It was too early on a Monday morning to make sense of the massive machine in front of me. The tour guide introduced it as North America’s largest geotechnical centrifuge. But it looked like a poorly designed carnival ride: two boxes, each big enough to hold a human, on either end of a massive rotating arm. The official size: 200G. Not grams, but ‘g’ forces (I had to look it up). This thing can produce forces equal to 200 times Earth’s gravity. And there it was, sitting in a cement bunker at Memorial University – on the edge of the Atlantic Ocean and as far east as one can get while still being in the Americas. I was confused: why here? What for? (And where’s the coffee?)

I had come to the city of St John’s, in the province of Newfoundland and Labrador, to attend Oceans ’14 – an international ocean technologies conference. This tour of the Memorial University facilities was a warm-up event for the highly technical conference sessions. It had been six years since the IEEE Oceanic Engineering Society and the Marine Technology Society had convened their joint annual conference here in Canada. So, this was a rare chance to get a ‘broader’, ‘global’ sense of ocean science and technology. It was also a chance to see ocean technology at play in a different province. (Never mind that my wife, toddler son, and newborn daughter could tag along to visit relatives.)

Some people had told me that ocean technology was ‘different’ in Newfoundland than in Nova Scotia. And the differences began to resonate in those first hours of the conference. I was disoriented by a deluge of engineering talk as our tour group walked through Memorial’s facilities. At home, my tours of university and public research facilities were almost exclusively focused on the work of ocean scientists. Their gizmos and gadgets – even the massive ‘Aquatron’ research tank at Dalhousie – were props for stories of scientific intrigue. But at Memorial, the stories were about the technologies themselves. The centrifuge was one thing: it is used to model interactions between sea ice and the land. Elsewhere at the C-CORE lab, my tour group heard about the development of technologies to detect icebergs from space and to protect vessels/oil rigs from iceberg impacts. At the Autonomous Ocean Systems Lab, we heard about the development of technologies to propel and control underwater autonomous vehicles. These vehicles might carry scientific instruments and enable new oceanic observations. Indeed, we heard a bit about the amazing deep-sea cameras developed by my wife’s high school friend, Dr Adam Gobi. He was the lead engineer on the camera systems that James Cameron took into the Mariana Trench. Cameron’s film Deep Sea Challenge had been released only a month before the conference. But even here, when the facility tour turned to questions of exploration, the emphasis was on the engineering feats – the advancements in techne. New scientific discoveries – epistemic advancements – were mere ellipses at the end of each story.

I left Newfoundland convinced that ocean technology was very different there than at home in Nova Scotia. One place privileged ocean engineering and the other privileged marine biology. And yet, there were people in my ear suggesting that I should merge these into one region – that I should include all four Atlantic Provinces in my study. This would match the Government of Canada’s approach: its Atlantic Canada Opportunities Agency has a four-province mandate. I could find very little in the way of ocean technologies in the other two of those provinces. But apparently federal officials wanted to encourage inter-provincial collaboration and thereby reduce resource competition between Memorial University (in Newfoundland) and Dalhousie University (in Nova Scotia). There was one speaker at the conference whose talk extended the boundaries even further. Dr Doug Wallace, the Canada Excellence Research Chair in Oceanography at Dalhousie University, lumped all of Eastern Canada – the 2,500 km drive (and ferry ride) from Montréal to St John’s – into one ‘ocean observation technology cluster’. He also spoke about promising tripartite discussions between Canada, the US, and the European Union that might lead to a collaborative scientific observation system for the ‘whole’ North Atlantic Ocean. I liked his ideas. But frankly, I just needed a clear sense of where to start and stop my PhD research. All this talk of shifting boundaries, and the ‘global’ nature of this conference, made that problematic.

In this chapter, I confront the problems of boundary specification for innovation studies. I begin by considering how regional or volumetric thinking – one form of spatial reasoning – drives the systems of innovation literature right into a theoretical brick wall. That wall is made of the surveyable, measurable, Euclidean spaces we call innovation systems. The wall is buttressed by a static form of institutional theory; institutional fields are conceived as containers for interactive learning networks/relationships. But this makes it difficult to observe innovation processes that fold people, things, and places together in new ways. In the first parts of this chapter, I discuss the ideas underpinning this approach to space, place, and innovation. I then describe how those ideas bounded my survey of innovation in ocean science instrumentalities. All this requires epistemologically positivist and ontologically realist assumptions. We will see that this form of thinking allows us to observe only certain kinds of innovation, in defined fields of technology, and in particular places.

In the conclusion to this chapter, I explore what might be missing from the regional/volumetric view. I argue the need for other forms of spatial reasoning in innovation studies. I take ‘object lessons’ from John Law and Vicky Singleton, and briefly consider the different implications of framing innovation as a region, network, fire, or fluid object (Law and Singleton, 2005). To understand the limits of thinking with region and network metaphors, I share ‘excess’ data that flooded and set fire to the boundaries I had established for my survey of scientific instrumentality innovation. In this chapter I ask: what observations about innovation are left in the dark by an over-reliance on one or two spatial metaphors?

Regional networks

Innovation system boundaries

Let’s begin with the dominant spatial perspective: regional networks. Much innovation research works from the premise that mappable regions contain networks of interactive learning. This is almost taken for granted today. Of course, regional approaches to innovation have a long history. We could go back to the knowledge that was suspended ‘in-the-air’ of Alfred Marshall’s (1890) industrial districts. Or Chris Freeman (1995) would have us go back further and notice the (possible) influence of Frederich List’s National System of Political Economy (1841). But hardly anyone today has read those works. The ideas they contain are dead ancestors to the modern systems of innovation. And please note the double entendre in my use of the word modern. Many scholars would argue that innovation systems are the most modern theoretical framework. But in this chapter I will contend that the systems of innovation approach is an outdated exercise in social science modernism.

In Chapter 2, we saw the beginnings of the innovation systems approach in Christopher Freeman’s writing about Project SAPPHO. In that preincarnation, the UK was an implicit boundary within which Freeman could think about the systemic nature of innovation. Freeman brought geopolitical boundaries to the forefront in his 1987 book. There, he explained how unique institutional arrangements, such as industrial groups or keiretsu, developed into an effective ‘national innovation system’ (NIS) for postwar Japan (Freeman, 1987). Over time, innovation systems would take on a variety of subnational and supranational boundaries. Charles Edquist (1997), who appreciated the value of vague boundaries, emphasized the many overlapping possibilities: ‘innovation systems may be supranational, national or subnational (regional, local) – and at the same time they may be sectoral within any of these geographic demarcations. There are many possible permutations. Whether a system of innovation should be spatially or sectorally delimited depends on the object of study’ (Edquist, 1997, p 12). And so, there is a wide menu of boundary options in this approach. Sometimes, researchers justify their boundary choices in writing. But even then, many innovation systems are bounded by convenience. Indeed, this has been so common that Edquist and Björn Johnson (1997) felt the need clarify the theory involved. They argued that the boundaries around an innovation system are ‘always’ (Edquist and Johnson, 1997, p 60) or at least ‘normally’ (1997, p 40) defined in terms of institutions. In other words, boundary specification is meant to be guided by institution theory.

This often does not happen because there are two different uses of the term ‘institution’ in the innovation studies literature (Edquist, 2001; Coriat and Weinstein, 2002; Grønning, 2008). Some work uses the term for a special category of organizations (Edquist and Johnson, 1997; Grønning, 2008). Here, ‘institution’ is a euphemism for various public organizations, especially public research organizations (Coriat and Weinstein, 2002). This is linguistically accurate. But at this point it should be clear why I would oppose that use of the term ‘institution’. It lumps a variety of public organizations into the support function of an innovation system. It helps us assume an a priori role for certain types of organization. Unfortunately, this is the approach that has been taken by prominent innovation scholars like Freeman, Nelson, and Rosenberg (Edquist, 1997). But some scholars have argued for a more theory-laden use of the word ‘institution’. Lundvall, Edquist, and others have engaged directly with institutional economics, focusing on the formal and informal rules and routines that shape organizational interaction (Edquist and Johnson, 1997).

Those who have tried to untangle this issue (see Edquist and Johnson, 1997; Coriat and Weinstein, 2002) have turned to Nobel Laureate Douglass North (1990). North argued for a distinction between manifest institutions (that is, organizations) and abstract institutions (that is, rules/routines). Invoking a sports metaphor, he said that ‘what must be clearly differentiated are the rules from the players’ (North, 1990, p 4). North argued that because any institutional theory must ‘begin with the individual’ (1990, p 5) and focus on ‘groups of individuals bound by some common purpose to achieve objectives’ (1990, p 5). In short, his approach encourages the separation of organizations, professions, and other social groupings from the explicit and implicit rules they follow. Edquist and Johnson (Edquist, 1997, 2004; Edquist and Johnson, 1997) argued that research on innovation systems should be grounded in North’s institutional theory approach. Based on North’s work, Edquist and Johnson (1997) concluded that we should ‘deduct’ legally constituted organizations from the interactive learning that occurs outside or between organizational entities. Similarly, Casper et al discuss a separation between institutions and ‘the interaction of individuals and groups within a particular institutional setting’ (2005, p 197). If we fully deploy North’s (1990) sports metaphor to an innovation system, we can think of organizations as the players, institutions as the rules, and interactive learning as the sport or the play. Since the work of Edquist and Johnson (1997), there has been an ‘increasing consensus’ within the innovation systems literature that institutions and organizations should be defined in this way (Grønning, 2008).

So now we have organizational players taking to the field. What sport do they play? Lundvall (1992) describes the innovation systems approach as a ‘focusing device’ – a kind of social scientific theory – that places our attention on different processes of ‘interactive learning’. This emphasis arises from an underlying assumption that ‘the most fundamental resource in the modern economy is knowledge, and, accordingly, that the most important process is learning’ (Lundvall, 1992, p 1). Note that Lundvall’s ‘focusing device’ does not point towards the noun knowledge or any other static outcome. Instead, Lundvall’s (1988, 1992) earliest contribution to the innovation systems approach was to embed an understanding of innovation as a verb: an ongoing, ubiquitous, and cumulative learning process. Meeus and Oerlemans explain that, in the context of the literature on innovation systems, learning is ‘a process in which all kinds of knowledge are (re-)combined to form something new’ (2005, p 159). And so, the contents of innovation systems have been increasingly conceived as networks. As we will see later, these are not networks in the rhizomatic poststructural sense. ‘Networks’ in innovation studies are almost exclusively plotted within Euclidean, regional spaces. This is positivist, realist, modernist thinking. Social network analysis is the primary analytical tool. It is used to chart the learning relations within and between geographical regions.

In summary, an innovation system can be described as an institutional field in which interactive learning takes place. The boundaries of that field typically correspond to national borders (Meeus and Oerlemans, 2005), regional economies (Gertler, 2010), or sociotechnical regimes (Fuenfschilling and Truffer, 2014). Whether these boundaries are geopolitical, socioeconomic, or sociotechnical, the institutional rules within the system – the cultural norms or laws and regulations (North, 1990) – must be relatively homogeneous. This relative homogeneity can be justified even if we engage in richer institution theory and think about institutions as isomorphic pressures (DiMaggio and Powell, 1983; Irwin et al, 2021). Now, this is not to deny ‘institutional work’ (Lawrence and Suddaby, 2006). For example, Stine Grodal (2017) has shown how the boundaries around the field of nanotechnology were constructed by various communities over time. However, a relatively static view of institutions has proven useful for innovation systems research. Institutions become part of the ‘context’ for a case study. Holding them constant lets us move institutions into the background and interactive learning into the foreground. Furthermore, it lets us compare regional contexts. It is the relative homogeneity within one system that sets it apart from the next. This is the essence of the approach and it is not without value. Innovation systems are zones on a map, full of rules, organizations, and interactions. Their boundaries are the points at which the rules change.

But the deeper we dig into institutional theory, the more we start to see that innovation systems must intersect one another (Freeman, 2002; Castellacci, 2009). It is common to acknowledge that institutional fields are nested and overlapping. And this is abundantly clear in research on any ‘regional, sectoral innovation system’ such as a regional biotech industry (for example, Cooke, 2002) (or a regional ocean tech industry). Some studies, like that by Belussi et al (2010), will fix the boundaries of an innovation system and then explore the openness of those boundaries. Yet, this problem of regional boundary specification has been one of the major unresolved issues in innovation research (Doloreux and Parto, 2005). David Doloreux and Saeed Parto (2005) called it the ‘unit of analysis’ problem. It is the problem of determining whether innovation system boundaries align with a city, metropolitan region, local district, subnational region, and so on. They proposed that proper deployment of institutions – the ‘key variable’ in regional innovation systems (Doloreux and Parto, 2005) – helps resolve this issue. But they also warned that ‘there is a danger of getting “lost in the woods” while searching for the institutional component’ (Doloreux and Parto, 2005, p 146). And this is where the boundaries around innovation systems start to crumble.

Regions matter (Chaminade and Plechero, 2015), but innovation system boundaries can only ever be semi-coherent (Fuenfschilling and Truffer, 2014). We have already seen Jerry Davis’ (2022, p 44) argument that the biggest technology firms today are ‘indifferent’ to sectoral boundaries. Now, consider John Allen’s decade-old, matter-of-fact observation that ‘parts of global cities like New York and London, for instance, predominantly the corporate finance sectors, are seen to be partially detached from the geographically circumscribed authority of the state’ (Allen, 2011b, p 287). Truly, it is hard to imagine any innovation system boundary that is closed off from the outside world (Edquist, 2001). This is because institutions and learning interactions are not necessarily bounded within one geopolitical or sociotechnical space. Innovation processes need not conform with the regions drawn on any map, or with the sectors described in any taxonomy. No amount of institution theory will help innovation researchers resolve this dilemma. One simply cannot have a ‘system’ without an inside and an outside. To study an ‘innovation system’, one must establish an analytical boundary. In short, innovation researchers are preoccupied with finding the edges of their subject matter, but have a sense that these edges are more of an analytical tool than an empirical reality. The limitation is paradigmatic.

Testing regional boundaries

What does this mean in practice? Let’s play within the modernist paradigm a bit longer and consider my Canadian context. Where many European innovation scholars default to ‘national’ boundaries, and must then grapple with multinational European Union policy, there is a different subnational challenge to innovation system boundaries in Canada (Holbrook and Wolfe, 2000). This country is a federation of dramatically different socioconomic regions. Under the Canadian Constitution, the individual Provinces retained regulatory power over nearly all industries, businesses, and professions. If you are not from here, then you cannot navigate this dilemma without a map (refer back to Figure 1) (but maps aren’t everything). In ‘Atlantic’ Canada alone, one must consider whether the three ‘maritime’ provinces are one region or many (Holbrook and Wolfe, 2000) – and whether the province of Newfoundland and Labrador should count at all (it is physically and socioculturally distant). There is also increasing emphasis on major cities as the appropriate contexts for regional innovation system research in Canada (Wolfe, 2014).

I noted this dilemma in the introduction to this chapter. I could have chosen to study ocean science and technology in the City of Halifax or the City of St John’s. I had to choose between one province or many. Some readers might set all this aside and argue that the appropriate boundaries for any study of ocean science and technology should correspond with the topography of the ocean. For example, one might argue for a Northwest Atlantic region. And I could argue that the ocean has material agency. However, I would quickly hear back from some mainstream reviewer that the ocean does not impose institutional rules in the way that provincial or national governments do.

The stop-gap solution in my PhD thesis was to offload the dilemma on others. I chose to study ocean science instrumentalities more or less around the region where I live. Then I devised a way to avoid any debate over where the exact boundaries might be found. Based on media reports, attendance at sector events, and referrals through personal networks, I identified all the individuals based in Halifax whose work responsibilities included understanding and supporting the ocean science and technology sector. I called these five individuals my ‘system experts’. Three of them held industry policy roles – one in a federal government organization and two in different provincial agencies. The other two individuals held ocean technology sector support roles in separate not-for-profit organizations. All five of these ‘system-level key informants’ (Borgatti et al, 2013) agreed to participate in an interview. During these highly structured interviews, the respondents and I logged all their responses on a touchscreen laptop running the network data collection platform EgoWeb 2.0 (Kennedy and McCarty, 2016). This was all performed in a way that would be acceptable within a positivist social science paradigm.

I asked these ‘experts’ to identify organizations involved in using and producing ocean science instrumentalities ‘in this region’. I deliberately avoiding defining what I meant by ‘region’. Once they listed the ‘players’, I was able to infer where each of them thought the playing field might be located.

I did define the sociotechnical field: ocean science instrumentalities. Onscreen and aloud, I explained that my focus included scientific instruments (such as hydrophones that can be used for collecting data on marine life) and research techniques (such as methods for processing data from those hydrophones) but not new marketing or organizational techniques (such as the way hydrophones are packaged for sale, or the way human resources are managed). Even here, I intentionally left vague boundaries around ‘ocean science’. As we will see at the end of this chapter, the field of ‘ocean science’ is no easier to pin down than the socioeconomic ‘region’.

Thankfully, the experts had no problem naming organizations that use and produce ocean science instrumentalities. I engaged them in a free-recall exercise where they could name as many organizations as they wished. They each worked independently to name the relevant organizations, except for one expert who engaged an outside colleague in helping to establish what they felt was a complete list. Following the best practices in network data collection, I used elicitation prompts to ensure that different kinds of organizations were identified (Brewer, 2000). The experts were prompted to include academic, private-sector, government, and not-for-profit organizations. I also anticipated issues in defining ‘organization’. So, I explained the concept of a ‘kind-of-activity unit’:

For the purposes of this study, an organization is not necessarily a standalone legal entity. In many cases, the parent organization (e.g., Saint Mary’s University) is less relevant to this study than a particular department, unit, or division (e.g., the Sobey School of Business). An operating unit can be considered an ‘organization’ if it engages in one kind of activity and has some decision-making autonomy. (OECD, 2005)

This helped sort out some issues with different units of Dalhousie University, the Nova Scotia Community College, and a few multinational enterprises. Notice how this sidestepped many of the ‘organism’ metaphor issues raised in Chapter 5.

By the end, the regional boundary was clear. All five experts named a key organization outside the city of Halifax. Four experts restricted their lists to organizations in Nova Scotia. The fifth expert included organizations located elsewhere in Atlantic Canada, including organizations in New Brunswick and Newfoundland and Labrador. Following the majority opinion of these experts, I decided that the innovation system’s regional boundary must be the Province of Nova Scotia (it didn’t hurt that my gut was already telling me this or that this narrower boundary would save on travel costs).

I also followed the majority opinion in defining the list of ‘players’ within the innovation system. In total, the five experts named 126 organizations: 60 public organizations and 66 private companies. All the academic, government, and not-for-profit organizations that were named by the experts were ‘public’ organizations, based on the definitional criteria set out by Perry and Rainey (1988). All five experts independently named the same 11 organizations. An additional six organizations were named by four experts, and ten more were named by three experts. The first four experts named between 25 and 40 organizations each, while the fifth expert named 92 organizations. The fifth expert named disproportionately more organizations because they defined the regional boundary much wider than the other four experts. All these details are represented in Tables 1 and 2. Again, following the majority opinion, I created a ‘fixed list’ of organizations that comprised the ‘core’ of my innovation system. All those organizations that were named by three or more experts were included in the list.

Table 1:

Levels of agreement among experts
Level of expert agreement Organizations
No agreement (one expert: 20%) 72
Two experts agree (40% agreement) 27
Three experts agree (60%) 10
Four experts agree (80%) 6
All five experts agree (100%) 11
Total organizations named 126
Table 2:

Number of ocean science instrumentality organizations identified by experts
Expert No. of organizations Agreement with other experts
Two or more others One other No others
#1 32 18 6 8
#2 25 19 5 1
#3 28 17 7 4
#4 40 19 15 6
#5 92 23 16 53
Total 126 27 27 72

Note: Totals indicate the number of unique organizations named.

Mapping the system

I identified key informants for all 27 organizations on the fixed list. The key informants were either the head of the organization (such as the President, CEO, or Executive Director) or a vice-head with sufficient knowledge of the organization’s research and development activities (such as a Vice-President or Director). These individuals were invited to participate in an interview and presented with a research consent agreement based on the best practices template developed by Borgatti and Molina (2005) and approved by the research ethics boards at both Saint Mary’s University and Acadia University. With respect to confidentiality, the consent agreement specified:

All data obtained from private sector companies will be kept confidential and will only be reported in an aggregate format (by reporting only combined statistics and by representing all private companies using one common colour/shape on network diagrams). No one other than the primary investigator and supervisor listed above will have access to the data about individual interviewees and the data about private companies. Data about public and not-for-profit organizations will be treated as public-record (i.e., not confidential), except where relationships with private sector companies are noted. To protect the strategic interests of private companies, this data will remain confidential.

The consent agreement included a sample network graph which was used at the outset of each interview to explain the risk that private sector organizations may be identifiable in the research outputs. Given sufficient knowledge of the research context, an informed reader may be able to infer the names of different companies from their relations or positions on a network graph (Borgatti et al, 2013). This consent agreement was signed by all participants.

I conducted face-to-face interviews with key informants from 25 of the 27 organizations on the fixed list. Only two of the 27 organizations on the fixed list did not participate in an interview. After multiple interview requests, senior officials at one private company and one academic PRO did not respond. To maintain confidentiality, these two organizations are unnamed in my work.

As with the preliminary round of expert interviews, the organizational key informant interviews were conducted using a touchscreen computer running the network data collection platform EgoWeb 2.0 (Kennedy and McCarty, 2016). Draft interview questions were reviewed by the five system-level experts and then revised to clarify language, expedite data entry, and better reflect the context of ocean science and technology. The same structured interview questions were used for all types of organizations – private companies, PROs, other public organizations, and not-for-profit organizations. Interviews ranged from approximately 30–90 minutes in length.

First, I asked respondents simple questions about the type of organization they were representing: academic, government research, other government, private company, or not-for-profit; the total number of full-time equivalent employees working at the organization, and the number working in R&D; the kinds of outputs the organization produced over the previous five years and the novelty of those outputs; and whether there had been any changes to the way these outputs were produced over the previous five years, and the novelty of those changes. These questions were the preamble to the more complex task of network data collection.

Most of the interview followed a standard ‘personal-network research design’ (Borgatti et al, 2013). This is a process to produce ego-network data: data on the network of alters, or relations, around each ego, or focal organization. A standard personal-network interview instrument includes two phases of questions: a name generator to establish a list of alters, followed by name interpreter questions to collect data about the alters and about ego’s relationships with them. For the name generator, I presented each respondent with a roster that included the 27 organizations on the fixed list, plus the 20 additional organizations that were named by only two experts. Respondents were asked to review the roster and identify those organizations that their own organization usually interacted with over the past five years. Then, during the name interpreter, the EgoWeb 2.0 software (Kennedy and McCarty, 2016) produced a grid where all the selected organizations appeared as rows and seven different types of interactive learning relationships appeared as columns (see Figure 2). These seven types of interactive learning relationships were adapted from the work of my PhD supervisor, Claudia De Fuentes (see De Fuentes and Dutrenit, 2012). I only needed to expand her model to add the transfer and sharing of equipment and technical services. All those relationships that could have directionality were presented twice on the screen during my interviews. For example, respondents could say that they licensed or transferred intellectual property to an alter organization, and/or that they licensed or transferred intellectual property from an alter organization. This meant that respondents could choose from among the ten different interactions listed in Figure 2 and could select all that applied.

A screenshot of the EgoWeb software with a table of all the selected organizations.
Figure 2:

The multigrid interactive learning relationships component of the interview instrument

Source: Produced by author using EgoWeb 2.0 (2016).

Personal-network research designs sometimes include a third phase of ‘name interrelator’ questions. This is where one might collect data on the connections between alters. However, this introduces considerable interviewee burden because it is time-intensive (Borgatti et al, 2013). Rather than using a name interrelator to collect nearly identical whole networks from each respondent, I used the common approach of aggregating ego-alter data across all respondents (Borgatti et al, 2013). This involved overlaying and merging network data from multiple respondents into one whole network that includes all respondents and the relations among them. From a positivist perspective, a convenient by-product of this data fusion process is triangulation of the network data (Borgatti et al, 2013). It also provides an opportunity to reliably reconstruct minimal levels of missing data (Stork and Richards, 1992; Huisman, 2009). This reconstruction is possible because the responses provided by each respondent overlap with missing values from each nonrespondent (Huisman, 2009). Although two of the 27 organizations on my fixed list did not participate in an interview, all 25 participating organizations provided data on their relations, or lack of relations, with the nonrespondent organizations. These two nodes can therefore be included in the network, as data were collected on 25 of their 26 possible relationships. Only two of the 702 possible relational paths – 0.003 per cent of the observations – were missing from the whole network dataset. Rather than computationally estimating the missing observations, I was able to run all analyses twice – turning the missing path ‘on’ or ‘off’. Since the results did not change, I followed the conservative assumption that these two organizations had no interactive learning relationships.

In the end, I produced a network graph of interactive learning in ocean science instrumentalities around Nova Scotia. I mapped one ‘strongly-connected component’ that comprises 27 organizations (see Figure 3). In other words, no organizations were isolated and all organizations were reachable through paths of interorganizational interactive learning relationships. The network included 12 scientific instrumentality companies and ten PROs. The ten PROs are listed and described in Table 3. The network also included five organizations that were highly engaged in ocean science, but did not directly engage in scientific investigations. One of these is a teaching unit of the Nova Scotia Community College. Four of these are not-for-profit organizations that meet two of the three criteria developed by Perry and Rainey (1988) for being classified as public organizations. I therefore labelled all five of these organizations as public support organizations.

A network graph of interactive learning in ocean science instrumentalities around Nova Scotia.
Figure 3:

Interactive learning network for ocean science instrumentalities in Nova Scotia, Canada

Note: Nodes sized by degree.

Source: Graph produced by author in NetDraw (Borgatti, 2002).
Table 3:

Public research organizations in the interactive learning network
Organization FTEs R&D Intensity (%)
Acadia Tidal Energy Institute 11 98
Bedford Institute of Oceanography (Department of Fisheries and Oceans) 700 21
Bedford Institute of Oceanography (Natural Resources Canada) 55 82
Verschuren Institute, Cape Breton University 40 90
Oceanography Department, Dalhousie University 118 97
Defense Research and Development Canada: Atlantic Research Centre 165 61
Applied Geomatics Research Group, Nova Scotia Community College 20 75
Applied Oceans Research Group, Nova Scotia Community College 10 100
Ocean Tracking Network, Dalhousie University 12 88
Academic Kind-of-Activity Unit (non-participant) – –

Note: FTEs = full-time equivalent employees (a measure of size). All organizations in this table are public, according to the criteria developed by Perry and Rainey (1988): they are all under public ownership, receive public funding, and operate under polyarchal social control. FTEs and R&D intensity for the nonparticipating PRO were available from online sources, but are suppressed in this table to maintain confidentiality.

This brings us to what I considered the most important feature of my data collection process: the equal treatment of public and private organizations. The roster of organizations that was identified by system-level experts included any type of organization that fell within the sociotechnical field. Then, key informants with those organizations all responded to the same structured interview, regardless of any public/private sector distinctions. Because all organizations were treated equally for data collection, I could test some hypotheses and produce some statistics about public innovation in goods. I will return to the data that fell within my survey boundaries and how I analysed it in the next chapter. But first, we should consider the hints of data that did not conform with a regional-network approach. This concluding discussion is not a confessional account of my technical errors (which are likely plentiful) or my poor use of method.1 Rather, I turn to the more fundamental question of how regional and network spatial metaphors focus our attention on some aspects of innovation, but not others.

Topological alternatives

Thus far in this chapter, I have surveyed the boundaries around an innovation system, laying the groundwork for an innovation survey. I defaulted to the regional type of spatial reasoning that underpins the systems of innovation literature. To think about innovation systems, we must first think about space as measurable, divisible, Euclidean fields. We saw that, in the mainline innovation literature, institutional fields are conceived as arenas for interactive learning networks/relationships. Whatever the ‘size’ of the regional space, it is volumetric. It contains and captures certain kinds of innovation, in defined fields of technology, and in particular places. Boundaries can be plotted. In my case, this spatial reasoning enabled a survey of organizational learning interactions, involving the production and use of ocean science instrumentalities, in Nova Scotia, Canada.

The regional/volumetric approach is adopted widely and without question in innovation studies. Any debate is about scale; it is about the size of the regional volume(s) under investigation. And we have seen that the debate about scale is really a search for the ‘institutional component’ of an innovation system (Doloreux and Parto, 2005, p 146). This is where innovation researchers are getting ‘lost in the woods’ (Doloreux and Parto, 2005, p 146). I have suggested that innovation research could use more sophisticated institutional theory tools, and these might provide a less static and bounded perspective. But this will only make it more difficult to trace the ‘edges’ of an innovation system. An alternative, which we began exploring in Chapter 5, would be to change our metaphors and thereby engage in different puzzle-solving activities. And I have already argued, in Chapter 3, that context is not a container. So perhaps innovation research could learn from a more radical set of spatial metaphors? Let us consider new ontological instruments – intellectual tools borrowed from social topology – where regions and networks are only two of the many options for thinking about space.

Topology is the study of how we understand and represent spatial relations. It is conceptually borrowed from mathematics, where methods were devised to account for geometric objects that can be stretched, folded, or otherwise deformed. It is silly to mathematically pin down the corners of a rubber sheet if you are interested in how its surface bends and warps. Topological shifts have therefore proven useful for understanding mathematical problems where the solutions are obscured by a focus on the absolute position and size of objects in space (and time). For example, physicists are making frantic use of alternative topological reasoning in their research on dark matter (for example, Derevianko and Pospelov, 2014; Afach et al, 2021). This was the subject of the 2016 Nobel Prize in Physics. Many believe that dark matter and dark energy might only be observable through topological defects: wrinkles in gravity or time. Less abstractly, topology was useful in the 1700s, when Leonard Euler shifted the famous Königsberg bridges problem from regional to network space. By disregarding the absolute position of the seven bridges around the city, Euler produced a proof showing that it was impossible to plot a route that crosses each bridge only once (Barabási, 2013). This might seem like a trivial maths puzzle, but Euler’s spatial reasoning is now extremely useful in everyday life. Consider the topological map of the London Underground. It shifts attention away from the confusing twists, turns, rises, and dips of the underground tunnels and instead focuses riders on the routes and connections between stations. This is an easier way to understand how the subway system moves beneath the city. It is a map of network space. It doesn’t deny that the trains move through real physical space; rather, it demonstrates that there are other useful ways of understanding our sociomaterial world.

This is why topology has become ‘one of the putative core topics in geography’ (Lata and Minca, 2016, p 439). As John Allen observes, ‘something seems to be happening to the way that we think about space and time – as non-linear, intensive, folded even – that increasingly chimes with our experience of the world’ (Allen, 2011a, p 317). In his work, topology helps reveal the non-Euclidean ways in which nongovernmental organizations (NGOs), governmental organizations, and multinational corporations exert power remotely and over great distances. Beyond questions of geopolitical power, Allen has argued that topology provides poststructuralist geographers with a ‘looser, less rigid approach to space and time that allows for events elsewhere to be folded into the here and now of daily life’ (Allen, 2011b, p 283). This has enlivened thinking about ‘local’ and ‘global’ relations (Lata and Minca, 2016; Latham, 2011). Indeed, Alan Latham has argued that ‘topological notions of space-time are most useful when they are used to challenge the very idea that there is a “global”’ (2011, p 315). In other words, topology can help us problematize the notion of ‘scale’ (Asdal, 2020; Oppenheim, 2020). It helps us tackle ‘the elusive character of borders, scales, territories, regions or networks’ (Lata and Minca, 2016, p 440). This is why the geographer of boundaries Anssi Paasi (2011) has argued for ‘a need to move from “absolute” to relative and relational space or from Euclidean metric spaces to some other spaces’ (Paasi, 2011, p 300, emphasis added). Notice here that the word ‘absolute’ stands in for a modern, realist ontology. Paasi’s (2011) alternatives are relative/postmodern and relational/amodern.

When geographers have imagined these ‘other’ spaces, many have turned to the rhizomatic plateaus of Deleuze and Guattari (1987) and to the material-semiotics of ANT (Lata and Minca, 2016). The ‘topological ethos’ of John Law and Annemarie Mol (2001) has been an especially ‘prolific’ source for new theorizing in geography (Lata and Minca, 2016). Organization studies has also benefited from very similar ‘object lessons’ offered by Law and Singleton (2005). (The same arguments were presented a decade earlier in STS: see Mol and Law, 1994.) In these various contributions (Mol and Law, 1994; Law and Mol, 2001; Law and Singleton, 2005), Law, Mol, and Singleton argue that our default topologies suggest stability and spatial integrity – impeding other ways to make sense of sociomaterial dynamics. Those default topological metaphors are the region and the network. Law, Mol, and Singleton advocate for ‘fluid’ and ‘fire’ as alternate metaphors. A brief discussion of these four topological metaphors will help conclude my argument.

Regions

First, we default to thinking in regions. Research objects appear stable and grounded (pardon the pun) when they are framed by a regional topology. This is the ‘common sense’ view where ‘we tend to think of objects as physically constituted items that occupy a volume in Euclidian space’ (Law and Singleton, 2005, p 335). And yet, ‘regionalism’ is undoubtedly a set of ‘topological rules about areal integrity and change’ (Law, 1999, p 6). The rules of regionalism are socially constructed and contested. Much of this chapter has been a description of the rules that I applied to set boundaries around a regional innovation system. In making the rules explicit, we can see how ‘objects are clustered together and boundaries are drawn around each cluster’ (Mol and Law, 1994, p 643). We can also see past the rules to the underlying assumption: we are working with a surface or volume that must be ‘broken up into principalities of varying sizes’ (Law, 1999, p 6). Of course, this topological assumption extends well beyond regions on a map – it helps us divide all manner of sociomaterial objects.

This topology is especially suited to discussions of national and regional policy, provided that the principality of the state aligns with the principality of the policy phenomena. But here is where innovation research must depart from regional thinking. Innovation is not geographically stable. We might sometimes think about how innovation is anchored – around a large R&D organization (Niosi and Zhegu, 2010) or a key piece of instrumentation like the Aquatron in Halifax, or the geotechnical centrifuge in St John’s that I mentioned at the beginning of this chapter. Innovation can be understood regionally, but we know it is not regionally delimited.

Networks

The network metaphor helps us think without regional limits. I have also approached the idea of a network topology in this chapter, but not in the way that is common to STS. In STS, and then other fields, ANT provided a topological understanding where ‘elements retain their spatial integrity by virtue of their position in a set of links or relations’ (Law, 1999, p 8, emphasis in original). In other words, the integrity of an object is maintained through a pattern of sociomaterial links that remain stable across space and time. Bits and pieces of scientific knowledge – instruments, diagrams, texts, etc. – can carry action across distances if they become punctuated as a ‘black box’ (Latour, 1987). They can be thought of as ‘immutable mobiles’ (Latour, 1987) – objects that transcend physical space without losing their shape. This is the network/rhizome topology – an ontological instrumentality – provided by ANT.

I noticed many immutable mobiles of ocean science during my data collection. At the Applied Oceans Research Group in the Nova Scotia Community College, there was a big round device called a CTD rosette (for conductivity, temperature, and depth), loaded with a variety of sensors and sampling apparatus. One would need at least a pick-up truck to transport this device, which is lowered into the ocean to collect samples and data at various predetermined depths. It was that lab’s turn to physically house the device – but its logoed surface was like the back bumper of some American cars. It was loaded with stickers from ocean science and technology organizations nearby and afar. This was a well-travelled device. I also saw and heard about many devices whose relations stretched beyond Nova Scotia. Some of the science organizations I surveyed had important instrumentation partners outside my research boundaries. Similarly, some of the instrumentation companies I surveyed had their principal scientific partners elsewhere in the world. However, none of this became data. None of these relations found their way into my study and none will appear in Chapter 7 of this book. When I had established my boundary – defined my innovation system – I had imposed a Procrustean transformation on the relations of ocean science technologies. I established the size of the ‘bed’ and then crafted a survey instrument that would deftly, rigorously, but quietly cut off any rhizomatic shoots. I captured network data, yet I did so within a regional/volumetric topology.

The network topology common to STS – the one that is rhizomatic, not regional – is suited to understanding how a scientific instrumentality is taken up and used from one laboratory to the next. Indeed, that network topology has its ‘roots’ in the ANT laboratory studies. It helps us think about the translation of objects through sociomaterial space and time. It was useful earlier in this book when I explored different ways of knowing the past. But notice that this topology is not about plotting relations between organizations or laboratories – as I have done in this chapter and as is common in ‘network’ analyses of innovation. That modernist network approach is about understanding the geometry of relations contained within a region. It is a form of regionalism. An amodern ANT-inspired network topology is ‘not about a volume within a larger Euclidean volume’ (Law, 1999, p 6). In this way the rhizomatic ideals of ANT actually ‘helped destabilize Euclidianism’ in many fields of social science (Law, 1999, p 8).

And so, we know that technology and innovation can be understood differently as network phenomena. There is a tremendous literature on this in STS and I agree with Martin (2016) that innovation studies must catch up with that work. Yet, many STS scholars have moved on. By the late 1990s, it had become clear that the network metaphor ‘had the effect of limiting the conditions of spatial and relational possibility’ (Law, 1999, p 8). More recently, in the journal Organization, Tommy Jensen and Johan Sandström have argued that ‘ANT is still haunted by a conceptual conundrum: while opening up for spatial complexity, the lure of the Network risks drawing representations into a singular, network space, “othering” spaces that might be critical for organizing’ (2020, p 702, emphasis in original). In that same journal 15 years earlier, Law and Singleton (2005) had characterized the ANT network topology as ‘an inappropriately rigid and centred version of relations’ (2005, p 341). This is where they advocated for fluid and fire as alternate topological metaphors.

Fluids

Regions appear stable because we position objects within Euclidean space-time. Networks appear stable because we position objects in relation to one another. But the idea of ‘fluid’ space dispenses with stability altogether.

The ‘fluid’ metaphor was developed in a study of ‘the Zimbabwe Bush Pump’ by Marianne de Laet and Annemarie Mol (2000). Many people speak of this pump as a fixed entity, but de Laet and Mol (2000) describe how instances of the pump are different from one another in interesting and incremental ways. Key to the success of the pump is the way in which the inventor and manufacturer ‘dissolve’ their ‘actorship’ or ‘authorship’ (de Laet and Mol, 2000, p 249). No one controls the various sociomaterialities we might call a bush pump. No one polices the boundaries of how a pump should be installed, used, and repaired. This is why de Laet and Mol say that ‘the Zimbabwe Bush Pump is solid and mechanical and yet … its boundaries are vague and moving, rather than being clear or fixed’ (de Laet and Mol, 2000, p 225, emphasis in original). Law and Mol (2001) later explained that, in fluid space, multiple instances of an object are the same, but not identical. Because one object flows into the next, ‘a fluid world is a world of mixtures. Mixtures that can sometimes be separated. But not always, not necessarily’ (Mol and Law, 1994, p 660, emphasis in original). This means that ‘in fluid spaces there are often, perhaps usually, no clear boundaries’ (Mol and Law, 1994, p 659, emphasis in original). For some readers, this topology might seem unreal. But consider how unreal it is to split and label our planet’s one ocean into sections. Andrea Ballestero (2019) has produced a wonderful book on some of the devices we use to separate water: formulas, indices, lists, and pacts. A fluid topology helps us see how water flows through us and this planet.

In practice, a fluid topology can reveal some of what I missed by surveying stable boundaries. For example, I needed to stabilize ‘ocean science’ so I could use it as an innovation system boundary. But we could see multiple, interrelated understandings of ocean science and ocean engineering from the very beginning of this chapter (and also Chapters 3 and 4). If we were to switch into other languages, the differences would be more self-evident. The Russian term translates to oceanology rather than oceanography and thereby denotes important differences in the scientific practices (Hamblin, 2005; Mills, 2011). Despite those apparent differences, we saw in Chapter 4 that Russian oceanology was not separate from Canadian oceanography – even during the Cold War. In his autobiographical history of oceanography, William Bascom notes that ‘oceanography is not so much a science as a collection of scientists’ (1988, p xiii). Depending on where you sit, ocean science might be focused on physical and/or biological and/or chemical processes. Benson and Rehbock go further, saying, ‘oceanography is a hybrid, a mixed science … [that] cannot be said to be a single scientific discipline’ (1993, p ix). Fluidity is evident in these scientific practices.

At issue is the multiplicity with which humans relate to the ocean. For example, Bascom credits four factors with the rapid growth of ocean science after the Second World War: a ‘doubling’ in submarine warfare, a ‘tripling’ of the global fish catch, the shift to offshore oil production, and a new public interest in marine conservation and archaeology (Bascom, 1988, p xiv). Eric Mills’ history of the field uses a slightly longer list. He writes that due to demand from ‘fisheries, shipping, sewage disposal, ocean mineral exploitation, and submarine warfare, the field [of ocean science] had expanded too rapidly for the supply of personnel from the pure sciences to keep pace’ (Mills, 2011, p 254). Ocean science is enacted in so many ways that disciplinary boundaries are problematic. We saw this in Chapter 4, with the challenges Dalhousie University faced advocating a biological focus against the University of British Columbia’s physical oceanography focus. Eventually, Dalhousie found some value in a ‘fluid’ collection of ocean scientists. But even then, the ocean does not provide a hard disciplinary boundary. Notice that ocean science and technology do not always stop at the shoreline: some ocean science and technology travel into space, and vice versa. If I had been situated in Massachusetts, like Helmreich (2009), I would have encountered the astrobiologists who are mixed into the Woods Hole Oceanographic Institution. The fluid topological metaphor helps us notice this endless mixing.

Like Mol and Law (1994), I must try not to favour the fluid metaphor too much here. They say that ‘fluid spaces are no “better” than regions or networks’ (Mol and Law, 1994, p 663). They simply give us a different understanding – an alternative to the Procrustean transformations of regionalism and relational rigidities of networks. Some say that a fluid topology gives us gradual and incremental change (Law and Mol, 2001). And we sometimes write and talk about innovation in this way. But I favour the way in which Law unpacks the metaphor differently in After Method:

in this way of thinking the world is not a structure, something we can map with our social science charts. We might think of it, instead, as a maelstrom or a tide-rip. Imagine that it is filled with currents, eddies, flows, vortices, unpredictable changes, storms, and with moments of lull and calm. (Law, 2004, p 7)

This tells me that a fluid topology could be useful for thinking about more than incremental innovation. It could also be applied to breakthrough innovation – after all, water can cause tremendous, sometimes destructive change. But what might we make of the destroyed and the absent? Those things are dissolved, or washed away, with the fluid metaphor. And so, the topology of ‘fire’ might also have value.

Fire

The topology of ‘fire’ emphasizes discontinuity (Law and Mol, 2001; Law and Singleton, 2005); it is about instability (vs. region/network) and disconnection (vs. network/fluid). Fire objects are ‘energetic and transformative, and depend on difference – for instance between (absent) fuel or cinders and (present) flame’ (Law and Singleton, 2005, p 344). Thinking with this metaphor keeps our attention on what must be made absent to make something else present. This becomes a valuable tool when we accept that ‘not everything can be brought to presence. Or, to put it differently, to make things present is necessarily also, and at the same time, to make them absent. Presence, in short, depends upon absence (just as absence depends on presence). This is a matter of logic, of definition’ (Law and Singleton, 2005, p 341).

Through a ‘fire’ topology, we might begin to understand the presences and absences that were created during my boundary survey. My attention was on bringing public organizations and scientific instrumentalities into presence. I knew that a focus on the market – as prescribed by the OECD’s innovation survey manual (OECD, 2005) – is a problem because it conceals user innovations (Gault, 2012, 2020). I followed the advice of Fred Gault (2012, 2018, 2020) and wrote a survey that would count any innovation that had been brought ‘into use’ anywhere in society. This was an absence in other research that I made present in my own work.

Of course, we cannot make everything present. Some absences are intentional. For example, I knew that some organizations I surveyed develop secret military technologies that respondents could not legally discuss. But there are also absences that we cannot imagine: the flame is so bright and exciting that it casts a shadow over the ‘other’. Some critical scholars of innovation have already argued that the brightness of the pro-innovation bias shifts attention from moments where ‘no’ or ‘slow’ innovation might be appropriate – like opportunities for degrowth (for example, Cañibano et al, 2017; Leitner, 2017). Also, consider Vinsel and Russell’s (2020) argument that ‘innovation’ deludes us into ignoring the importance of maintenance and repair. It is clear, then, that the idea of ‘innovation’ is focused on the ‘flame’ of novelty and thereby casts shadows over other important socioeconomic phenomena.

After several of my interviews, I noticed that the word ‘innovation’ was creating these kinds of ‘other’ interesting absences from my survey data. For example, I was fascinated by a set of oceanographic instruments that were being constructed in a lab at my university. I saw bits and pieces of everyday construction materials, like one might see in the shop of a plumber or carpenter. In this lab, an interdisciplinary team predicts and studies the ecological impacts of tidal energy turbines. Physical oceanographer Brian Sanderson builds and repairs various ‘low-cost’ drifters for their collaborative studies in the Minas Passage, the Minas Basin, and the Bay of Fundy. In their publications, the research team describes limitations of the more expensive and typical approach: mooring instruments in place (Adams et al, 2019; Sanderson et al, 2021). They are working with the world’s highest recorded tides and the tidal flow produces so much flow noise that standard instrument assemblies produce poor data. Sanderson et al drastically reduced flow noise in their data by allowing their instruments to drift with the current and not be disrupted by waves – figuring out how to control that movement with assemblages of plywood (called drogues) that were also carefully fashioned to avoid entangling lobster fishing gear. For one set of studies, instrument drifters were built from ‘38 mm diameter ABS pipe and common plumbing fittings with flotation fashioned from 50 mm foam board insulation’ (Sanderson et al, 2021, p 58). Inside those pipes were batteries, inexpensive GPS trackers designed and marketed for dog owners, and a Nova Scotia-produced Vemco brand acoustic receiver. This receiver is designed to detect ‘pings’ from Vemco tracking tags which other researchers have surgically implanted into fish, such as Atlantic salmon. The receivers are normally fixed in place (O’Dor et al, 1998), like at the mouth of a river, and are therefore subject to the ‘noise’ of rushing water and jiggling waves. This inexpensive drifter solution, bootstrapped by scientists in their own lab, took measurements differently and thereby proved far superior to moored devices for certain scientific issues. They have done similarly for hydrophone detection of marine mammals (Adams et al, 2019). Their approach to constructing drifters has even allowed local school children to build their own instrument assemblies and collect high quality scientific data (Redden, 2016). In short, these researchers built ingeniously simple and inexpensive devices so that they could produce novel science. Their science and their instruments are uniquely crafted for a particular purpose at a particular place and time. No piece of their equipment is really ‘new to the world’ or eligible for patent protection. I would like to call it ‘innovation’ and yet it would not meet any of the standard OECD survey criteria. So, while their lab appears as a public research organization in my study, there was no space in my innovation survey for the details in this paragraph.

Because I was studying innovation, I could not study non-innovations – the things Godin and Vinck call ‘novation’ (2017b, p 3). And so, there are always absences like ‘novation’ that we overlook because our instrumentalities point us another way. Certain forms of novelty are valued above all else. I will turn to one of these absences in my final chapter. But for now, my point has been made. I am suggesting that there is unexplored opportunity to theorize innovation through a fire metaphor. This is not only because ‘destruction’ and ‘discontinuity’ are common ways of thinking about innovation; it is also because innovation studies should involve some examination of unrealized potentials. Some of today’s absences will become tomorrow’s innovations. So, Law and Singleton do not go far enough when they celebrate fire objects for ‘their novelty, their creativity, their destructiveness’ (2005, p 349). Fire topology also allows us to acknowledge absences – the seemingly non-innovations that are cast into the shadows and the combustive materials that are not yet aflame. This is likely difficult research because fire objects cannot ‘be domesticated’ (Law and Singleton, 2005, p 347).

Other topological metaphors

In this chapter, I have argued for innovation research to break from its singular topological perspective. The field’s most common instrumentalities – collected under the innovation systems approach – rely on regional or volumetric thinking. If you prefer, this has also been called ‘arborescent’ (Deleuze and Guattari, 1987) thinking. Other topological metaphors can stretch, twist, and distort our observations in other fruitful ways. In the words of Iulian Barba Lata and Claudio Minca, topology gives us ‘a spatial lexicon’ (2016, p 441) that can account for the intersecting multiplicity of space-time. Surveying multiple topologies will give us multiple understandings.

The four topological metaphors considered here are not the only possibilities. We need others. Some say these ones have serious limits. For example, Allen (2011b) has argued that metaphors like fluid and fire ‘serve only to confuse rather than enlighten’ (2011b, p 283). He suggests that they are ‘failed metaphors, words which, after their initial promise faded, nobody was much interested in using’ (Allen, 2011a, p 317). He argues that playing with topological metaphors ‘may be colourful, but owes little to the eye-opening possibilities that topology offers’ (Allen, 2011a, p 318). He writes well, but seems to miss the point. Working with multiple metaphors means ‘we can avoid naturalizing a single spatial form, a single topology’ (Law, 1999, p 7). The scholarly task is to expand our ‘spatial imagination’ through ‘metaphorical proliferation’ (Latham, 2011, p 315). Thinking about space and place through multiple metaphors provides an ‘“intertopological” effect … a spatiotemporality not only of the connected and unconnected, but also of the potentially, or not yet, connected’ (Oppenheim, 2020, p 318).

One approach might be through meta-metaphor. ‘Hyperobject’ has recently been proposed as a way to appreciate the many topological possibilities for innovation research (Rehn and Örtenblad, 2023). As a hyperobject, innovation can be understood to be ‘massively distributed in time and space, to the point where most things can be seen as innovation depending on which spaciotemporal position you choose to occupy’ (Rehn and Örtenblad, 2023, pp 6–7). Another approach might be to create space for new understandings through the ‘interruption of topology’ (O’Doherty, 2013, p 211). Damian O’Doherty (2013) did this in an ethnographic methodological experiment where ‘an arbitrary set of rules and constraints following the rigor of a mathematical series of calculations and measurements were devised to generate a sequence of random walks traversing the city of Manchester’ (O’Doherty, 2013, p 215). Based on the resulting insights, O’Doherty would have us proceed without any single metaphor: ‘bereft of any abstract principle’ and therefore open to ‘the conditions of possibility for thinking topology’ (O’Doherty, 2013, p 226).

The alternative to all this topological play is accepting the ‘hegemony’ (Sepp, 2012, p 47) of regionalism. That is the innovation studies norm. But in geography, it is normal to investigate – intertopologically – the construction of boundaries, regions, and territories (for example, Sepp, 2012; Asdal, 2020; Oppenheim, 2020). What if I had done that here? While I was busy pinning down the boundaries of ocean science instrumentality innovation, there was work being done to shift public and government attention from a Nova Scotian ‘ocean technology cluster’ to an Atlantic Canadian ‘ocean supercluster’ (see Doloreux and Frigon, 2021). And as I write this book, there has been a further shift in the governmental language: Canada’s regional ‘innovation superclusters’ have been rebranded as ‘global innovation clusters’ (see Sá, 2022). Note the hyperbole, but also the opportunity to apply unstable topologies.

Some innovation scholars might say that we need the region/volumetric metaphor so we can pin things down, survey them, and quantify them – otherwise they do not count. But topology provides ‘a way of understanding space and time when the numbers no longer quite add up to anything significant’ (Allen, 2011a, p 316). And as we will see in the next chapter, statistically significant results can become meaningless in the face of staunchly held values and beliefs.