Fences or stacks: theoretical considerations regarding references in patents

Introduction

Interest in patent citations was heightened in the early 1980s because of their usefulness for studying linkages between science and technology (Hammarfelt 2021). Citations from patents to scientific articles were introduced as method for showing how scientific research was supporting technological advancements, and linkages between patents and papers became an important argument for the continuing funding of basic science (Carpenter, et al., 1980). Simultaneously, the improvement of databases, such as the one developed by Francis Narin and colleagues at Computer Horizon, gradually allowed for more comprehensive analyses of patent citations. Eventually, citation analysis of patents, later also called patentometrics, became an established method for studying relations between science, technology and innovation across a range of fields, including for example, bibliometrics, economics, innovation studies and research policy. However, while patent citations have become established items for analysis, or even indicators in their own right, there has been little work done on developing a conceptual and theoretical understanding of citations and referencing practices in patents.

There are plenty of suggestions for a theory of scientific citations, but very little has so far been achieved in terms of a citation theory of patents. A few attempts exist, Collins and Wyatt (1987), Meyer (2000a and 2000b) and a report by Jaffe, Trajtenberg and Fogarty (2000) appears as some of the more ambitious empirical studies, while, Oppenheim (2000) offers a more conceptual review of the topic. These studies foremost operate on the level of taxonomies for reasons to cite, and generally follow closely to previous attempts of classifying scientific citations. Collins and Wyatt (1988) outline the legal requirements of patents, and how these may influence citation patterns, and they contend, “little has been published on the cognitive and sociological functions of citations in patents” (1988, p. 66). Meyer (2000b) suggests that the assumptions made by authors and inventors are different. The scholarly author assumes a very substantial familiarity with the subject matter of the article, on the part of the intended audience. The patent applicant only assumes the ability to understand the specific application for which patent protection is sought. In contrast, Jaffe, Trajtenberg and Fogarty (2000) suggest two main functions of references in patents: showing similarity of application and pointing to similarity of technology. These functions are quite narrow compared to the many reasons to cite which has been identified in the scientific literature. For example, a recent study of citer motivations listed no less than 35 different reasons for giving a reference in scientific papers (Lyu, et. al. 2021). The analysis presented here draws upon these earlier studies, and previous findings on reasons to cite constitutes an important background to the argument made. The ambition is, however, to advance beyond these accounts by adding an additional layer of analysis in which the overarching rhetorical function of references in patents and papers is considered. Yet, before a theory of patent citations can be discussed in detail there is a need to shortly recapitulate some key features in the discussion on the possibilities of a citation theory.

Theories of referencing, or theories of citation: revisiting the citation theory debate

Is there a need for a theory of citations? This was the question discussed by a range of renowned scholars in a special issue of Scientometrics published in 1998 (Leydesdorff 1998). The question asked in this collection of papers was not new as there had been a long history of trying to conceptually understand citations and their role in scholarly communication. Partly these discussions regarding a theory of citations could be seen as a response to a critique from neighboring fields such as science and technology studies, which questioned citation analysis based on its lack of theoretical underpinnings. An influential contribution to this discussion was an article written by MacRoberts and MacRoberts (1989) in which several criticisms against citation analysis and its use in research and policy were raised. A key problem according to these authors was the lack of theory for explaining citations.

Generally, two main viewpoints, the normative and the constructionist, can be outlined in the debate regarding a theory of citations (Nicolaisen 2007). The normative, which partly builds on the Mertonian norms, basically suggest that references are made to relevant work that has contributed to the citing study, and the citing of a source can be interpreted as a form of endorsement of previous works. The so-called normative theory of scientific citations is generally associated with the idea of scientific norms (Merton 1973/1942), which guide scientific practices and conventions. In the constructionist tradition, on the other hand, citing is a matter of persuasion (Gilbert 1977), and referencing is viewed as social act that is only partly dependent on epistemological matters. To a large degree the patent citation literature tends to align with the former, normative, position. For example, Tijssen and colleagues based their understanding of patent citations “…on the premises that these citations are primarily meant as “positive” citations reflecting significant contributions of scientific research to essential elements in the patented invention”. (Tijssen, et al., 2000, p. 394). A similar, argumentation is found in a paper by Li and colleagues (2014, p. 1009) where it is claimed that “…the ultimate goal for applicants in a patent application is to assert uniqueness/betterness”. Such views align well with the normative view of references in scientific papers, where the primary reason for citing is to give credit to important earlier work.

A response to the critique against citation analysis, which tries to avoid both normative and constructivist explanations, is to argue that many of the problems raised concerns references rather than citations. Hence, a conceptual separation between citations (received citations that are registered in a citation index) and referencing (giving references in written text) was suggested (Wouters, 1999). The attractiveness of such a division is obvious as it separates referencing, which is a complex and messy activity involving both epistemological and social deliberations from citation analysis. Citation analysis, according to van Raan (1998), is then a purely quantitative approach where large numbers solves problem associated with individual preferences and practices. Hence, in the debate about developing a theoretical understanding of scientific citations the need for theory has been dismissed by the argument that explaining why references are given cannot be used to explain citation distributions on a larger scale.

However, while such a view may be relevant when dismissing studies of individual referencing behavior it largely disregards that practices of referencing which are being manifested on a group level rather than on individual basis. An illustration of this is how historians tend to use references in specific ways, and for particular purposes and these are different from how biomedical researchers, or physicist, use references in their text (Hellqvist 2010). Consequently, while individual behavior might be less relevant for theory development, the group, or practice level, is indeed important for understanding citation distributions on the macro-level. Furthermore, the normative argument might be less convincing in the field of patent citations as datasets generally are much smaller than their bibliometric equivalents, and this especially holds true for early studies operating with rather small samples in specific specialties. It should also be mentioned that differences between national systems in their requirements for patent applications may have considerable influence on how many, and which type, of references that are made. The following section will further outline six important differences that are important when considering how references in patents differ from those found in the scholarly literature.

Referencing in patents and papers: six distinctions

Patent and scientific papers share many similarities as formalised documents containing scientific and technological information. Yet, the referencing practices of scholarly authors and patent applicants are different, which have consequences for how references are given, and some notable features distinguishing the two genres are outlined below.

Who cites? Or the continuing debate on applicant versus examiner references

In scholarly articles the author(s) generally add references, although suggestions made by colleagues, editors, or reviewers may influence their decisions. In contrast to the scientific paper two types of references exists in patent applications: those inserted by the applicant (or sometimes referred to as the inventor), and those inserted by the examiner. The references added by the examiner are often found on the front page, or on a separate data sheet, while references given by the applicant is found in the body of the text (Meyer, 2000b, p. 98). Generally the references inserted by the examiner are easier to access, and these are indexed in citation databases to a larger degree than applicant citations. For certain purposes, for example when looking at the interaction between patents and the scientific literature it has been claimed that “…inventor [applicant] citations rather than the total set of citations, that should be taken as indicators of knowledge flows”. (Criscuolo and Verspagen, 2008, p. 1907). Indeed Meyers (2000b, p. 98) argue that the connection to scientific findings and scholarly articles is more visible in the references added by the applicant compared to the references added by the examiner. However, subsequent work has questioned this conclusion and pointed to examiner references as the most accurate measure of science-technology interaction. Recent studies based on text similarity Chen (2017) suggests that examiner citations better describe the relation to prior art compared to the more scattered references inserted by applicants. The overall conclusion is that “…examiner citations are a good indicator of knowledge linkage rather than an incomplete and noisy indicator”. (Chen, 2017, p. 76). Hence, a contested issue has been how the two types of references found in patents – inserted by either examiners or applicants – should be handled and understood. Which type of reference reflects science and technology interaction most accurately? Many early studies consider references by applicants to be a more promising method, as applicants generally cite the scientific literature more frequently. The arguments used in this debate has however changed over time, and more recent studies has focused primarily on examiner citations (Hammarfelt 2021), yet if this is the due to them being a truly better indicator, or if its due to examiner citations being more easily accessed and indexed remains an open issue.

The temporality of citing and the stability of reference lists

An important difference between patents and scientific papers is the temporality of referencing. Generally most references in journal articles are given before submission to a journal, and thereafter may additional references be added in the revision of the paper. However, when the articles are published the reference list becomes permanent, and new references will not be added to the text. In patent application references may be added or withdrawn at different stages of the process, and references themselves can be questioned, or even be part of legal deliberations and negotiations. Hence, the notion of a fixed and stable list of references is less established in patents, which makes citation patterns more dynamic and interchangeable. That references can be added or removed during the lifetime of a patent has also been identified as a possible problem when analysing patent citations (Warr and Suhr, 1992).

Global or local references – the influence of legal systems on citation patterns

That referencing practices are shaped by disciplinary differences is established knowledge in the field of bibliometrics (Larivière, et al. 2006). However, there are other contextual factors that shape how references are given. For example it is often so that authors are more prone to cite scholars close to themselves such as colleagues at the same department, university or from the same country. However, the norm in scholarly papers is still to cite globally – the most relevant papers should be cited, not the most familiar ones in terms of geographical proximity. For patents, however, the situation is rather different as they predominately act within national legal contexts. Hence, several studies have shown that characteristics of patent citations are influenced by the legal system in which they appear. For example, Meyer (2000b) emphasises the legal character of patents, which influences how references are given, and which also results in large national differences in the number of references given (more references in the U.S compared to European countries) as well as differences in what and who is cited (more international references in Sweden compared to what is common in the German legal system).

Transparency of referencing practices

How and why academic authors cite previous literature has been studied quite extensively, and several attempts have been made to list various motivations and practices (Erikson and Erlandson 2014; Kunnath, et al., 2021). Moreover, referencing is part of the everyday practices of researchers, and practical knowledge about norms and conventions of citing is part of academic literacy. The interest in the social and political aspects of referencing has been heightened in a time when citations are used as important indicators in evaluation, and serves as important merits for individual researchers. Hence, referencing in academic texts is increasingly an ethical issue, which is discussed in terms of responsible referencing (Penders 2018), and the political act of citing specific authors is highlighted. Consequently, inclusive practices, such as citing underprivileged groups like ethnic minorities or research from countries that generally are less visible in the scientific community, has been proposed.

For patents, the norms and deliberations associated with referencing have not been studied to the same extent. Partly this might be due to referencing practices of patents being less visible and accessible. Patent applicants, and patent examiners, may be less inclined to take part in surveys regarding their work, and the openness and transparency, which is supposed to be part of the ethos of science, is not always present in the legal sphere. The few studies made of how examiner use references (Meyer 2000b) usually build on a few selected informants, and more systematic studies are rare or non- existent. Similarly, there is little knowledge – except for general assumptions – regarding differences in how examiners and applicants cite.

Incentives for gaming

References can certainly be used strategically in patent applications, yet there are little incentives for the type of strategic referencing discussed in scholarly publishing. Citing yourself, your colleagues, or the journal in which a paper is published in order to boost citation scores is not unheard of in scientific publishing, and so-called coercive citations where reviewer or editor demands that certain sources should be cited has been documented in the literature (Wilhite and Fong 2012). Although patents may be studied using citation analysis there is little incentive to cite sources with the purpose of increasing ones own, or a colleagues, citation scores, as the patent or the applicant, rarely is evaluated, or valued, based on the number of citations received.

Quantity of references

Generally, there are much less references given in a patent compared to a scientific paper. While the number of references in patents, including references to the scientific literature, has increased it is still low in comparison to scholarly papers. Indeed, it has been argued that in some cases the numbers are so low that analyses of patent citations may in some cases be statistically unreliable (Vinkler 1994; Oppenheim 2000). The possibility to access patent citations through online databases has gradually increased the accessibility of larger data sets, yet for specific studies of certain sectors, or time periods, the generally low number of references – especially references made from patents to the scientific literature – may still be a problem.

The differences identified and discussed above are summarised in table 1 below. Note that this list is not exhaustive but focuses on aspects, which have been found to be of most important for the comparative approach of the paper.

Fencing or stacking: towards a theory of referencing practices in patents and papers

Identifying differences may be useful when attempting to formulate a possible theory of referencing, but solely pointing at dissimilarities is not enough. Rather a theory should, at least partly, be able to explain why these differences exist, rather than just registering them. A basic premise in this paper is that a theory should consider that references are always made within texts, and while they may serve intellectual as well as social functions, their purpose is always also rhetorical. This means that references will be used differently depending on the genre of a text, and its purpose. Hence, in order to understand referencing practices, and their effects on subsequent citation patterns, we need to consider genre specific uses of references as rhetorical devices. So then, what kind of rhetoric is used in patent applications? According to Rip (1986, p. 91) “Patents are still arrays of forces that are designed to oblige the reader to follow certain itineraries. The itineraries are traced out more explicitly because they are legal documents and not simply texts emanating from a laboratory”. Patents both point to a specific knowledge area, which are of interest for exploitation, and at the same time patents tries to fence of competing claims. This dual nature of patents, as being documents that both open and enclose information (Wirtén 2019), has consequences for referencing practices. In fact it has been argued that patents seldom cite the most relevant and related scientific papers, but rather those that are relevant more generally. Or applicants may, as suggested by Oppenheim (2000) be inclined to cite earlier documents which are most useful for displaying the invention as being novel and important, and not the sources which are most relevant. This would be a referencing strategy, which is open, showing related yet not always the most relevant sources, yet at the same time disclosing the most important and novel aspects. Similarly, the references by examiners are usually given in order to delimit the extent of the claims made by the applicant. Consequently, the main task of a patent application is not transparency but “that of marshaling a set of forces that will withstand skeptical scrutiny” (Rip 1986, p. 85). The applicant thus needs to erect fences of interest, and one of the resources used for this purpose is scientific claims made in scholarly publications. Hence, references to scientific papers are here used for a very specific purpose compared to the many reasons to cite identified in scholarly texts.

The rhetoric of scientific papers is a vast subject, and similarly the use of references, and reasons to cite has been covered in numerous papers and books (Baerman, 1988; Bornmann and Daniel, 2008). However, specific reasons to cite – like references made for the purpose of clarification, support, or contrast – are less relevant for the analytical comparison between patents and papers which is sought after here. Rather, for this purpose a principal idea of the role and use of references in scientific papers is needed. Such a conceptualisation is provided by Latour (1987) in his description of the role of literature in science. Important in his analysis is the “question of numbers” and the “stacking of masses of references” (Latour 1987, p. 33). The more allies (references) a paper has, the harder it will be to criticise the claims made. A higher stack of cited papers will thus strengthen the argumentation and rhetorical power of a paper. The skilled academic author will adapt the given references so that they actively support the argument made in the written text, or as Latour puts it “replacing crowds of uncertain allies by well-arranged sets of obedient supporters” (Latour 1987, p. 35). The arrangement of references is both a craft and a strategy – Latour compares it to the building of a stone hut – were each layer of cited texts builds on previous literature, and each added references is supported by earlier ones. Consequently, the strategy of stacking consists not only of the piling of references, but these must also be carefully arranged as one misplaced, or flawed references, may result in a collapse of the rhetorical strategy.

Importantly, the metaphors of fencing and stacking serves to illustrate the overarching purpose of references in the distinct legal and scientific context in which these two genres operate, but stacking and fencing should not be understood as two neatly separated rhetorical strategies. Rather the stacking of references may be a tactic used when defending against possible opposing claims in a patent application, and similarly fencing could be employed to ward of criticism of arguments made in a scientific article. To avoid simplification it should therefore be stated that in practice these two strategies often intermingle, and the type of conceptualisation provided here builds on a comparative approach, which focuses in differences rather than similarities.

Conclusions

Despite their joint history, our current understanding of scientific citations, and attempts of formulating a theory of citations largely refrains from engaging with the legal history of referencing. In framing citation practices in patents as distinctively different from those found in science this paper contributes with a novel approach for understanding the function of references. Six important characteristics – who cites, the temporality of citing, national context, transparency, incentives, and quantity – in which patent citations differ from their counterparts in the scientific literature are outlined. All of these aspects have been mentioned in the literature before, albeit some of them very briefly, but they have not been assembled and analysed in a systematic and integrated effort. In fact, insights regarding referencing in patents and its implications for citation analysis is scattered across the literature. By bringing key observations and theoretical contributions together this paper contributes to a further integration of concepts and approaches.

The comparative approach of juxtaposing patents and scientific papers in terms of referencing practices is warranted due to their close association in quantitative studies of citations in the fields of bibliometrics and patentometrics. Still, the ambition has not only been to provide a simple dichotomy, but to provide an explanation to why these differences appear. Hence, by focusing on the rhetorical function of references in relation to constraints and possibilities offered by the genre of papers and patents are two specific strategies identified. The concept of fencing (Rip 1986) is used to illustrate how references in patents are primarily used to demarcate claims, while stacking (Latour 1987) of references is an act of cumulative persuasion. These strategies can then be discussed in relation to the identified characteristics of references in patents and papers. The demands of legal demarcation is one explanation to why patents generally contain much fewer references compared to scientific articles, and reversely the continually increasing number of references in scientific papers may be explained by the rhetorical strategy of stacking. Similarly, adding additional references, also called citation gaming, for the purpose of boosting the citation scores of others or yourself would not be a motivation for applicants or examiners, while academic authors may see a benefit both socially and strategically. Especially as the potential cost of adding an extra reference is generally low in scholarly papers. Moreover, while stacking references is a visible practice in science which has been discussed in terms of perfunctory citations for some time, the practices of adding references, and demarcating borders, in patent applications still is secluded activity where few, except the professional experts involved, have real insight.

Probably, an all-encompassing theoretical framework, which covers intellectual, social and rhetorical aspects of referencing, is not achievable. Still, the theoretical discussion initiated in this study will hopefully encourage further reflections on the advancement of a theory of citation which also includes the genre of patents. Important, for the argument made here is the realisation that citations when used as indicators of technological development, or scientific excellence, are derived from references which are imbedded in a (con)text. Detailed consideration of the text from which citations are derived is seldom achievable, or even desirable, in large-scale bibliometric analysis. However, a pre- understanding of the rhetorical role that references have in a specific genre –in an essay by a humanist scholar, a scientific paper or a patent – could contribute to more substantiated conclusions based on citation statistics. Given that patents for many researchers – also those in research policy and bibliometrics – is a less familiar genre than scientific papers the need of understanding the conventions and practices of referencing within the patent literature becomes even more insistent. To take a concrete example, an assumption made in the literature is that highly cited patents might be more valuable that less cited ones (Sampat and Ziedonis 2004). This idea echoes the notion that citation scores reflects the importance (and quality) of scientific papers. Yet, if we consider this assumption in relation to the idea of patent references as mainly an act of fencing then being cited takes on another meaning, as strategic positioning becomes more important that acknowledging important prior work. In such way the metaphors of stacking and fencing could be useful as a starting point for further analytical deliberations on genre specific referencing practices and how these may influence distributions and patterns in citation analysis.

Acknowledgements

This research was funded by a grant from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (Grant Agreement No. 741095-PASSIM- ERC-2016-AdG)

About the author

Björn Hammarfelt (Ph.D.) is an associate professor at the Swedish School of Library and Information Science (SSLIS), University of Borås. His research is situated at the intersection of information studies and sociology of science, with a focus on the organisation, communication and evaluation of research. He can be contacted at bjorn.hammarfelt@hb.se.

How to cite this paper