‘I’m studying ocean tech’, I would say whenever anyone asked about my PhD. That answer was conveniently short and sweet. The Chronicle Herald provincial newspaper kept everyone abreast of this important ‘new’ industry. So, on the surface, my topic was easily understood by family and friends – even those who had no idea how a PhD works. Everyone in my network seemed to understand what I meant by ‘ocean technology’. But, secretly, I didn’t. Something about that contrast of understandings nagged at me for months. Sure, I was caught up in the common academic problem of needing to define everything. But, as we started to see in Chapter 3, defining an industry is also a practical problem (Kirsch et al, 2014). And I couldn’t get my head around the sociopolitics of this industrial category.

Over that first year or two, I met with several key players and policy makers to get the lay of the land (or, in this case, to ‘find my sea legs’). I attended as many ‘industry events’ as I could. I collected names and business cards that might be useful for future fieldwork. But mostly I listened and learned. Sometimes these public events were in big conference venues (for example, see the preambles to Chapters 3 and 6). Sometimes they were the small ‘ocean connector’ events hosted by the Institute for Ocean Research Enterprise (IORE), a newly rebranded industry-facing unit at Dalhousie University. I followed many of the same speakers across these events, and no one was more central than IORE’s Executive Director, Jim Hanlon. An electrical engineer by training, Hanlon had decades of senior leadership experience in several technology companies, including multinationals and two of his own start-ups. Much of his career has been oceans related. And anyone who followed these industry events would recognize him as a positive driving force for ocean technology development in the province. They would also regularly hear the ironically dismissive opening lines of his stump speech. He would joke that if we are going to talk about an ocean technology sector in this province, then there must only be two other technology sectors: land and aerospace.

Each time I heard Hanlon deliver this line, I would laugh along with the audience at the notion of ‘land technology’. His coarse, comedic categories were clearly nothing like the animal-vegetable-mineral categories in Linnean taxonomy (Linnaeus, 1758). But regular folk do talk of aerospace technology (as do statisticians). And around these events, everyone was talking about ocean technology. On the surface, these two categories seemed viable. But they would break down whenever anyone (Hanlon included) mentioned the importance of ocean technology to space exploration. Indeed, the National Aeronautics and Space Administration (NASA) and other space agencies are heavily invested in deep-sea exploration; NASA co-chaired the international Oceans ’22 engineering and technology conference. So, ocean technology is not even close to a discrete sectoral category.

Here in Canada, ‘ocean tech’ is the most common label for what some other parts of the world call blue, offshore, or marine tech. None of these labels is based in any system of standardized industrial classification, although I once heard rumours of lobbying efforts to gain that kind of legitimacy by ‘updating’ the North American Industrial Classification System (NAICS). Merely scratching the surface of these labels will reveal that they are ‘folk taxonomy’ – groupings only loosely connected with real, naturally material distinctions. This is akin to the way in which many people speak of spiders as ‘bugs’. Biologists make clear distinctions between arachnids and insects, as they do between ‘true fish’ and the strikingly different species that we call starfish, shellfish, and jellyfish.

Indeed, I discovered quite a few slippery jellyfish when I tried to reproduce the list of Nova Scotia’s ‘over 200 companies’ in ocean tech (Government of Nova Scotia, 2012, p 1). For a paper at our regional academic conference (MacNeil, 2014), I reviewed the public membership list of the Ocean Technology Council of Nova Scotia (Ocean Technology Council of Nova Scotia, 2013) (66 private sector members), the Canadian Ocean Technology Sector Map (Almada Ventures Inc., 2013) (31 private companies), a government-commissioned ‘Ocean Technology’ value chain analysis (Gereffi et al, 2013) (35 companies), and an internal provincial government working list (72 companies). Removing the duplicates gave me a spreadsheet of only 120 ‘ocean tech’ companies in the province – and it included law firms, accounting firms, machine shops, and several others only vaguely linked to technology and the ocean. For 17 of these companies, I could find no trace on the internet of any product or service linked with the ocean. Nonetheless, I proceeded to assign six-digit NAICS codes to each of the firms on my list using the product descriptions on their websites and/or the NAICS codes provided in Industry Canada’s (2013) directory of Canadian Companies Capabilities. I found that the firms were distributed across 45 different NAICS categories ranging from the 210000 level (Resource Extraction) through to the 610000 level (Education Services). This confirmed what was already clear: ‘ocean technology’ is pretty a ‘folksy’ departure from the accepted norms of industrial classification. But it also pointed towards some bigger questions about technology, innovation, and sectoral classification.

As I pulled the thread, some basic assumptions of innovation studies started to unravel. You see, product-based industrial classifications (NAICS, SIC, etc.) are embedded within innovation theory through the methods that were used by Keith Pavitt (1984) and his heirs (Archibugi, 2001; Castellacci, 2008). The resulting ‘taxonomies’ of innovation have been widely used as ‘a predictive tool’ (de Jong and Marsili, 2006, p 215) for firm and sectoral innovation performance, and this has been widely applied in public policy – most notably at the OECD (de Jong and Marsili, 2006). When I tried to map my list of ocean technology companies into these taxonomic systems, I ran into trouble (MacNeil, 2014). Four firms did not match any of the six patterns/categories identified in Castellacci’s (2008) widely cited update to Pavitt (1984). For example, the website for Jasco Applied Sciences indicated that this one small company was doing two very different things: providing acoustic impact assessment services (for example, ‘will this underwater activity interfere with marine mammal communication?’) while also developing and manufacturing underwater acoustic sensors. The first of these matches Castellacci’s ‘knowledge intensive business services’ innovation mode (and falls within NAICS #541712, ‘Research and Development in the Physical, Engineering, and Life Sciences’), while the second matches his ‘science-based manufacturing’ mode (and falls within NAICS #334511, ‘Search, Detection, Navigation, Guidance, Aeronautical, and Nautical System and Instrument Manufacturing’). The two modes are meant to be starkly different technological regimes, yielding starkly different technological trajectories. However, this one scientific instrumentality firm was defying the classification.

Like some Victorian-era zoologist, I became excited that I had ‘discovered’ the platypus – the one species that might rewrite innovation taxonomies. Biologists would label a specimen like Jasco Applied Sciences as incertae sedis (uncertain placement)1 – a designation that calls for further taxonomic investigation. But then I remembered: Jasco is an organization, not an organism. It cannot be classified like an animal, vegetable, or mineral. Indeed, the entire grouping of ‘ocean tech’ seemed to suggest that sectoral boundaries are not as ‘natural’ as innovation research assumes. That is the thrust of this chapter.

As we will see, taxonomic classification is an important – and often problematic – instrumentality in the innovation studies toolkit. To understand the problems of taxonomic classification for innovation research, I build upon Gareth Morgan’s (1980, 1986) work on metaphors in organization theory. Morgan (1980, 1986) argued that academic schools of thought are powerfully shaped by their acceptance and use of certain metaphors. For example, innovation studies have biological metaphors at their core (for example, Nelson and Winter, 1982) and the idea of taxonomic classification is one of those that we have explicitly borrowed from biology (see Archibugi, 2001; de Jong and Marsili, 2006). This conceptual borrowing has been theoretically fruitful, but we have long known that that there are limits to the use of biology metaphors when studying economics and organization (for example, Penrose, 1959; Nooteboom, 2000; Ziman, 2003c; Langrish, 2017). Jeroen de Jong and Orietta Marsili (2006) argue that it was important to borrow the idea of taxonomy from evolutionary biology so that we might identify the many variables that lead to differences in innovation behaviour among firms. But even Geoffrey Hodgson – the most ardent proponent of Darwinian biology metaphors in economics – notes that there are explanatory limits (Hodgson, 2002).

Taxonomic classification is what Morgan (1980, 1986) would call a ‘puzzle solving activity’. Such research activities are enabled by their underlying metaphorical assumptions; the metaphors are instrumental to the research activity. In this chapter, I will argue that an implicit organism metaphor has driven the many decades of taxonomic puzzle solving in innovation studies. Unfortunately, the assumptions carried by any single social science metaphor ‘are rarely made explicit and are often not appreciated, with the consequence that theorizing develops upon unquestioned grounds’ (Morgan, 1980, p 619). To this end, metaphors are not merely ‘literary frill’ (Hodgson, 2002, p 263) or rhetorical flourish: ‘the logic of metaphor has important implications for the process of theory construction’ (Morgan, 1980, p 611). ‘Metaphors are never innocent’ (Derrida, 1978, p 17); they are instrumental.

I begin this chapter by reviewing Pavitt’s formative taxonomy and its known limitations. I then use analogies from taxonomic biology to further our understanding of the three major methodological problems already described in the innovation taxonomies literature. This includes, but is not limited to, the taxonomic separation of public and private organizations. We will see how public organizations become excluded from classification. I then turn to the deeper problems that arise from the implicit use of an organizations-as-organisms analogy. This requires some general discussion of the organism metaphor and its implications for theorizing and classifying organizations. Then, I explore the theoretical assumptions of inheritance, determinism, and functional unity that are embedded in the idea of taxonomic classification. I conclude by returning to the metatheoretical implications of biological metaphors and the corresponding political assumptions that are inscribed in sectoral classification tools. I argue that we will need other metaphors if we hope to observe other dark innovation patterns.

The taxonomic puzzle

Pavitt’s taxonomy

Among Keith Pavitt’s countless contributions to innovation studies (Verspagen and Werker, 2004), his taxonomy (Pavitt, 1984) is said to have been the most significant (Archibugi, 2001). It sparked decades of investigations into innovation behaviour patterns and remains extremely influential to this day (Archibugi, 2001; Peneder, 2003, 2010; Castellacci, 2008, 2009). Martin (2010) goes so far as to suggest that Pavitt’s taxonomy ended the race for one unifying innovation model:

By the 1980s we had dozens, if not hundreds, of innovation studies, all coming up with their own models of the innovation process. Keith Pavitt (1984) showed that if you classify firms into a number of sectors, then you could begin to make sense of the rather baffling picture that was beginning to emerge. (Martin, 2010, p 4)

In the original paper, Pavitt used data on 2,000 ‘significant’ manufacturing industry innovations in the UK (1945–79) to develop five broad patterns or categories of innovation behaviour. He called these his ‘sectoral technological trajectories’ (Pavitt, 1984, p 354). First, his ‘supplier-dominated firms’ were generally small, used nontechnical means to appropriate process innovations, and relied primarily on technological inputs (knowledge flows) from suppliers. His ‘science-based firms’ were much larger; used know-how, patents, and secrecy to appropriate a mix of product and process innovations; and drew knowledge from both internal R&D and public science. Pavitt divided his third category, ‘production-intensive firms’, into two subcategories: (a) ‘scale-intensive firms’ and (b) ‘specialized suppliers’. The former were large firms that drew knowledge from both suppliers and internal R&D, and used a variety of means to appropriate process innovations. The latter category, ‘specialized suppliers’, were small firms that relied on their customers/users and their internal R&D activities to develop product innovations, which were appropriated through know-how and patents.

While Pavitt (1984) spoke of sectoral trajectories, some might make the distinction that he also used variables associated with ‘technological regimes’ (Breschi and Malerba, 1997). A technological regime is ‘the particular combination of technological opportunities, appropriability of innovations, cumulativeness of technical advances and properties of the knowledge base’ within which an industry operates (Breschi et al, 2000, p 388). Indeed, Castellacci (2008) describes Pavitt’s taxonomy as more than simply a model of technological trajectories. The taxonomy’s categories are said to help us anticipate the trajectory of a firm’s technological change because the direction of that change is shaped by the firm’s technological ‘paradigm’ (Dosi, 1982) or ‘regime’ (Breschi et al, 2000). Pavitt’s work was an early articulation of these trajectory and regime/paradigm concepts (Castellacci, 2008).

But Pavitt’s taxonomy also goes further, helping to explain the vertical linkages that tie together various manufacturing sectors. Archibugi argues that the taxonomy made it ‘easier to explore how different economic units are interconnected and to identify the main knowledge flows and user-producer linkages’ (2001, p 422). Castellacci has argued that these linkages are ‘a crucial aspect’ (2008, p 980) and the ‘most original contribution of Pavitt’s taxonomy’ (2009, p 324). This taxonomy was therefore an early attempt to create an integrated paradigm-regime-trajectory-linkage model of organizational innovation (Castellacci, 2008).

While Pavitt’s taxonomy has been used extensively by innovation scholars and policy makers, it has also been heavily critiqued (Cesaratto and Mangano, 1993; De Marchi, Napolitano, and Taccini, 1996; Archibugi, 2001; Gallouj, 2002; Hollenstein, 2003; de Jong and Marsili, 2006; Leiponen and Drejer, 2007; Castellacci, 2008). Many of the critics forget that Pavitt readily acknowledged the limitations of his work: ‘Given the variety in patterns of technical change that we have observed, most generalizations are likely to be wrong, if they are based on very practical experience, however deep, or on a simple analytical model, however elegant’ (Pavitt, 1984, p 370). Here, he recognized that his taxonomy was only a starting point. He called for further exploratory research, extensions, and alterations.

Gallouj (2002) provides an extensive survey of the problems with Pavitt’s taxonomy. He notes that Pavitt’s work excludes nonmarket firms, lumps all service sector firms into the ‘supplier dominated’ category, assumes that product-based sectors are heterogeneous, overlooks the possibility of innovation co-production, assumes a clear distinction between product and process innovations, neglects organizational innovations, and uses firm size as a determinant of technological trajectory (whereas the causality should likely be reversed) (Gallouj, 2002). It is generally accepted that these problems arise from methodological limitations within the taxonomic literature. Two broad remedies have been discussed: extensions and replications in other research contexts, and taxonomic research that is not linked to product-based industrial classification. Maintaining the biological metaphors for now, I call these the problems of ‘mare incognitum’ and ‘parataxonomy’. I will explain each in turn. But first, let’s consider a third unresolved problem: the ‘wastebasket taxon’ that Pavitt (1984) created for government organizations.

Government as wastebasket taxon

As he was concluding his original article, Pavitt noted some limitations of his work and began to suggest future alterations. In particular, he suggested that another category might be needed ‘to cover purchases by government and utilities of expensive capital goods related to defence, energy, communications and transport’ (Pavitt, 1984, p 370). This throwaway line tells us that Pavitt knew his exclusion of public organizations was problematic. Surprisingly, however, the taxonomic literature does not include any subsequent alterations or alternative taxonomies that integrate ‘nonmarket’ (public and social sector) organizations. There have been multiple attempts to develop entirely separate taxonomies of public sector innovation (Hartley, 2005; Koch and Hauknes, 2005; de Vries et al, 2016). And, of course, we know that public organizations do innovate (see de Vries et al, 2016). But generally, the public sector has been kept apart – and mostly absent – from the literature on taxonomies of innovation.

Some might say that this makes sense. They might argue that public organizations are different from private companies at the most basic taxonomic level. If so, then it would be fair to classify the whole ‘public sector’ (and perhaps the ‘social sector’) as a separate ‘domain’ or ‘kingdom’ in our innovation taxonomies. In other words, organizations that serve any public or social purpose might be analogous to bacteria and/or archaea, while for-profit companies are the eukaryotes: the animals, plants, fungi, and other familiar organisms with closed nuclei. Notice where this analogy takes us. Biology has long been dominated by the study of eukaryotic organisms. Although bacteria were first identified beginning in the 17th century – thanks to advancements in microscopic lenses, they only rose to prominence in the 19th century with the help of Louis Pasteur (among others) (Latour, 1993). Meanwhile, microbiologists only began to understand archaea as distinct from bacteria in the late 20th century. Indeed, the now common three-domain biological taxonomy – bacteria, archaea, eukaryota – was only proposed in 1990 (Woese et al, 1990). As Stefan Helmreich (2009) explains, ‘alien’ microbes had been found thriving in the extreme heat of deep-sea hydrothermal vents. Those microbes forced microbiologists to reconsider basic taxonomic structure and assumptions. Some say that these microbes are our ancient ancestors (hence ‘archaea’). And given their love for extreme conditions, some say they might also be found on other planets. But what is most curious about archaea is that some of them have a genetic structure resembling both bacteria and eukaryota. These microbes tell us that the fundamental genetic boundaries of biological taxonomy are not as sharp as everyone assumed. This has left the tree of life ‘in a brambled state’ (Helmreich, 2009, p 81). I will return to the odd genetic structure of these microbes later. But for now, it should be noted that domains and kingdoms have recently been opened for major revision in biology. The biological analogy suggests that we should always be open to revising basic taxonomic distinctions.

I, for one, seriously doubt any claims to a universal distinction between market and nonmarket organizations. Yes, there are legal distinctions between publicly governed and privately owned organizations (and let’s not forget the different legal structures of membership-based societies and cooperatives). But there are also so many hybrids: crown corporations, social enterprises, community interest corporations, and so on. There are all those public organizations that act ‘business like’ and all those private companies that purport to have a public purpose (see Rhodes, 2021). I agree with Henry Mintzberg, who once said ‘it is time we recognized how limited that dichotomy really is’ (Mintzberg, 1996, p 76). However, I disagree with his solution: breaking the tie with a three-domain classification of public, private, and ‘plural’ sectors (Mintzberg, 2015). That merely introduces a third, even more heterogeneous category. It still neglects other nonmarket innovations, like those undertaken by criminals and terrorists.

The deeper issue here is the idea of sectoral boundaries. As Patricia Bromley and John Meyer have argued, ‘it is increasingly difficult to distinguish between these historically separate entities’ (2014, p 939). They emphasize how difficult it now is ‘to determine an organization’s form (business, government, or charity) based on functional activity alone’ (Bromley and Meyer, 2014, p 957). And so, I support Mark Moore’s (2005) argument that we should not hold sectors as fixed. He contends that ‘when we define a sector, we hold a purpose relatively constant’ (Moore, 2005, p 48) and this restricts possibilities for breakthrough innovation. It is a self-fulfilling prophecy to assume that public and private organizations exhibit distinctly different innovation behaviours.

Remember, this is all moot because Pavitt’s taxonomy did not include a separate category for public organizations, and nor have its successors (for example, Castellacci, 2008). Rather, Pavitt and his heirs have treated the public sector as what biological taxonomists call a wastebasket taxon. This is the type of shadow category that catches everything seen as taxonomically unimportant. It is a parking spot for those entities that are so uninteresting that they do not really need classification (or so uninteresting that they could be classified later by some lower-level researcher). The public organizations I described in Chapter 4 – BIO, the Naval Research Establishment, and the Dalhousie Oceanography Department – are all lumped into the same taxonomic wastebin as the motor vehicle registration office. BIO’s technology development unit would only have become worthy of taxonomic classification when it spun-out as the company Brooke Ocean (to name but one example). Until that time, no one following Pavitt-style taxonomic logic would have noticed the development of technologies within BIO. The issue here is that public and private organizations are assumed to always exhibit fundamentally different innovation behaviours. Because public organizations are seen as substantially less innovative than private ones, they are seen as unimportant to the work of taxonomic classification. The careful study of public sector innovation has not placed them in a separate domain or kingdom; rather, systematic bias and lack of study have placed everything other than business in a taxonomic wastebasket.

Of the 120 ‘ocean tech’ companies I attempted to classify in Nova Scotia, one exemplified this point. The Fundy Ocean Research Center for Energy Inc. (FORCE) is an R&D joint venture (a not-for-profit corporation) established to provide shared infrastructure facilities for the testing of four different tidal power turbines in the Bay of Fundy (site of the world’s highest tides). It is an R&D facility jointly owned by the four competitor companies and their regulator (the Government of Nova Scotia). FORCE is also partnered with a variety of companies, government agencies, and research organizations (including my university). With its partners, FORCE developed a $50 million scientific sensor platform for real-time environmental monitoring, called Fundy Advanced Sensor Technology (FAST). This complex organization is not government, or business, or truly not-for-profit. It is also clearly intended to have a limited lifespan. Perhaps it will eventually become the site of a power generation facility – and then Pavitt’s (1984) proposed extra category might apply. But for now, what is important is that it does not fit in our innovation taxonomies. FORCE defies the taxonomic separation of market and nonmarket organizations – of the public, private, and social sectors. It is like a mythological creature concealed in an underexplored context. This brings us to our next problem.

The dragons of mare incognitum

As we know, Pavitt’s work was conducted in one specific temporal and geographical context. His taxonomy was based on data from the UK during the period 1945–79, but ‘is intended to be universally applicable’ (Gallouj, 2002, p 6). As Castellacci says: ‘Pavitt’s model … provides a stylized and powerful description of the core set of industrial sectors that sustained the growth of advanced economies during the Fordist age’ (2008, p 980, emphasis added). The limitation here is obvious. A large volume of research has documented the great variation in innovation patterns across temporal and geographical contexts rather than only sectoral ones (for more on this argument, see Castellacci, 2008). One test of Pavitt’s taxonomy across modern-day Europe found greater variability in innovation patterns between countries than between sectors (Castellacci, 2008). The model clearly cannot be fully generalized beyond the context in which it was created. And so, if we intend to continue solving this taxonomic puzzle, we must begin to explore unusual and understudied contexts; we need what Robert Yin (2009) calls ‘revelatory’ cases.

This is what drove major developments in biological taxonomy. It was a thriving field in Victorian-era England. In that time and place, societal expectations of animal diversity were being upended by the kangaroos, platypuses, and other creatures being ‘discovered’ during colonial explorations/exploitations of places like Australia (Ritvo, 1997). Back then, European explorers had little sense of what might be found on terra incognito. Indeed, the unexplored and potentially dangerous places on old European maps and globes were marked with illustrations of dragons and sea monsters. But over time, Europeans crossed the oceans and surveyed much of Earth’s landmass. Their biological taxonomy became a slowly dying field. Then came reports of Archaea. Mare incogitum – the unknown sea2 – captured the scientific imagination. And starting in 2000, a multidisciplinary and transnational team of 2,788 researchers undertook a ‘Census of Marine Life’ (Vermeulen, 2013). By 2010, those researchers had reported the discovery of 1,200 new species and had a backlog of 4,800 potential new species to be confirmed (Ausubel et al, 2010). New organisms were found in new contexts. Perhaps innovation studies could discover its own new ‘species’ within underexplored contexts like government or ‘ocean tech’? Some of these might be as odd as FORCE.

But let’s take this analogy one step further. Notice that the ‘revelatory’ contexts of Victorian-era taxonomy were only novel to Europeans. Before contact, hundreds of thousands of people were living in what the Europeans considered to be a hypothetical land – terra australis. Those Indigenous peoples were completely familiar with the local kangaroos and platypuses. As Harriet Ritvo shows, the biological taxonomy practices of 18th- and 19th-century England were situated within a culture of fear and fascination around ‘hybridity and cross-breeding’ – ‘monstrosity and monsters’ (1997, p xiii). Taxonomic classification systems and practices were entwined with Victorian England’s subjugation of women (especially sexual) and racialization (even ‘speciation’) of anyone without white skin. As we saw in Chapter 3, context is a construct. Shifting the sociopolitical ‘context’ helps reveal the violence of colonization, slavery, and subjugation. In her book, Ritvo (1997) (re)worked context to show the unnaturalness of zoology and biological taxonomy. Similarly, we can (re)work the context(s) that frame innovation taxonomies.

And it is not hard to see the contextual constraints of Pavitt’s postwar England. The geopolitical context is one matter. The temporal context is another. Indeed, the past 50 years have been given multiple labels – for example, ‘post-industrial’ (Bell, 1973), ‘late capitalism’ (Mandel, 1978), and ‘post-modern’ (Lyotard, 1984) – that all signal a departure from earlier times. But it is Jerry Davis (2022) who has most clearly articulated the implications of economic change for the classification of firms. In his book Taming Corporate Power in the 21st Century, he examines the problem of classifying ICT firms:

big tech companies such as Alphabet and Amazon use their expertise in ICTs to make money. They are indifferent to industry boundaries; they look for opportunities to apply ICTs in new ways that yield profits … They are not analogous to the railroads, oil, electricity, the telephone, radio, or a superhighway, because ICTs have become inescapable in human interaction. (Davis, 2022, p 44)

Pavitt’s taxonomy was built for the era of railroads, oil, and electricity – not the era of smartphones. Countless scholars have tried to ‘update’ Pavitt’s taxonomy, but the many adjustments and alternatives all maintain a product-based industrial structure (Archibugi, 2001; Gallouj, 2002; de Jong and Marsili, 2006). And as Davis shows us, product and process-based classifications do not work for ‘unicorns’ like Airbnb, Coursera, Uber, DoorDash, and Lyft (Davis, 2022). He names these companies as samples, drawn from

a rich and diverse buffet of tech nerditude that spans hotels, restaurants, schools, highways, infrastructure construction, spying, and, apparently, clouds. And yet all these businesses are classified in SIC code 7372 (‘pre-packaged software’). By tradition, this means we should regard them as competitors. Uber also classified itself as 7372, while its most obvious direct competitor, Lyft, went with 7389 [‘business services, not elsewhere classified’]. (Davis, 2022, p 39)

Davis is arguing that industrial classification systems are ill-suited to many of the most popularly innovative organizations in the early 21st century. He insists that the nature of these companies will contradict most fixed industrial categories. Godin (2005) made a similar point: ‘biotech’ does not fit into existing industrial classifications. These folk categories – biotech, infotech, ocean tech, clean tech, and so on – are rhetorical framings of different industrial contexts. We could get value from ‘sensibly’ mixing formal and folk classifications (Bowker and Star, 2000). But my point here is that these folk categories relate to a different industrial context than the formal ones. They reveal (and conceal) different patterns. This brings us to our third problem: innovation taxonomies are derived from Fordist-era industrial classification systems.

Industrial parataxonomy

In biological taxonomy there is a highly contentious methodological practice called parataxonomy. This approach involves sending nonspecialists into the field to collect large numbers of specimen samples and to pre-sort those samples roughly based on their most obvious physical characteristics (see Goldstein, 1997). Parataxonomy is meant to be an efficient division of labour between data collection and taxonomic analysis. However, in his critique of the method, Paul Goldstein (1997) explains that it is not necessarily more efficient – and it is certainly less effective at identifying priorities for diversity conservation. This is primarily because ‘the ability of parataxonomists to sort various groups must be tested repeatedly, as must the readiness with which various groups lend themselves to sorting by amateurs’ (Goldstein, 1997, p 572). Innovation studies is similar in that most taxonomic research has involved large datasets that were collected and ‘pre-sorted’ by others before the taxonomic researchers stepped in. Of course, I am not the first to notice this: de Jong and Marsili (2006) reviewed 11 important taxonomic contributions on organizational innovation and found that these all used some type of rough pre-sorting based on industrial product class. Indeed, a great deal of effort has been expended in innovation studies to highlight and test what I am calling the ‘parataxonomy problem’.

Unfortunately, product-based industrial classifications have remained embedded in the taxonomic literature since Pavitt (1984). Pavitt established explicit correspondence between his taxonomic categories and standard product-based industries. For example, he said that his ‘scale-intensive’ category included firms from industries such as food product manufacturing and shipbuilding. Meanwhile, ‘science-based firms are to be found in the chemical and the electronic/electrical sectors’ (Pavitt, 1984, p 362). However, Pavitt had intended for his model to be a firm-level taxonomy: ‘the basic unit of analysis is the innovating firm’ (1984, p 353). Archibugi (2001) argues that Pavitt failed to make this clear, instead giving the impression that this is a taxonomy of industrial sectors. That impression is reinforced throughout the original paper. Pavitt (1984) performed his econometric analysis using data he had aggregated up to the industry level, defined by the UK’s ‘Minimum List Heading’ (that is, categories of the Standard Industrial Classification). This made a great deal of sense, given his background in economic policy. But even in those times, it neglected considerable heterogeneity within each industrial class.

Since the Second World War, economists and statisticians have established a relative consensus around the product-related classification of firms. ‘Standard’ classification systems around the world (such as North America’s NAICS, Europe’s NACE, or the United Nation’s ISIC) are all very similar in this respect. They sort business establishments based on the primary products they produce. However, businesses with similar product outputs can exhibit completely different innovation behaviours. A good example might be two footwear companies: one mass-producing slippers, the other collaborating with NASA to produce moon boots (Archibugi, 2001). This is the aspect of Pavitt’s taxonomy that has been most heavily critiqued. For example, Archibugi (2001) has noted that multi-product, multi-technology firms defy the classification system. Gallouj (2002) has argued that the classification logic leaves no room for firms to change their product offerings or technological trajectories. Hoberg and Phillips have also lamented that none of the existing industry classifications ‘reclassifies firms significantly over time as the product market evolves’ (2016, p 1427). They add that the SIC, NAICS, and similar classification systems cannot ‘easily accommodate innovations that create entirely new product markets’ (2016, p 1427). And finally, Moore has pointed out that ‘organizations can succeed by migrating from one sector to another, while a sector can improve only by getting better at producing the goods and services defined by the sector’ (2005, p 48). In short, radical organizational innovation disrupts our existing industrial classifications. This means that many forms of radical innovation are excluded from data that has been pre-sorted by industry.

Cesaratto and Mangano (1993) may have been the first to statistically demonstrate that Pavitt’s sector-level analysis was problematic. They argued that Pavitt’s general idea was sound, but that innovation behaviours varied greatly among individual firms in Italy. De Marchi et al (1996) also tested Pavitt’s model using Italian firm-level data. They found some statistical support for Pavitt’s taxonomic categories, but also observed a high degree of firm-level variability within sectors. Other studies have confirmed that small firms in Switzerland (Hollenstein, 2003) and the Netherlands (de Jong and Marsili, 2006) are more diverse in their innovation patterns than Pavitt’s sector-based approach might allow. Leiponen and Drejer (2007) demonstrated that firm-level capabilities and strategies were more important to innovation behaviour than sectoral characteristics in both Finland and Denmark. Considering studies like these, Archibugi argued that improvements to Pavitt’s taxonomy should provide us with ‘a categorization of firms entirely independent from the product-based one’ (2001, p 420). He called for further innovation taxonomy research at the firm level rather than the industry level.

This is akin to Goldstein’s (1997) call for biologists to spend more resources on taxonomy and to engage in less parataxonomy. Indeed, it is easy to see parallels in the policy problems that result from parataxonomy in both biology and innovation studies. Goldstein was very concerned with the serious problems biological parataxonomy can create when its results are applied in conservation policy. Parataxonomy means sacrificing the effective identification of rare and potentially endangered species in favour of more efficient rough estimates of ecosystem diversity. Similarly, de Jong and Marsili expressed concern that sector-level analysis might make for easier innovation policy, but ‘it does not account for intra-industry diversity of innovation across firms’ (2006, p 216). In other words, it conceals novelty within and across its categories. It encourages policy makers to apply homogeneous innovation policies to heterogeneous categories of innovators. Ironically, stimulating novelty and heterogeneity is almost always a goal of these innovation policies. We are aware that this is problematic, and yet it continues.

The organism metaphor

There have been many published adjustments and alternatives to Pavitt’s taxonomy (see the discussion in Archibugi, 2001; Peneder, 2003; Castellacci, 2008). Because my focus is on the overall limitations of taxonomic logic, it is not my intention to discuss all published adjustments and alternatives. Rather, I have now reviewed the methodological challenges that persist in this literature (see also Archibugi, 2001; Gallouj, 2002; de Jong and Marsili, 2006). Many contributions to the innovation taxonomies literature end with some discussion of the various methodological limitations. These major methodological challenges – which I have summarized as the problems of wastebasket taxon, mare incognitum, and parataxonomy – must be addressed in future research.

But now that I have reviewed the challenges identified within the taxonomic literature, I would like to take a step back and explore the challenges that lie beneath it. I must now stop working with the biological metaphors and start working against them. In the following sections, I link taxonomic problems to three ways that organizations are not like organisms. The first of these is a reminder that organizations do not genetically ‘inherit’ their characteristics. Next is the agency of real organisms (humans) in the face of technological/organizational determinism. Finally, because organizations are not actually individual entities, they do not have any natural boundaries and so they often do not have functional unity. As we will see, these three limits of the organism metaphor are boundary conditions for the taxonomic classification of organizations. But first, let’s quickly review Gareth Morgan’s (1980, 1986) important contributions on the organizations-as-organisms metaphor.

Organizations as organisms

Forty years ago, Gareth Morgan (1980, 1986, 1997) wrote on the paradigmatic nature of metaphors in organization studies. He argued that the relatively ‘normal’ (cf. Kuhn, 1962) and coherent perspectives within an academic community ‘are based upon the acceptance and use of different kinds of metaphor as a foundation for inquiry’ (Morgan, 1980, p 607). Any one broad paradigmatic community is often home to more than one foundational metaphor. For example, the metaphors of ‘machine’ and ‘organism’ have long dominated studies of economics (Nelson, 1995), businesses, and organizations (Morgan, 1980, 1986). Academic communities operationalize their metaphors in the various ‘puzzle-solving’ activities (areas and methods of inquiry) that are seen to be normal (Morgan, 1980). They are tools for solving knowledge puzzles. Metaphors thereby enable and constrain research.

Good metaphors can advance our understanding of interesting phenomena, but only to a certain extent. As Morgan explained, ‘metaphor stretches the imagination in a way that can create powerful insights, but at the risk of distortion’ (1997, p 5). Through metaphor we produce ‘constructive falsehoods’ that can be useful until they are taken literally or to an extreme (Morgan, 1997, p 5). Researchers can neglect the limits of a metaphor and, more fundamentally, can also forget that certain theories are grounded in certain metaphors (Morgan, 1980). This is how a discipline can become ‘imprisoned’ (Morgan, 1980, p 605) by its metaphorical assumptions. The remedy is to recognize the limitations of a metaphor. Only then can a metaphor’s theorizing potential be fully realized. As Morgan said, ‘in recognizing theory as metaphor, we quickly appreciate that no single theory will ever give us a perfect or all-purpose point-of-view’ (1997, p 5).

In the next three sections, I am concerned with the limitations of the ‘organism’ metaphor for innovation studies. One might argue that innovation studies have ‘inherited’ the organism metaphor – and various related biological analogies – from economics. Freeman once warned that biological metaphors from economics presented ‘serious dangers’ (Freeman, 1991, p 211) for the study of innovation. Indeed, there is a longstanding debate about the utility of biology metaphors within economics, especially about the idea of Darwinian evolution (for example, Penrose, 1959; Nelson and Winter, 1982; Nelson, 1995; Hodgson, 2002). But I will only skim the debate about universal Darwinism (see Hodgson, 2002). The broad debate about the limits of Darwinian/biological metaphors has led to nuanced differentiation of economic ‘evolution’ from biological evolution (for example, Nelson and Winter, 1982; Nelson, 1995). However, the idea of ‘speciation’ is still contested; there is still debate around which characteristics of technology and organization can map onto a biological metaphor (for example, Ziman, 2003c; Langrish, 2017). Indeed, the neo-Darwinians who study innovation know that organizations are not exactly like organisms. But they nonetheless write of ‘taxonomic classification’ – and this carries the organism metaphor forward. Thus far, the puzzle-solving activity of taxonomic classification has been beyond the scope of the biological metaphor debate. My focus is therefore not on ‘evolutionary’ processes in economics per se; it is more tightly constrained to certain ‘taxonomic’ implications that arise from the organization-as-organism metaphor.

The ‘organism’ metaphor works because it allows us to treat organizations as relatively stable entities that exist within an environmental context. Morgan explains that

in the organismic metaphor the concept of organization is as a living entity in constant flux and change, interacting with its environment in an attempt to satisfy its needs. The relationship between organization and environment has stressed that certain kinds of organizations are better able to survive in some environments than others. (Morgan, 1980, pp 614–15)

This has led to some of the most important theoretical insights on business and organization (Morgan, 1986). For example, it is the metaphor that inspired systems theories and life-cycle theories of the firm (Morgan, 1986, 1997). It is a core assumption of the ‘population ecology’ approach to organization theory (Morgan, 1980, 1986), which brought Darwinian ideas into organization studies (for example, Hannan and Freeman, 1977). It is at the centre of the sociotechnical systems approach (Morgan, 1986), which examined the mutually constitutive nature of organizing and technology (for example, Trist and Bamforth, 1951). And in the context of this chapter, it is particularly important to note that the organism metaphor is the foundation for contingency theory (Morgan, 1980, 1986), which theorized different forms of organization as adaptions to different environmental conditions (for example, Burns and Stalker, 1961). It was early contingency theorists who noticed how different forms of organization were linked with different technologies (for example, Burns and Stalker, 1961; Woodward, 1965). This, of course, led to decades of research into the various ‘species’ of organization (Morgan, 1997). However, all this theorizing has found its limits. Following Morgan (1980, 1986), I argue that those limits exist because organizations are not biological organisms. In fact, what these theoretical traditions all share is that they reify the processes of organizing; they metaphorically create a real and concrete entity (an ‘organization’) out of sociomaterial interactions.

‘Inherited’ characteristics

Throughout the debates on biology metaphors in economics, the most hotly contested question has been whether organizations ‘inherit’ their traits. And when it comes to innovation taxonomies, a central assumption ‘is that firm behaviour is shaped and constrained by the nature of the technologies they use’ (de Jong and Marsili, 2006, p 214). In other words, organizations inherit their innovation behaviours through their technologies. But how far does this ‘inheritance’ analogy (and, by extension, the evolution analogy) stretch? What are its limits? In building his argument for universal Darwinism, Hodgson admits that ‘the strongest reasons to be skeptical of “biological analogies” involves the detailed differences between the types of evolutionary mechanism applying to the socio-economic and to the natural domain’ (2002, p 274). He argues that the concept of inheritance (and replication) is where universal Darwinism will ‘find its boundary’ (Hodgson, 2002, p 273). In this section, I argue that the idea of inherited traits takes us beyond the useful bounds of the organism metaphor.

To rethink the ways in which organisms and organizations are different, consider one of biology’s ongoing taxonomic problems: the classification of larvae. For a period in the early 19th century, ‘zoea’ were considered a type of crab (a genus). Then taxonomists realized that ‘zoea’ was a larval stage in crab development. It is sometimes difficult to distinguish between larvae and adult animals because classic taxonomy relies on the identification of unique physical characteristics. Of course, these can change dramatically as the research subjects mature. When ocean scientists are searching for new species, the samples returned in their nets can often be difficult to sift through, and trained taxonomists tend to have large backlogs of work. In response, many scientists turned towards DNA barcoding (Snelgrove, 2010), which is not only faster but can also work around the maturity problem (although, as we will see later, even this has its limits). Generally, larvae and adults of the same species contain the same genetic material. Indeed, that genetic material is what determines the stages of an organism’s life cycle and its physical features at each stage. Although there have been attempts to develop life-cycle taxonomies of firms (for example, Abernathy and Utterback, 1978), it is difficult to argue for material determinism in organizational life. As Freeman (1991) asserted, the life cycle of firms is not like the life cycle of organisms. There is no DNA barcoding solution for the study of organizations.

And there is no debate on this point: organizations do not have DNA. Nelson and Winter may have argued that, in economics, ‘routines play the role that genes play in biological evolutionary theory’ (1982, p 14), but they were aware that this was an analogy. Nonetheless, Hodgson (2002) took their argument one step beyond metaphor: ‘One possible and relevant example is the propensity of human beings to communicate, conform and imitate, making the replication or inheritance of customs, routines, habits and ideas a key feature of human socio-economic systems’ (Hodgson, 2002, p 270). Based on this kind of reasoning, some scholars have come to accept ‘memes’ as a corollary to ‘genes’ within evolutionary economics. But Ziman (2003a) characterizes the move towards ‘meme’ as merely a ‘convenient’ way to ‘sustain the overall analogy’ (2003a, p 5). Witt (1996) has forcefully argued that there is nothing resembling genetic material at the social level. Louçã and Cabral (2021) have asserted that ‘no economic analogue exists for the replication unit in biology’ (2021, p 4). Patterns of human behaviour – like organizational routines – are simply not analogous to genes; rather, because humans enact organizations, their genetic material is only part of any organizational process (Vromen, 2006).

Organizing involves social interaction between humans. It also involves layers of social behaviour between other constituent materials, such as technological artefacts. Since the earliest days of the organization-as-organism metaphor, we have known that social and technological realities are mutually constitutive (Trist and Bamforth, 1951; Burns and Stalker, 1961; Woodward, 1965). We know that organizing is a sociomaterial process. But the organism metaphor still implies that organizations are materially determined – in the same way that genetic material determines the make-up of an organism. Indeed, by 2004, Nelson was recommending that we abandon ‘the search for credible analogues of genes in economic processes’ (cited in Vromen, 2006, p 544). There is no definitive DNA barcode that we can sample; firms do not have fixed, technically determined characteristics.

But this has not stopped us from overextending the analogy of inherited characteristics. In taxonomic logic, organizations are seen as being ‘locked’ into a particular sectoral trajectory (Gallouj, 2002). Or, as de Jong and Marsili (2006) explain, technological regimes determine ‘the directions, or ‘natural trajectories’, along which incremental innovations take place within the regime’ (de Jong and Marsili, 2006, p 215). There is most certainly a path dependency to innovation behaviours (Cohen and Levinthal, 1994). But the organization-as-organism metaphor neglects the very important role of human agency (Morgan, 1986). Morgan (1986) notes that this is one of the major limitations of the organism metaphor: it easily gives way to an ideology of natural determinism and Social Darwinism.

Innovative agency

I join others in arguing that the organism metaphor leads us towards an overly deterministic perspective on innovation and organization(s). These arguments go back at least as far as Edith Penrose (1959), who said that we should reject theories that treat firms as if they are organisms because such theories neglect human agency (see also Levallois, 2011). She said ‘to abandon their [firms’] development to the laws of nature diverts attention from the importance of human decisions and motives, and from problems of ethics and public policy, and surrounds the whole question of the growth of the firm with an aura of “naturalness” and even inevitability’ (Penrose, 1959, p 809). Chris Freeman applauded her argument and applied it to innovation theory. He agreed that we should not ‘give explanations of human affairs that do not depend on human motives’ (Freeman, 1991, p 219). I might add the agency of other actors to the entanglements we call organizations/organizing (MacNeil and Mills, 2015). But then we simply get a more complex and distributed sociomaterial agency. Nonetheless, Penrose was right to assert the agentic power of human managers and thereby confront the conservative bias (Levallois, 2011) that we find entwined with the organism metaphor.

Contingency theories of the firm developed around this sense of agency. These theories focus on how managers can adapt their strategies (and organizational forms) to their environments. But some theorists argued that this allocated ‘too much flexibility and power to the organization and too little to the environment’ (Morgan, 1997, p 60). This led to the more Darwinian ‘population ecology’ approach to organizations (see Hannan and Freeman, 1977). Although it produced some valuable insights, many organization theorists see population ecology as far too deterministic (Morgan, 1997). Through an overextension of the organism metaphor, the population ecology view can take the ‘rather pessimistic stance that this choice [human agency] will never count for much because environmental forces ultimately have the upper hand’ (Morgan, 1997, p 68). This downplays or discounts the role of human creativity in organizing and the ability for organizations ‘to create market niches that never existed before’ (Morgan, 1997, p 63). In short, it discounts many possibilities for innovation.

We cannot afford to take the organism metaphor this far in innovation studies. As Vromen (2006) suggests, ‘the irregular, unpredictable parts in firm behaviour might be considerable’ (p 559). Indeed, my view is that those irregular, unpredictable parts should be a primary focus of innovation studies. But the organism metaphor focuses us deterministically on inherited traits and the ways that organizations conform to their contexts. There are certainly material aspects of organization. But unlike organisms, organizations are also socially constructed (Morgan, 1986).3 This is the greatest limitation of the organism metaphor: ‘Organizations are very much products of visions, ideas, norms, and beliefs, so their shape and structure is much more fragile and tentative than the material structure of an organism’ (Morgan, 1997, p 69). The social shaping of technology (MacKenzie and Wajcman, 1999) is taken for granted in science and technology studies. However, innovation taxonomies are much more deterministic.

Taxonomies of innovation assume that organizations exhibit patterns of behaviour that are determined by – or inherited from – their technical reality. This assumption is what allows us to place organizations into mutually exclusive categories. But it only works for organizations that have stable technological regimes/trajectories. They must have characteristics that are determined as if by some genetic/inheritance mechanism. An organism cannot choose to change its phylogenic characteristics, but organizational decision making can change innovation behaviours. As I have already noted in this chapter, some organizations undertake radical change. In fact, some organizations are known to completely redefine their domains (Covin and Miles, 1999). Taxonomic logic could be applied to these organizations, but they would need to be awkwardly classified as different species before and after any radical change (much like our crab larvae, zoea). I therefore conclude that the boundaries of the organism metaphor do not extend to radically innovative organizations that are in the process of establishing new sociotechnical realities. This metaphor leaves those organizations and innovations in the dark. Again, we see that taxonomic classification is unsuited for observing some of the most interesting innovation phenomena.

Functional (dis)unity

The organism metaphor’s other major boundary condition is interesting because it relates directly to the ‘real’ boundaries of organizations. Treating organizations as organisms allows us to reify clear limits around our unit of analysis: to treat organizations as open systems. Morgan explains this part of the analogy:

If we look at organisms in the natural world we find them characterized by a functional interdependence where every element of the system under normal circumstances works for all the other elements. Thus, in the human body the blood, heart, lungs, arms, and legs normally work together to preserve the homeostatic functioning of the whole. The system is unified and shares a common life and a common future. (Morgan, 1997, p 70)

However, organizational processes are not unified in this way. Organizations do not have natural boundaries (except sometimes when they operate in a single building). And so, they seldom have functional unity. At the extreme, some criminals and terrorists find innovative ways to avoid functional unity in their organizing.

Elsewhere, Albert Mills and I have shown that the idea of an organization is a black box, and the processes of organizing are much more precarious than we typically realize (MacNeil and Mills, 2015). The organism metaphor helps us to temporarily suspend the messy processual realities of organizing and thereby study discrete, stable ‘entities’. To put it another way, ‘organizations are but temporary reifications, because organizing never ceases’ (Czarniawska, 2004, p 780). It is often empirically useful to temporarily ‘pin down’ organizational life like an entomologist might pin down an insect. The problem is that the organism metaphor determines our unit of analysis. As Morgan says, organizations ‘are not discrete entities, even though it may be convenient to think of them as such’ (1997, p 64). Ontologically, organizations are not really (materially) individuals.

This has manifest as a messy methodological problem for taxonomies of innovation. Although they followed the normal logic of the organism metaphor, de Jong and Marsili (2006) noticed this problem. In their final sentence they say: ‘A suggestion for future work is to integrate the different levels of analysis and disentangle the influence that conditions specific to single innovations and to industrial sectors have on the diverse clusters of innovative firms’ (de Jong and Marsili, 2006, p 227). And so, it has been acknowledged that sometimes ‘innovations’ are embedded within ‘organizations’ which are embedded within ‘sectors’. But with so many possible units of analysis, which is the right one? Do we go ‘down’ to the level of the technique or routine, as argued by Joel Mokyr (2013)? Or do we follow Nick Oliver and Michelle Blakeborough (1998) towards the distribution of innovation across networks of multiple organizations? If we had only these two choices, I would be inclined towards the latter – because Oliver and Blakeborough (1998) included two scientific instrument innovations among the seven aggregated within their study. However, when it comes to taxonomic classification, the organism metaphor tells us what to do: we treat organizations as the embedded unit of analysis within industrial sectors. Beyond the metaphor, we know that innovation processes are not bounded within organizations in the way that biological processes are seen to be bounded within organisms.

To understand this limitation, let us turn to the ‘coral reef problem’ from early biological taxonomy. (Please hold this example loosely; I am merely turning the organism metaphor in on itself, before tossing it aside.) The ‘rocks’ that most people consider to be coral are actually the limestone exoskeletons of many individual polyps, smaller than your littlest finger, living in densely packed colonies. These polyps also live symbiotically with algae, sometimes inside their bodies. Turning the original Linneaen taxonomy on its head, corals are therefore part animal, part plant, and part mineral. Today it is much easier for a taxonomic biologist to distinguish between each part of a coral. Algae living within a polyp can be identified as separate and distinct organisms through genetic barcoding. But two major difficulties arise when we try to do the same for discrete, mutually exclusive types of organization: the hybridity problem and the symbiosis problem.

Hybridity

First, we struggle to classify hybrid organizations. De Jong and Marsili (2006) explain this as a methodological limitation of taxonomic research: ‘the same firm may implement various types of innovations … each single innovation may display different patterns’ (2006, p 227). However, shifting the unit of analysis to a smaller (or larger) scale does not guarantee that we will find functional unity. That would be akin to searching for algae inside a polyp. But with organizing we will never find DNA. It makes more sense to address this problem by shifting away from the organism metaphor entirely, since ‘unlike in nature, where species are distinguished by discrete clusters of attributes, organizational characteristics are often distributed in a more continuous way. One form often tends to blend with another, producing organizations that have hybrid characteristics’ (Morgan, 1997, p 55).

This hybridity is possible at all sizes and scales of organizing. Neglecting it is extremely problematic for innovation studies because if we follow Schumpeter’s (1934) logic – if innovation is, indeed, the implementation of new combinations – then innovation processes will always produce hybrids that defy classification. Certainly, Ritvo (1997) has shown that biological taxonomy developed through cultural opposition to hybridity. Hybrid animals like the platypus ‘were stigmatized … not only as mongrels but as “monsters”’ (Ritvo, 1997, p 132). She quotes the famed English wood-engraver and naturalist Thomas Beswick to illustrate some of the beliefs that informed biological taxonomy: ‘Nature has providently stopped the … propagation of these heterogeneous productions, to preserve, uncontaminated, the form of each animal; without which regulation, the races would in a short time be mixed with each other, and every creature, losing its original perfection, would rapidly degenerate’ (Beswick, cited in Ritvo, 1997, p 89). Presumably, no innovation scholar today would ascribe to such a position. Hardly any would be aware of the racist context surrounding early taxonomic practices. But the fact remains that taxonomic classification in biology was imbued with the values of purity and pedigree. Taxonomic classification will always be at odds with hybrid entities.

Symbiosis

Attempts to identify discrete, mutually exclusive types of organization can also miss symbiotic, mutually dependent organizations. Again, this has been framed as a methodological limitation for innovation taxonomies. For example, Gallouj (2002) criticized Pavitt’s taxonomy for missing the possibilities of innovation coproduction between service providers and their customers. But I argue that the limitation here is found in our metaphor, not our analytical methods. Consider the food chain and food web metaphors for a moment (again, for brief rhetorical effect). When scientists were working within the ‘food chain’ paradigm, marine bacteria did not seem to be very important. The linear chain started with phytoplankton performing photosynthesis and proceeded through progressively larger marine animals. Then better techniques allowed biologists to understand the important role of various bacteria in digesting dissolved organic matter. These bacteria are consumed by single-cell zooplankton, some of which are taken up by larger organisms, but many of which cycle back into dissolved organic matter (and become food for the bacteria). These previously unnoticed microbial loops form the foundation of the marine food web. It took a switch from the food ‘chain’ metaphor to a food ‘web’ metaphor to recognize the importance of this marine subsystem.

The logic of existing innovation taxonomies predicts vertical chains between organizations in different categories. This is based on the linear thinking of inputs and outputs. But we know that there are interesting innovation behaviours in this world that depend on the strategic mutual dependence of two or more different organizations. For example, consider Mowery’s work on innovation within the military-industrial complex (for example, Mowery, 2009). As Mowery and Nelson (1996) have said, ‘corporations are one part of a complex institutional system, and their role cannot be understood in isolation’ (p 189). Also consider Gorm Hansen’s (2011) research on the symbiotic relationship between a biology research laboratory and a biotech firm. These two organizations were innovating as one – a phenomenon that is invisible to innovation taxonomies. The organism metaphor fails to classify these forms of organizing because the innovation processes do not have functional unity within a concrete boundary.

Moving beyond speciation

Away from biological metaphors

This chapter has considered the challenges of taxonomic classification in innovation studies. Many scholars have attempted to solve the taxonomic puzzle and their insights have had a considerable positive impact on innovation theory and policy (de Jong and Marsili, 2006). Overall, taxonomic classification has helped us to reduce empirical complexity by establishing ‘few and easy to remember categories’ (de Jong and Marsili, 2006, p 214). From Pavitt (1984) onwards, the resulting taxonomies have guided researchers and policy makers in finding, recognizing, and reinforcing innovation behaviours. But as we have seen, these taxonomic classification tools have clear methodological and ontological limitations. They conceal many interesting forms of innovation.

We could reveal some dark innovation by improving these taxonomic tools. For example, we could drop the practice of industrial parataxonomy. But, as we have seen in this chapter, some of the taxonomic issues are inseparable from the biological analogy that enables this whole exercise. The organism metaphor is useful in many ways, but all metaphors have limits (Morgan, 1980, 1986). As Morgan said, ‘any one metaphorical insight provides but a partial and one-sided view of the phenomenon to which it is applied’ (1980, p 611). The challenge is that individual metaphors can powerfully shape social scientific paradigms (Morgan, 1980, 1986).

In the 1990s, a collaboration of British researchers spent four years considering the extent to which biological metaphors should be applied to technological innovation. This “Epistemology Group” could not establish a consensus on either maintaining or abandoning analogies to biological evolution (Ziman, 2003c), but they did conclude that we should reject the idea of speciation: ‘we no longer feel impelled to find technological analogies for the most familiar evolutionary concept in biology – the notion of a species’ (Ziman, 2003c, p 313, emphasis in original). John Ziman was lead author of the book that arose from this collaboration, and his chapters were the only ones to directly confront taxonomic classification (albeit briefly). In his opening chapter, Ziman argued that we should ‘give up such Procrustean exercises as trying to make industrial firms look just like organisms’ (2003a, p 11). And in a later chapter he bemoaned that ‘taxonomies that mirror “the tree of life”’ are ‘all too tempting’ (Ziman, 2003b, p 46). Unfortunately, many innovation researchers still succumb to the temptation.

Perhaps this is because taxonomic classification is such a unifying tool for such a loosely defined field of study. Although Chris Freeman ‘voiced strong reservations about the uncritical translation of biological concepts’ throughout his career (Louçã and Cabral, 2021, p 2), he once said that ‘a taxonomy is essential both for analytical purposes and as a tool for empirical research [in innovation studies]’ (Freeman, 1991, p 222). He knew that this tool both helped and hindered innovation research. Indeed, he argued that ‘any such taxonomy or classification system must of course do some violence to the infinite complexity of the real processes of technical and economic change’ (Freeman, 1991, p 222). I have concurred with him on this point. But my arguments in this chapter have diverged from Freeman’s very early assertion that ‘all schemes of classification are to some extent arbitrary and artificial’ (Freeman, 1974, p 261). Innovation taxonomies are not arbitrary. By exploring the metaphors that enable this exercise, I have surfaced some of the assumptions carried from past sociopolitical contexts. Pavitt’s taxonomy was key in the formation of innovation studies as a field (Fagerberg and Verspagen, 2009). But it was also, perhaps unintentionally, one inscription point for conservative, neoliberal ideas about innovation.

In their book Sorting Things out: Classification and Its Consequences, Geoffrey Bowker and Susan Leigh Star explain that ‘standards and classifications, however dry and formal on the surfaces, are suffused with traces of political and social work’ (2000, p 49). And we have now seen some of the social and political traces that shaped taxonomies of innovation. Nonmarket organizations were set aside. Assumptions about industrial structure were anchored in postwar England. Standardized industry categories were adopted for methodological efficiency. These processes, which I discussed in the first half of the chapter, have been somewhat easy to identify in the taxonomic literature. However, the literature treats these decisions as apolitical. Of course, they cannot be. Bowker and Star note that ‘each standard and each category valorizes some point of view and silences another. This is not inherently a bad thing – indeed, it is inescapable’ (2000, p 5). The bad thing is to deny these politics.

We have ‘consistently ignored’ the insidious conservative bias attached to biological analogies (Levallois, 2011). Harriet Ritvo (1997) showed us how biological taxonomy became imbued with Victorian English values. Taxonomic practices developed in favour of purity and against hybridity. And Edith Penrose could see the conservative bias in her time as well – a time when Social Darwinism was resurging and her friends were being persecuted under McCarthyism (Levallois, 2011). In addition to the material harm caused to people with progressive views (such as McCarthy’s ‘communist’ witches), Penrose could see the harm to organizational theory: ‘the chief danger of carrying sweeping analogies very far is that the problems they are designed to illuminate become framed in such a special way that significant matters are frequently inadvertently obscured’ (Penrose, 1952, p 804). At various points in this chapter, we have seen that the combined organism-species-taxonomy analogy obscures many forms of innovation. There is the neoliberal effect where firms are framed as the primary organisms for innovation. But there is also the obscuring of heterogeneity, divergence, hybridity, and symbiosis – processes that ought to be celebrated in innovation research.

In Chapter 2, I argued that theoretical models are social scientific instruments. I said that even when models are tacit, we can think of them as noncorporeal actants (Hartt, 2019). They work through physical traces, and through our sensemaking, to co-construct knowledge. Now, we can add theory-laden metaphors to that mix. The organization-as-organism metaphor is a noncorporeal actant in our taxonomies of innovation. It is entwined in a sociomaterial knot of taxonomic instrumentalities. This enables and constrains our understanding of innovation. Once we start looking, we find bits and pieces of biological analogy everywhere in the taxonomic toolkit. As Bowker and Star state, ‘all classification and standardization schemes are a mixture of physical entities, such as paper forms, plugs, or software instructions encoded in silicon, and conventional arrangements such as speed and rhythm, dimension, and how specifications are implemented’ (2000, p 39). In other words, classification systems appear highly structured, but they are a messy sociomaterial stew.

Towards other classification tools

Because the organism metaphor is taken for granted, my arguments here are likely to encounter considerable resistance: ‘Schools of theorists committed to particular approaches and concepts often view alternative perspectives as misguided, or as presenting threats to the nature of their basic endeavour’ (Morgan, 1980, p 613). For those who might object in this way – those who are wedded to some adaptation of the biological analogies – let me point out that biological taxonomy is undergoing a similarly uncomfortable paradigm shift. You see, those microbes that were found in the deep ocean – the ones we now call Archaea – have been found to actively swap genetic material. They cannot be classified based on their DNA. Helmreich explains that this means ‘the stability of the category of species for microbes has been called into question’ (2009, p 87). He suggests that ‘the tree of life might turn out to be a net’ (Helmreich, 2009, p 82) and that biology might be headed in a direction that ‘strikes a familiar chord with readers of the maniac philosophy of Gilles Deleuze and Félix Guattari’ (Helmreich, 2009, p 83). Wherever biology might be headed, it too must grapple with the design of its taxonomic tools. Too many organisms in the deep sea do not fit within a classification system that carries assumptions from the land. Something similar is true for innovation studies, where there are known instances of innovation that do not fit within old, British, industrial assumptions. And there are most certainly ‘known unknowns’ – readily identifiable instances of hybridity, symbiosis, and other kinds of dark innovation – that cannot be directly observed using the organism metaphor.

Interestingly, innovation studies might be primed for a shift. Ziman (2003a) has already suggested that technological innovation might be better classified using a ‘neural net’ metaphor. Meanwhile, Hoberg and Phillips (2016) have developed an approach to classifying publicly traded firms in the US based on the interrelatedness of text in their public disclosures – an approach they call ‘textual network industry classification’. But networks (and rhizomes) are not the only possible alternative classification devices. At one time, before the tree of life became locked in, there were other unusual zoological classification systems, such as the ‘quinary’ approach of grouping organisms around interlocking circles (Ritvo, 1997). In the next chapter, I look to Law and Singleton (2005), who considered four metaphors – region, network, fluid, and fire – for theory building in science and technology studies. But regardless of the specific metaphors we choose, let’s heed Morgan’s advice: ‘in order to understand any organizational phenomenon many different metaphorical insights may need to be brought into play’ (Morgan, 1980, p 613). The overuse of this one metaphor has focused research on those innovation phenomena that fit. Other metaphors will help illuminate dark innovation.

The organism-species-taxonomy exercise has found its limits. This does not mean that classification should cease; rather, it means we should explore new metaphors ‘which overcome the weaknesses and blindspots of traditional metaphors, offering supplementary or even contradictory approaches’ (Morgan, 1980, p 612). I take up the challenge of new metaphors in the next chapter.