Linking science to technology: the “patent paper citation” and the rise of patentometrics in the 1980s

Basic science and the linear model of innovation

The importance of basic science for technological development has been a central theme in the research policy debate since the end of Second World War. An often mentioned landmark is Vannevar Bushs Science, the Endless Frontier: a Report to the President published in 1945 (Bush, 1960, 1945). Stokes (2011, p. 3) provides an in-depth discussion on Bush’s use of the concept “basic research” and highlights two central notions as especially important: (1) “basic research is performed without thought of practical ends”, and (2) it should contribute to “general knowledge and understanding of its nature and its laws”. The main message is that too much focus on practical use thwarts creativity, and the purpose of separating basic from applied science is to protect basic research from outside influence. Indeed when Carpenter et al. (1980), in their first paper on patent to paper citations, refer to basic science as “external” to the patent literature with its connections to technological and commercial interest, they evoke the image of a science untainted by outside interests. Bush’s deliberate use of the concept helped in arguing for a continued support for basic research, and the longevity of the concept could in turn be explained by its integration in economic theories, its use in official statistics of research and development and the lack of viable alternative concepts (Pielke, 2012, p. 357).

Eventually, as the USA became more focussed on the economic progress rather than military needs, the challenges against Bush’s canon became more frequent (Stokes, 2011). The idea of basic research and its role in promoting innovation had come under increased scrutiny during the 1960s and 1970s, and the overarching framework for understanding the relation between science and technology – the so-called “linear model of innovation” – was questioned (Schauz, 2014, p. 313). The focus on growth, the importance of knowledge generation for the economy and the important role of patents were further emphasised in the legal sphere by the introduction of the Bayh–Dole Act in 1980. This law allowed universities to claim ownership of inventions based on federal funding, and it resulted in an increase of university patents from roughly 300 per year in 1980 to 1200 in 1990. The increasing focus on patenting in university settings was reinforced by two high-profile cases in the Supreme Court that allowed for patenting of living organisms and computer software (Rooksby, 2016, p. 132f.) [2]. According to Slaughter and Rhoades (2004, p. 103), these changes marked a shift from a “public good knowledge regime” to a “capitalist regime” in academia. Similar developments have been documented in the United Kingdom where patents have had an increasing impact on the culture and economy of academic science (Sherman, 1994).

During this period, the criticism of the linear model became more persevering as the focus of US research policy changed from military advancement to economic growth (Stokes, 2011). The legal emphasis on patents as property forms further strengthened the focus on universities as engines of the economy. Consequently, those championing the importance of “basic research” needed concrete evidence of its importance for social and technological development. Yet, while being used in official statistics, basic science is inherently hard to define and operationalise empirically. A common solution in many of the early studies of patent citations is to define research as “basic” on the basis of journals. Hence, a certain set of journals are defined as basic, and patents citing these journals are deemed to be relying on basic research (Carpenter et al., 1980; Narin, 1984). For example, in their 1997 study, Narin and colleagues write (p. 322, italics added): “The journals shown on Table 2 are clearly prestigious and influential, and for the biomedical and chemistry papers, quite basic.” However, the journals in physics and technology show another pattern: “The physics journals, however, are not basic, with much of the physics cited in patents published in applied physics journals, rather than the more basic theoretical and high energy physics journals.” (Narin et al., 1997, pp. 323–324, italics added). Hence, the problem of defining basic research is complex, especially as the notion has different interpretations depending on the field. The most common method is to define basic research using classification of journals (for another example, see Van Vianen et al., 1990). However, some authors refrain from separating between different kinds of research, as Coward and Franklin (1989), who instead refer to the “science universe” (papers) and the “technology universe” (patents).

Patent citations as the solution to the objectivity versus relevance problem

As shown earlier, the difficulty in providing systematic empirical data on the influence of science on technology was a major challenge. A central problem for many analysts was to find a method that was both relevant and objective. The idea of basic science driving development and innovation, the so-called “linear model” (Godin, 2017), supposed a direct link between science and technology. However, empirical evidence for the connection between basic or fundamental research and technological development was missing. An important project directed towards the question of how science influences technology was launched by the US Defence department in the 1960s. Project Hindsight, as it was called, investigated the role of basic science in technological development. It studied “100 critical events” in the development of 20 weapon systems and found that only 1 in 10 could be traced back to research and only 1 in 100 to “basic research” (Stokes, 2011, pp. 55–56). These results were troublesome for those claiming a strong link between basic science and technology. Hereafter, as a response, the National Science Foundation launched a project of their own, called TRACES, which looked at the trajectory of five technological innovations – videotape recorders, oral contraceptives, electron microscopes, magnetic ferities and matrix insulation – and, perhaps not surprisingly, strong links to basic science were found in all these cases.

However, a problem, associated with both Hindsight and Traces, was that events or technological developments were picked subjectively. Indeed, it could be claimed that these studies served as well-chosen and illustrative examples, rather than analytical and complete studies. Francis Narin, who himself was a driving force behind the TRACES study, did position these approaches as being highly relevant yet not very objective. In fact, he outlined an “objectivity/relevance” scale which is of particular interest if we are to understand the role that the “patent citation” was given in subsequent work. In a paper titled “Objectivity versus relevance in studies of scientific advance”, he developed a scale on which studies of science could be projected (Narin, 1978, p. 38) (Figure 1).

In reviewing earlier studies, Narin related the challenges of science studies with that of Heisenberg’s uncertainty principle in quantum physics, and his conclusion was that “those science policy studies which are most relevant to measuring the true rate of contribution of a science have the greatest uncertainty as to objectivity, while those which are the most objective have the greatest uncertainty as to relevance.” (Narin, 1978, p. 36). He argued that qualitative studies using “hand-picked” examples of research that were used in technological development often targeted the relevant question: How is research used? Yet, it is doubtful how representative these examples are for science at large. This argument follows the same line of reasoning as expressed by de Solla Price (1965, p. 564) when arguing that the few times science has had direct influence on technological development is when “…the effect is brilliant and startling, and the situation is of considerable historical important so that the incident becomes glamorized and mythologised.” Thus, single examples of how sciences have influenced technology may give an overly optimistic view on the relation between the two, and such studies may, according to de Solla Price (1965), not reflect the normal function of science in relation to technological development.

On the other hand, quantitative data, such as publication or citation counts, can be collected using clearly defined parameters and can therefore be said to be more objective, yet the relevance of such bibliometric measures for studying how research contributes to society can be questioned. Hence, a cut-off between relevance and representativeness (and objectivity) was constantly an issue when analysing the impact of basic research on technological development.

Several of the examples in Figure 1 are from Narin’s and his colleagues’ work in the Computer Horizons Inc (CHI) consultant company. Two methods are discussed as especially promising for the future; citation counts and especially citation networks, and the possibilities offered by the analysis of key words and concepts. Patents are not mentioned, although the analysis of patent statistics, with a particular focus on citations, would be Narin’s main focus during the coming years. Over time, Narin would develop an approach that he found being both highly relevant and representative. The method used the “objective” techniques of bibliometrics and citation analysis, but on a new material: patents.

Indexes and entrepreneurs: the rise of patentometrics in the 1980s

The paper “Linkage between basic research literature and patents” published in the journal Research Management in March 1980 marks the start for a series of papers, published during the 1980s, which uses patent citations as an important link between “basic research” and technology (Coward and Franklin, 1989). Carpenter et al. (1980, p. 34) presented here “a direct technique for linking patent literature, a body of knowledge of technological and commercial interest and external to basic science itself, with the standard measure of scientific research – the scientific article”. While patents, and the references found within them, had been studied before, the analyses produced during the 1980s were larger, more systematic, and they primarily focussed on the link between patents and scientific papers.

The “patent citation” did offer a new and potentially fruitful response to the objectivity/relevance problem, and in arguing for the usefulness of patent citations, Carpenter et al. (1980, p. 34) listed three main advantages of using patents: (1) they were “external to science”, (2) they demonstrate active utilisation and (3) they can be searched without pre-selection of topic areas. Most importantly, the patent citation was “objective”, yet at the same time, the citations provided “direct links” between science (papers) and technology (patents). As Sherman (1994) suggests, the perceived “objectivity” of the patent is largely dependent on its origin in patent law, which allows it to “be presented as a bounded, stable object” (p. 529). Hence, in retrospect it appears as the “patent citation” was the solution to the very problem that Narin outlined in 1978, and during the 1980s, it would be studied extensively by researchers in the emerging field of patentometrics.

The work of Narin and his colleagues at Computer Horizons takes a central role in establishing citation analysis of patents and especially in studying the so-called “non-patent citations”. However, the use of patents as indicators of technological activity was not new, as noted by Pavitt (1985) in an early review of the patent literature. Such analyses were conducted by economist such as Schmookler (1950) and Scherer (1965) already in the 1950s and onwards, and early uses of “patent statistics” can be traced back to the beginning of the 20th century (Tissell, 1907). Already in 1949, Arthur H. Seidel described a “citation system” for patents, and in 1955, the chemist and information scientist Eugene Garfield presented a prototype of a citation index of 5000 chemical patents (Garfield, 1957). Evidently, Narin and his colleagues were not unique in taking an interest in patent citations. Patent citations had been studied for quite some time, yet then often in the context of studies into documentation and information use. For example, Clark's (1976) early study of patent citations involved both patent-to-patent citations and citations to patents from the periodical literature (ISI Journals). In fact, he envisioned how paper–patent citations may be useful for wider studies of science and technology: “patents are used for their technical content independent of their legal one, and we shall propose that citations to patents from serials can index this use, together with that as sources for the history of science and technology.” (Clark, 1976, p. 37). Similarly Charles Oppenheim, information scientist and for a long time working at the patent indexing company Derwent Inc. (Moore and Robinson, 2017), did publish several papers on patent citations in the late 1970s, and many of these studies were inspired by citation studies involving scientific papers (Oppenheim, 1976). For example, Ellis et al. (1978, p. 19) propose that patent citation networks could be linked to journal citation networks and mention that such attempts are “under active consideration”. Thus, methods for tracking citations from patents to papers, and from papers to patents, were developed already in 1980, and the potential of these approaches was recognised. Yet, these early attempts were often small and of an exploratory nature due to the painstaking and time-consuming work of matching references manually.

As outlined earlier, the interest in patent citation analysis during the 1980s was motivated by the search for concrete empirical data on the science–technology interface, which could support the notion of basic research as the driver of technological progress. Yet, a prerequisite for the emergence of patentometrics on a larger scale was the availability of data, and patent citations were available for analysis on a larger scale first in the 1980s. The increasing accessibility of patent citation data was in turn dependent on the establishment of a new industry of scientific information, where entrepreneurs such as Eugene Garfield with Institute of Scientific Information (scientific citations), Monty Hyams’ Derwent (patent database) and Francis Narin’s Computer Horizon (science indicators) saw the need for gathering and organising information on scientific and technological activities. Garfield was first in providing patent citation data, as the 1964 edition of Science Citation Index (SCI) indexed both references in patents and citations to patents from scientific articles. However, due to the cumbersome work of indexing this feature of the SCI, it was abandoned already in 1966 [3]. Derwent had indexed patents since 1951 in various databases, but it was first in 1994 that a “patent citation index” was added to the company portfolio (Oppenheim, 2000). In more general terms, the 1980s marked the emergence of an “online information database industry” (Smith and Tenopir, 2010), and the greater possibilities of distant retrieval of records and data would eventually accelerate the study of patents and papers. Yet, for the period in focus here, researchers still primarily relied on citation data collected by themselves or by the independent research company Computer Horizon (also known as CHI research).

In the early eighties, Narin and his colleagues at Computer Horizon realised the potential of information available in patents and how this data could be linked to other resources such as ISI Science Citation Index. The incorporation of citation data from ISI allowed for more advanced analyses, for example, in studying the relation between patents and highly cited documents. Notable however is that many of the early studies, especially the ones including applicant citations, were rather small due to the time-consuming process of manual or semi-manual extraction of references. Moreover, there were several problems to overcome in linking records from different data sources (e.g. in terms of journal titles and their abbreviations), and especially standardisation names of institutions such as universities and companies were a continuing problem (Narin, 1984). The classification of patents, which in the US system is based on art – Narin gives the example of the class “rotating devices”, which encompass such different entities as a jet engine and a fan – was another difficulty (Narin, 1984, p. 181). Obviously, there was a great degree of manual work involved in creating the CHI patent databases, which in 1984 covered 800000 patents and roughly 5 million references.

So, in the early 1980s, Narin and colleagues in the emerging field of patentometrics began to study patent citations in a more systematic and ambitious way, and the availability of data was a prerequisite for this development. While new databases and methods were an essential precondition for their research endeavour, it was also guided by established ideas and metaphors describing the interaction between science and technology.

Dancers: describing science and technology interaction

In their 1985 paper, “Is technology becoming science?”, Narin and Noma continue to investigate the connection between science and technology through patents. Drawing on de Solla Price, they use the metaphor of science and technology as a pair of dancers, which currently “are locked in an embrace from which it is virtually impossible to separate the partners.” (Narin and Noma, 1985, p. 370) [4]. The proof of this embrace is primarily citations from patents to scientific papers, but similar characteristics between papers and patents in the age distribution of references are important as well. These findings may, according to the authors, have great consequences:

If, for example, it is true that science and technology are converging, or at least converging in key high-tech areas, then this is a powerful demonstration of the utility of basic research to technology. Clearly, if very current scientific research is as important to some high-technology patent areas, as appears the case, the diminution in support for basic research could have negative implications for technology. (Narin and Noma, 1985, p. 380).

For Narin and Noma, this observation has methodological implications for how research should be studied, and the strong connection and integration make it necessary to include patents in studies of science. Yet, despite science and technology being highly connected, it is still basic research that leads “the dance” and technology that follows. As noted by Meyer (2000, p. 411) in his review on patent citations, to determine who actually leads the dance becomes increasingly difficult as science and technology draw closer to each other and eventually become indistinguishable from each other. Consequently, Meyer criticises Narin and colleagues for relying on the traditional “linear model of innovation” when describing the relation between science and technology.

The metaphor of science and technology as “dancers” is used by Narin several times in his writings (Narin and Noma, 1985; Narin and Olivastro, 1992), and unmistakably he picked this references from de Solla Price and his essay on “Is technology historically independent of Science: A study in statistical historiography” (1965). It seems obvious that he found the metaphor attractive and illustrative, but at the same time Narin’s interpretation is markedly different from de Solla Price’s. Throughout the essay de Solla Price questions the tight connection between science and technology, viewing them as two distinct spheres which develop at the same rate, but where there is little connection between them in the form of literature exchanges (de Solla Price, 1965, p. 563). In particular, de Solla Price emphasises the difference between science and technology in their use of literature where he positions scientists as “writers” and technicians as “readers”. In the conclusion, he summarises his view while emphasising the importance of more detailed studies:

I am afraid that the naïve image of technology as “applied science” will be difficult to refine and understand in greater depth. But until we know what the rhythm is and how both dancers move to it we shall not have a proper understanding of the history of technology, and until we know that we shall not be able to make intelligent judgments in such critical areas as the support of science and technology and medicine by state and industry. (de Solla Price, 1965, p. 567)

An important argument for de Solla Prices is the cumulativeness of literature and how the age span of citations can be seen as an indicative of science. This would indeed be what he refers to as the “rhythm” of science and technology. According to de Solla Price, a spectrum, from pure science to non-science, can be deduced based on the age of references used, or in his words: “the ratio of the research-front citations to archival citation.” (de Solla Price, 1965, p. 558). Price suggests that references in patents do not form a “cumulative network” which in turn would place them on the “non-scientific” end of the spectrum. Clearly, this is an assumption that Narin and several contemporaries challenge, and the age of cited references in patents therefore becomes an important argument when linking science and technology through patent citations.

The metaphors of the “dancers” and de Solla Price’s views on the relation between science and technology have obviously been influential in forming the research agenda in early studies of patent citations. Still, there is an inherent contradiction in the descriptions of “dances” and “interactions” and their subsequent operationalisation of the relation between science and technology. Meyer (2000, p. 410) argues that such a view builds on a simplified notion of “quasi-organizational” definitions of science and technology, in which university-affiliated academics produce science and industrial researchers produce technology. Studies of science and technology interaction tend to emphasise the increasing closeness of the two, making them almost indistinguishable, yet in designing a method for analysing the actual documents a distinct delineation is needed where papers represent science and patents represent technology and application. When the delineation is made, connections can be drawn. The main method for making this connection was, and still largely is, to study references in patents and papers. The idea was that references in patents, much like their scientific equivalent, could be studied as indicative of knowledge flows. Yet, referencing practices in patent documents did offer specific challenges both in terms of methodology and in their theoretical interpretation.

Linking through references: understanding the patent paper citation

A challenge recognised already early in the literature on patent citations is how to handle the two different kinds of references included in patent documents. In contrast to the scientific paper, two types of references exist in patent applications: those inserted by the applicant and those inserted by the examiner. Examiner references are often found on the front page, or on a separate data sheet, while references inserted by the applicant are found in the body of the text. As pointed out by Oppenheim (2000), the practice of examiner-added references is of a rather recent date starting in the beginning of the 20th century in Britain and as recently as 1947 in the USA. Generally, the references inserted by the examiner are easier to access for analysis as these are indexed in citation databases to a larger degree than the applicant citations. This is largely due to examiner citations being given in a specific section – usually on the front page of the patents, while applicant references are scattered throughout the document, which makes it more complicated to index them. This is why many studies of patent citations mainly focus on examiner references.

In the US systems, the applicant is obliged to cite relevant prior art, yet but as pointed out by Oppenheim (2000, p. 408), the applicant “…naturally wishes the invention to appear in the strongest possible light” and while such motivations may be rather similar to the motives of giving references in a journal article, it could be that applicants have an even stronger incentive to focus on the uniqueness and importance of the invention. Collins and Wyatt have a similar view when describing references from applicants as “…either related to but significantly different from, or else a useful step towards, the new invention or a use of the invention.” (Collins and Wyatt, 1988, p. 66).

Regarding the references added by examiners, Collin and Wyatt (1988 p. 67) suggest that they focus their reading and subsequent citing to a more limited section of the literature compared to applicants (e.g. they more often cite other patents), and when scientific publications are used, it is more frequently in the form of abstracting journals rather than in its primary form. Moreover, they claim that applicants are prone to cite academic journals, while examiners relay on a more extended set of publications including primarily patents but also reference books and abstracting journals (Collins and Wyatt, 1988). Furthermore, examiners may be more inclined to choose references that satisfy legal requirements, as they should focus on the claims made and are not obliged to cite additional sources (Oppenheim, 2000, p. 408). In more general terms, referencing practices of applicants resemble those of researchers, while examiners have a narrower focus on the legal claims. This relates to the community addressed where “[t]he author assumes a very substantial familiarity with the subject matter of the article, on the part of the intended audience of a journal article. The inventor only assumes the ability to understand the specific application for which patent protection is sought.” (Meyer (2000, p. 100).

A majority of the early papers on patent citations point to applicant references being a more important, or at least as important, link between science and technology compared to references given by the examiner (Carpenter et al., 1980; Narin and Noma, 1985). The importance of including applicant references is emphasised also in studies using references inserted by the examiner, and the need for more in-depth studies is stressed. For example, Vianen et al. (1990) find that there are considerably more applicant references to scientific papers compared to examiner given references, and they “…conclude that the analysis of inventor-given references is a necessary further step in the study of the science base.” (p. 81). Hence, an additional argument for including applicant references is that they would provide a more robust statistical material for analysis. The low ratio of non-patent citation is viewed as methodical problem, and there are concerns that “random samples” would distort the findings (Schmoch, 1993).

Overall, a gradual shift in the use and discussion of the two kinds of references found in patents can be identified. Initially their similarities are emphasised, and both types are analysed to varying degrees. Several of these pioneering studies discuss citer motivations and rely upon information from applicant and examiners involved in the process. Later studies, in which examiner citations are highlighted as the most reliable, foremost studies patent citations on a larger scale using solely quantitative approaches. For example, in the influential 1997 paper by Narin and colleagues, the decision to focus only on examiner citations is justified both methodologically and theoretically: “the front page references should be the most important ones on a patent, since they are the ones relied upon, as mentioned previously, by the examiner in establishing the patent’s novelty. Furthermore, from a practical viewpoint it is far more difficult to extract the non-patent references scattered through the text of a patent.” (Narin et al., 1997, p. 319). Gradually the differences between the two types of references are strengthened, and eventually it is argued that examiner citations are to be preferred both theoretically and practically. Again, the perceived “objectivity” of examiner citations rests on their origin in patent law, which posits the examiner as a more neutral and objective actor compared to the applicant. Evidently, this development coincides with larger studies, which do not allow for manual indexing of applicant references.

Fences or stacks: referencing practices in patents and papers

When the function of references in patents is discussed, the analysis often focusses on differences between applicant and examiner citations. Already in 1988, Collins and Wyatt suggest that very little has been published “on the cognitive and sociological function of citations in patents” (Collins and Wyatt, 1988, p. 66), and similar statements can be found in the subsequent literature (Oppenheim, 2000; Meyer, 2000). A few exceptions can be found however, and a more social and rhetorical perspective on the role of references in patents is suggested by Rip (1986). He argues that the main task of a patent application is “that of marshalling a set of forces that will withstand sceptical scrutiny” (p. 85). The applicant thus needs to erect “fences of interest”, and one of the resources used for this purpose are scientific claims (Rip, 1986). Hence, references to scientific papers are used for a very specific purpose compared to the many “reasons to cite” identified in scientific texts. Rip’s account of the use of references in patents highlights the social and rhetorical aspect of citations, and the analogy of “fences” points to the work that references do in the actual document. Importantly Rip’s account, contrary to many studies originating in the scientometric literature, emphasises the legal character of patents.

A similar argument, but in the context of scientific texts, is made by Latour (1987) when arguing for the role of references in persuading an audience. He describes the practices of citing as “stacking” where references are used as “evidence” for a certain view point, and by lining up (famous) names, a certain claim is made more credible in the eyes of other researchers. In both cases references are used as a defence against the questioning of a scientific or legal claim, yet “stacking” and “fencing” are two distinct strategies, which imply different citation practices. Stacking is mostly a matter of adding, the higher the stack, the stronger the claim. Consequently, plentiful of references should normally work to one’s advantage, as many “Allies” make it more difficult to question a specific claim. “Fencing” on the other hand requires exactness in the claims made and the borders drawn. On the one hand, the patent should include “prior art” yet too many references may compromise the “novelty” of the claim. Accordingly, giving references in patents becomes a balancing act, in which each reference is carefully selected. For example, it has been stated that “[…] because citations in patents have a specific legal function, they are likely to be much more carefully selected than citations in journal papers. The former will be scrutinized by patent examiners; the latter may not attract similar attention from referees.” (Collins and Wyatt, 1988, pp. 66–67). This is an interesting observation as it points to an important difference between paper and patent: adding to the “stack” of references in a scientific paper comes at a low cost, and with little risk, while adding “fences” may come at considerable cost, and ultimately claims made in relation to the literature could be challenged in court (for a recent discussion, see Donato et al., 2019). Moreover, this entails that references in patents can be added at various points – new “fences” can be erected – over the lifetime of a document, resulting in a complicating factor for citation analysis (Warr and Suhr, 1992).

It could be argued that comparisons with “scientific” references offer little in terms of understanding the referencing practices of patents. While this might be true, it does emphasise the problem of assuming that all citations are alike. Similarly Oppenheim (2000, p. 421) claims, on the basis of a study by Kaback et al. (1994), that

… patents are not governed by the same rules of etiquette as journal articles. The references that are made by the applicant rarely look like the bibliography of a journal article. The authors of the patent application wish to avoid any implication that the current application grew naturally out of earlier work. Thus, most prior art that is cited by the applicants relates unsuccessful approaches to the question.

In regard to references given by the examiner, Oppenheim (2000) emphasises that these are exclusively focussed on the claim and not used for background information. Hence, references in patents are used for distinctively different purposes compared to their function in academic papers. Moreover, while referencing practices in science are rather uniform across national contexts, the giving of references in patents is dependent on the specific patent legislation of a particular country or region. Meyer (2000) emphasises that the requirements of patent applications in particular legal systems result in large variances in the number of references given (more references in the USA compared to European countries) as well as differences in the kind of materials cited (more international references in Sweden compared to e.g. Germany). Another important difference is the social function of citations. The patent reference is, for example, rarely used for making friends and Allies, while an important function of the reference in scientific publications is to show membership and identity. The patent’s references are, as suggested by Rip (1986), more about drawing boundaries and showing independence. As, suggested by the “stack” metaphor, scientific references have emerged as a currency on the academic market, the problem of strategic referencing, for example, by citing key figures in the field, or for that matter, possible reviewers, has become a debated issue. Such considerations are less pronounced in the context of patent citations as examiner and applicants have little to gain from engaging in “citation gaming” or flattery citations.

To summarise, the specific purposes of scientific paper and patents, as well as the spheres (scholarly, legal and commercial), which they address, influence why and how references are given. Largely referencing practices have been “black boxed” in research on patent citations, and few attempts have been made to theoretically explain why references are made by applicants and examiners. A fully fledged theoretical framework of the patent citation is probably not feasible given the many possible layers and perspectives (social, legal, historical and rhetorical) involved. Still, informed considerations regarding the specificities of referencing in patents may result in a more nuanced and contextualised understanding of how the different practices, and “rhythms”, influence empirical findings.