifteen categories in the typology.  This is in the nature of the beast: knowledge used in innovation does not come in watertight boxes but is mutable and multidimensional, precisely because of the complex social processes by which it is generated and utilized.  In an attempt to “get a handle” on some of this complexity, I would tentatively suggest that at least two taxonomic axes cut across the specific categories of knowledge listed in Table 5.  The first axis concerns the object of the knowledge in question, which may be the knowledge of:

1. the natural world

2. design practice

3. experimental R&D

4. the final product

5. knowledge itself

These headings provide a convenient and relatively straightforward way of grouping specific knowledge types, as indicated in Table 5.


The second axis cutting across these categories refers to the more slippery but nonetheless significant distinctions concerning the broad character of knowledge.  To these I would add the frequently cited distinction between knowing and doing, and a distinction that, in my view, is still not sufficiently grasped in this “Information Age,” namely, that between knowledge and information (Wildavsky 1983).  This suggests a three-way distinction between knowing as understanding, knowing as holding information, and knowing as holding skills, alongside the main dualistic distinctions identified in Table 3:

1. understanding—information—skill

2. tacit—articulated

3. complex—simple

4. local—universal

5. specific/contingent—general/metalevel

The importance of these broad distinctions is likely to vary with different specific categories of knowledge.  Thus, for example, skill-based knowledge appears in my typology under Design, R&D, and Production.  Specific and local knowledge is likely to be particularly significant in Design, Production, and Operating performance (see Vincenti 1991, chap. 6), as is tacit knowledge, which is also likely to be dominant in Practical experience.

In summary, we may usefully conceive of the knowledge used in innovation in terms of three taxonomic dimensions, namely:

1. specific types of knowledge (the typology)

2. the object or activities with which they are associated (product, R&D, etc.)

3. broad distinctions in the character of knowledge (tacit, specific, etc.)

Although elements in the last two dimensions relate closely to the social and economic issues of power and tradability noted by Fleck, these dimensions do not in themselves account adequately for the more external aspects of knowledge.  They do, however, provide a reasonably rich and rounded framework for conceptualizing the cognitive and epistemological aspects of knowledge used in innovation.

On this basis, I suggest we need a conceptualization that integrates the three dimensions of type, object, and character.  The approach developed by us to investigate knowledge flows between industry and public sector research begins to achieve such an integration empirically.  As indicated in Table 4, we asked three sets of questions about the knowledge inputs to innovation.  First, we asked interviewees to specify the types of knowledge they use, under four headings: knowledge of particular fields, technical information, skills,

450 Index

and knowledge related to artifacts. [29]  We then asked them to characterize these types of knowledge in terms of whether they were predominantly tacit or formal in nature.  Finally, we asked them to indicate the “impact” or contribution of knowledge from various sources in terms of different company activities (also indicated in Table 4).  It is not difficult to see how each of these steps could be refined or extended by application of the composite typology proposed here, together with further questioning on the broad distinctions of character suggested above.



To further our conceptualization of the types of knowledge used in innovation, I have reviewed the literature on the science-technology distinction and on industrial innovation.  This review led to the conclusion that there is a strongly interactive relationship between science and technology, instrumentalities being an important area of overlap.  In some new fields, such as biotechnology, the relationship between science and technology is so intimate that the boundaries between them appear blurred, if not obliterated.  Nonetheless, technology can be distinguished from science because of its practical, artifactual orientation.  This has implications for both its sociotechnical organization and its cognitive or epistemological character.

Empirical findings confirm that what we commonly take to be scientific knowledge is used in the course of R&D leading to innovation, and that industrial organizations and PSR institutions both contribute to this knowledge.  Such conclusions must be placed in context, however, because the contribution of science is relatively small.  Other, more strictly technological, knowledge plays a greater role in innovation.  Moreover, technology also contributes to science - at least to the extent that instrumentalities relate to artifacts - and may thus be considered to encompass both scientific and technological knowledge.  With respect to the institutional distinction between industry and PSR, in-house R&D and expertise generally make a greater contribution to knowledge used in innovation than PSR, which is often less important than knowledge originating from other companies.  The dominance of knowledge from internal sources is explained by the need to appropriate knowledge, by the cumulative nature of innovation, and by the importance of specific knowledge.  Significantly, though, industrial R&D activities demand a synthesis of tacit and formal knowledge from internal and external sources.

This article has compared various attempts to capture the heterogeneity of technological knowledge. The categories proposed by Winter, Dosi, and


Fleck and Tierney identify broad distinctions in the character of technological knowledge (tacit-articulated, specific-general, etc.).  To a degree, these categories relate to the wider economic and power-related factors shaping the distribution of knowledge among individuals and organizations, although they do not account for them.  The categories developed by Vincenti, Gibbons and Johnston, and my colleagues focus in greater detail on the content of knowledge utilized in innovation.  The composite typology proposed here draws together the main features identified in these latter categorizations.  It identifies fifteen different types of knowledge, grouped under five headings that reflect the different activities or objects to which each knowledge type relates: the natural world, design, R&D, the final product, and knowledge itself.  It is argued that the broad distinctions in the character of knowledge identified by Winter, Dosi, and Fleck and Tierney also cut across the more specific categories in the proposed typology.  Thus a more complete conceptualization of technological knowledge should incorporate three taxonomic dimensions: the specific knowledge types (viz., our typology), the object or activities to which they relate, and the broad differences of character.

This conceptualization is a refinement of the categories used in our empirical study.  I believe that this refinement strengthens and extends the applicability of a research approach that seeks to characterize in detail the range of knowledge used in the course of R&D leading to innovation.  The studies by Gibbons and Johnston and ourselves have utilized this approach to good effect as a means of examining the particular knowledge contribution of PSR to innovation.  But there is a real need, and considerable scope, to improve and refine our conceptualization of the knowledge flows associated with all aspects of industrial innovation, not solely public-private research linkage.  For example, collaborations between user and supplier companies may also be addressed in terms of science and technology flows.  Indeed, work being conducted at Edinburgh specifically focuses on different types of expertise utilized in the development and modification of complex IT systems (Fleck 1992b).  There is also a connection with recent work at Manchester that investigates technology strategy in terms of the relationship between companies’ knowledge base and corporate strategies (Coombs and Richards 1991).

These developments appear to signal a more “holistic” approach to the study of industrial innovation that, in the spirit of this journal, genuinely spans sociological and economic approaches to the subject.  Moreover, the kind of work I have in mind offers a possibility to explore how knowledge itself changes during the innovation process.  For example, it is already widely recognized that technology transfer involves a transformation of knowledge (Gold 1980; Peláez 1991).  Perhaps our most exciting challenge is to “get

452 Index

inside” the processes of knowledge transformation, using research tools like the typology proposed here.

Substantial policy and management benefits could result from such work.  or example, the typology could be used by companies to investigate the extent and nature of their knowledge base and knowledge requirements in specific areas - perhaps in areas that are new to them - or to assess whether they are making the best use of available sources of knowledge to meet their requirements.  Similarly, government organizations might use this approach to assess the strengths and weaknesses of the R&D system in a particular field - perhaps one of strategic interest - or to assess the effectiveness of policy measures geared to enhancing the R&D system and flows of knowledge around it.  Discussions of these important issues are often lamentably superficial and policy interventions are not sufficiently targeted.

Perhaps the greatest need for more sophisticated tools for understanding the knowledge required in innovation lies in those countries and firms least advantaged in terms of scientific and technological infrastructure.  In such cases, technology strategy demands a means to address questions such as:

What types of knowledge about available technologies do we need and how do we gain access to them?  What types of knowledge do we need if we are to acquire an external technology, or to develop our own internally?  In principle at least, the conceptualization I have proposed here could act as a checklist and enable organizations to make informed decisions that avoid wasting limited scientific and technological resources.



1. An earlier version of this article was presented to a workshop entitled “Exploring Expertise,” which sought to draw together different conceptualizations.  See especially Fleck 1992 and Winter 1987.

2. I am using the conventional shorthand of research and development (R&D) to include design, but I would stress that design and development are often more important to innovation than R&D.

3. The study was titled “Public-Private Research Linkage in Advanced Technologies.”  It was funded by the U.K. Economic and Social Research Council under the Science Policy Support Group Initiative on Public Science and Commercial Enterprise.  Dr. Jacqueline Senker of the Science Policy Research Unit, University of Sussex, and Dr. Lea Velho, now of the University of Campinas, conducted most of the fieldwork for this study.  I am grateful to them both for permission to use some of the findings here.

4. Utterback(1971) and Rosenberg(1992) make a similar point about innovation in scientific instruments.

5. They also, contrary to de Solla Price’s assumptions, contribute to this literature (Hicks, Isard, and Martin 1993).


6. Fores may be justified in stressing the primarily empirical character of engineering, but it is hard to refute the case that the transition from craft to modern technology represents two “epochs” (Constant 1 984b) in the history of technology, even though elements of the former inevitably remain an important part of what we today call technology.  See Layton (1988) and Channell (1988) for their responses to Fores.

7. The greater number and importance of “relevant social groups” was, of course, also recognized in Pinch and Bijker’s seminal call for a sociology of technology built on conceptual frameworks from the sociology of science.  As David Edge (1992) notes, the greater social complexity of technology makes it more difficult to study than science.  More forcefully, Sørensen and Levold (1992) argue that scholars who transfer conceptual frameworks from the study of science to the study of technology are likely to have “blind spots” concerning technology.  Indeed, sociologists of technology have been strongly criticized for failing to grasp macrolevel forces shaping technology (e.g., Russell and Williams 1988).

8. This view is explored empirically in Edge (1992 seep. 158).

9. Collins (1974) himself does problematize what constitutes replication in science.

10. These techniques involve some theory (e.g., laws of similitude and dimensional analysis), but such theory is more engineering than scientific in character, in the sense outlined above.

11. Although, as Vincenti (1991, chap. 4) noted, it may involve an act of considerable reduction to reveal this in the case of engineering theory.

12. These studies attempted to identify the key cognitive or research “events” that contributed to specific innovations (in industry and defense, respectively) and then to analyze what proportion of these events took place in publicly funded laboratories.  The studies were criticized because the time frame adopted had a crucial bearing on the results produced, because the method of retrospective reporting was highly selective and assumed that the origin of an idea or “piece” of knowledge could be sensibly identified, and because of the inadequacies of the linear model on which the approach was based (Barnes and Edge 1982).

13. Their data have been reworked in two ways.  First, their content categories have been grouped under five headings, to reflect more closely the composite categorization developed below.  Second, their external sources have been broken down as far as the data allow to reveal the respective contributions of public sector research (PSR) and other companies.  A small proportion of their data could not be easily categorized in this way, and so only 254 of the total 300 units of information for external sources are represented in this table, and only 94 of the total 107 obtained from “public science.”

14. Originally, thirteen subheadings were used under these six main headings, but there was little variation between sources for the subheadings under Future Innovations, Production, and Technical Backup, so these have been aggregated.  The numbers of responses sometimes include halves because many of our respondents were unable to identify a single source as having the major impact on an activity and so gave dual responses; these numbers have been allocated equally between each source involved.

15. A full account of the methodology developed can be found in Faulkner (1992).

16. It should be stressed that the unit of analysis in Table 2 is numbers of responses and not units of information as in Table 1.  Also, the responses are to categories that are in no way equally weighted.  As a result, the absolute values are not significant although the relative ones are - at least within each “impact” category.

17. Our study revealed some variation in the relative importance of collectively derived and individually held knowledge, but we did not examine this systematically.

18. PSR accounts for 53 percent of the reported impact on scanning the research frontier and 43 percent of that on underpinning knowledge.

19. This explains the importance of “on the hoof’ or “person embodied” mechanisms of transfer.

454 Index

20. See, for a review, Rosenberg (1982, chap. 1); also his early work on the nineteenth-century machine tool industry (1976).

21. These terms have perhaps been adopted too easily: Fleck, Webster, and Williams (1990) demonstrates that trajectories of technological development do not always follow the path anticipated for them and that alternative trajectories can exist side by side.

22. We have found that industrial R&D workers and designers communicate quite frequently with their opposite numbers in competitor companies (Senker and Faulkner 1993).  Barden and Good (1989) found such discussions, although not frequent, to be highly influential in terms of the direction of projects.

23. They give as an example assumptions of technocratic rationality.

24. A common example is the failure to document fully programming code.

25. Although in the case of our categories, this took place at the pilot stage of the research.

26. This largely reflects our initial decision to subsume design into R&D, and the fact that our categories derived primarily from a study of biotechnology, where design is relatively unimportant compared with research.  The gap became particularly evident in our study of parallel computing and represents an obvious area of improvement for future work.

27. Vincenti (1991, chap. 7) gives as an example the knowledge that successful jets require a ratio of engine thrust to loaded aircraft weight of between 0.2 and 0.3.

28. Note that this category might be subsumed under research instrumentalities.  Although this would parallel Vincenti’s use of instrumentalities in relation to design, it would create a rather large category and so potentially lead to a loss of “resolution.”

29. Knowledge of particular field includes scientific theory, engineering principles, and knowledge of knowledge (after Gibbons and Johnston 1974); technical infonnation includes specifications and operating performance of products or components, plus experimental or test procedures and results; skills includes specific skills, such as programming, and more general research or production competence; artifacts includes knowledge relevant to process or research instrumentation (overwhelmingly the latter), and other intermediates used in R&D.



Barden, P., and B. Good. 1989. Information flows into industrial research.  London: Centre for the Exploitation of Science and Technology.

Barnes, B., and D. Edge. 1982. The interaction of science and technology.  In Science in context: Readings in the sociology of science, edited by B. Barnes and D. Edge, 147-54. Milton Keynes: Open University Press.

Battelle Columbus Laboratory. 1973. Interactions of science and technology in the innovative process: Some case studies. Final report to the National Science Foundation (NSF-C667). Ohio: Battelle Columbus Laboratory.

Bode, H. W. 1965. Reflections on the relation between science and technology. In National Academy of Science, Basic research and national goals; A report to the Committee on Science and Astronautics, U.S. House of Representatives. Washington, DC: Government Printing Office.

XXX Bud, R., and S. Cozzens, eds. 1992. Invisible connections: Instruments, institutions and science. Bellingham, WA: SPIE Optical Engineering Press.

Carpenter, M. P., M. Cooper, and F. Narmn. 1980. Linkage between basic research and patents. Research Management 23 (2):30.

Carpenter, M. P., F. Narmn, and P. Woolfe. 1981. Citation rates to technologically important patents. World Patent Information 3 (4):160-63.


Channell, D. F. 1982. The harmony of theory and practice: The engineering science of WJM Rankine. Technology and Culture 23 (l):39-52.

___________ . 1988. Engineering science as theory and practice. Technology and Culture 29 (1): 98-103.

Collins, H. 1974. The TEA set: Tacit knowledge and scientific networks. Science Studies 4:165-86.

Constant, E. 1984a. Scientific theory and technological testability: Science, dynamometers, and water turbines in the 19th century. Technology and Culture 24(2):l83-98.

______________ . l984b. Communities in hierarchies: Structure in the practice of science and technology. In The nature of technological knowledge: Are models of scientific change relevant? edited by R. Laudan. Dordrecht: Reidel.

Coombs, R., and Richards, A. 1991. Technologies, products and firm strategies: Part II- Analysis of three cases. Technology Analysis and Strategic Management, 3 (2):157-75.

de Solla Price, D. J. 1965. Is technology historically independent of science? A study of statistical historiography. Technology and Culture 6 (4): 553-68.

Dosi, G. 1982. Technological paradigms and technological trajectories: A suggested interpretation of the determinants and direction of technical change. Research Policy 11:147-62.

_____ .1988. The nature of the innovative process. In Technical change and economic theory, edited by G. Dosi, C., Freeman, R., Nelson, L., Soete, and G. Silvetherg, (eds.) London: Pinter.

Edge, D. 1992. Mosaic arrays in infrared astronomy. In Invisible connections: Instruments, institutions and science, edited by R. Bud and S. Cozzens. Bellingham, WA: SPIE Optical Engineering Press.

Faulkner, W. 1986. Linkage between industrial and academic research: The case of biotechnological research in the pharmaceutical industry. D.Phil. Thesis, Science Policy Research Unit, University of Sussex, Brighton.

_________ . 1989. Industry-academic linkages in the case of biotechnology in the UK: Practice and policy. Paper presented at ECPR workshop on Technology Transfer and Techno-Industrial Innovation, 12-14 June, Como, Italy.

_______ . 1992. Understanding industry-academic research linkage: Towards an appropriate conceptualisation and methodology. Edinburgh PICI’ Working Paper no. 35. Edinburgh: Research Centre for Social Sciences, University of Edinburgh.

Faulkner, W., J. Senker, and L. Veiho. (1994). Knowledge frontiers: Industrial innovation and public sector research in biotechnology, engineering ceramics and parallel computing. Oxford: Clarendon Press.

Fleck, J. 1988. The development of information-integration: Beyond CIM? Edinburgh PICT Working Paper no. 9. Edinburgh: Research Centre for Social Sciences, University of Edinburgh.

______ . 1992. Expertise: Knowledge, tradability and power. Paper presented at P1CT Workshop on Exploring Expertise, November, Edinburgh.

______ . 1993. Configuration: Crystallising contingency. International Journal of Human Factors in Manufacturing, 3 (1): 15-36.

Fleck, J., and M. Tierney. 1991. The management of expertise: Knowledge, power and the economics of expert labour. Edinburgh PlCT Working Paper no. 29, Edinburgh: Research Centre for Social Science, University of Edinburgh.

Fleck, J., Webster, J., and Williams, R. 1990. The dynamics of IT implementation: A reassessment of paradigms and trajectories of development. Futures, 22 (6):618-40.

Fores, M. 1988. Transformations and the myth of “engineering science”: Magic in a white coat. Technology and Culture 29 (1): 62-81.

Freeman, C. 1982. The economics of industrial innovation. 2nd ed. London: Pinter.

456 Index

Gambardella, A. 1992. Competitive advantages from in-house scientific research: The U.S. pharmaceutical industry in the 1980s. Research Policy 21:391-407.

Gibbons, M., and C. Johnson. 1982. Science, technology and the development of the transistor. In Science in context: Readings in the sociology of science, edited by B. Barnes and D. Edge, 177-85. Milton Keynes: Open University Press.

Gibbons, M., and R. Johnston. 1974. The roles of science in technological innovation. Research Policy 3 (3):220-42.

Gold, B. 1980. On the adoption of technological innovations in industry: Superficial models and complex decision processes. Omega 8 (5):505-16.

Granberg, A., and R. Stankiewicz. 1978. The production of knowledge in technological fields. Research Policy Studies Discussion Paper no. 122. Lund: Research Policy Institute.

Hicks, D., P. Isard, and B. Martin. 1993. An analytical comparison of research in European and Japanese corporations. Proceedings of the Kyoto Meeting of the R&D Dynamics Network, May, Kyoto, Japan.

Illinois Institute of Technology Research Institute. 1968. Technology in retrospect and critical events in science; Final report to the National Science Foundation (NSF-C535). Illinois: Illinois Institute of Technology Research Institute.

Latour, B. 1987. Science in action: How to follow scientists and engineers through society. Milton Keynes: Open University Press.

Layton, E. T. 1974. Technology as knowledge. Technology and Culture 15 (1):31-41.

_________ . 1976. American ideologies of science and engineering. Technology and Culture 17 (October):688-700.

________ . 1988. Science as a form of action: The role of the engineering sciences. Technology and Culture 29 (1):82-97.

Lieberman, M. B. 1978. A literature citation study of science-technology coupling in electronics. Proceedings of the IEEE 6(1):5-13.

Mayr, 0. 1982. The science-technology relationship. In Science in context: Readings in the sociology of science, edited by B. Barnes and D. Edge, 155-63. Milton Keynes: Open University Press.

Metcalfe, J. S., and M. Gibbons. 1989. Technology, variety and organisation: A systematic perspective on the competitive process. In Research on technological innovation, management and policy, edited by R. S. Rosenbloom and R. Bergelman. JAI.

Musson, A. E., and E. Robinson. 1969. Science and technology in the industrial revolution. Manchester: Manchester University Press.

Narmn, F, and E. Noma. 1985. Is technology becoming science? Scientometrics 7 (3-6): 369-8 1.

Nelson, R. 1959. The simple economics of basic scientific research. Journal of Political Economy (June):297-306.

_______ . 1982. The role of knowledge in R and D efficiency. Quarterly Journal of Economics 388:453-70.

Pavitt, K. 1984. Sectoral patterns of technical change: Towards a taxonomy and a theory. Research Policy 13(6):343-74.

______ . 1991. What makes basic research economically useful? Research Policy 20:109-19.

Peláez, E. 1991. From symbolic to numerical computing: The story of thinking machines. Edinburgh PICT Working Paper no. 23. Edinburgh: University of Edinburgh, Research Centre for Social Sciences.

Polanyi, M. 1966. The tacit dimension. London: Routledge & Kegan Paul.


Rosenberg, N. 1976. Perspectives on technology. Cambridge: Cambridge University Press.

__________ . 1984. The science/technology relationship, the craft of experimental science, and policy for the improvement of high technology innovation. Research Policy 13 (l):3-20.

__________ . 1982. Inside the black box. Technology and economics. Cambridge: Cambridge Uni­ersity Press.

__________ . 1990. Why companies do basic research (with their own money)? Research Policy 19: 165-74.

__________ . 1992. Scientific instrumentation and university research. Research Policy 21:381-90.

Rothwell, R. 1977. The characteristics of successful innovators and technically progressive firms. R&D Management 7 (3):191-206.

Rothwell, R., C. Freeman, A. Horsley, V.T.P. Jervis, A. B. Robertson, and J. Townsend. 1974. SAPPHO updated—project SAPPHO, phase II. Research Policy 3 (3):258.

Russell, S., and R. Williams. 1988. Opening the black box and closing it behind you: On micro-sociology in the social analysis of technology. Edinburgh PICT Working Paper no. 3. Edinburgh, United Kingdom: University of Edinburgh, Research Centre for Social Sciences.

Senker, J. 1993. The contribution of tacit knowledge to innovation. AI and Society, (7): 208-24.

Senker, J., and W. Faulkner. 1992. Industrial use of public sector research in advanced technologies: A comparison of biotechnology and ceramics. R&D Management 22 (2):157-75.

_____________________ . 1993. Networks, tacit knowledge and innovation. Paper presented at ASEAT Conference on Technology Collaboration: Networks, Institutions and States. Manchester: University of Manchester Institute for Science and Technology (UMIST) and Manchester University.

Sherwin, C. W., and R. S. Isenson. 1967. Project Hindsight. Science 156:571-77.

Staudenmaier, J. M. 1985. Technology’s storytellers: Reweaving the human fabric. Cambridge, MA: MIT Press.

Sørensen, K., and N. Levold. 1992. Tacit networks, heterogeneous engineers, and embodied technology. Science, Technology, and Human Values 17 (1):13-35.

Utterback, J. M. 1971. The process of innovation: A study of the origination and development of ideas from new scientific instruments. IEEE, Transactions on Engineering Management 18: 124.

Vincenti, W. 1991. What engineers know and how they know it: Analytical studies from aeronautical history. Baltimore: Johns Hopkins University Press.

von Hippel, E. 1988. The sources of innovation. Oxford: Oxford University Press.

Walsh, V. 1984. Invention and innovation in the chemical industry: Demand-pull on discovery-push. Research Policy 13 (4):211-34.

Walsh, V., R. Roy, M. Bruce, and S. Potter. 1992. Winning by design: Technology, product design and international competitiveness. Oxford: Blackwell.

Wildavsky, A. 1983. Information as an organisational problem. Journal of Management Studies 20 (1):29-40.

Winter, S. 1987. Knowledge and competence as strategic assets. In The competitive challenge: Strategies for industrial innovation and renewal, edited by D. Teece, 159-83. Cambridge, MA: Balhinger.

Wendy Faulkner trained in biology and, at the Science Policy Research Unit, Sussex, in science and technology policy. She is now based in the Science Studies Unit at the University of Edinburgh (Edinburgh EH8 9LN), where she convenes a postgraduate program in Technology Studies. Her research has explored various aspects of industrial innovation in both large and small firms. She has also written on gender, science, and technology.



The Competitiveness of Nations

in a Global Knowledge-Based Economy

April 2003

AAP Homepage