I - Prologues

1 Narrative: A General-Purpose Technology for Science

Mary S. Morgan *

1.1 Introduction

Scholars of scientific life see it filled with experiments, models, theories, descriptions, observations, categories, etc. It is equally full of narratives. Yet the levels at which narratives work, and the kinds of things that scientists come to understand through the activities of developing their narratives, are not easily described in terms of any specific ambitions or functions. Narratives themselves may be understood as a broad class of ‘epistemic genre’, to use the label that Pomata (2014) developed, essential to the representation of scientific knowledge.¹ But narrative is more than just a means of representing such knowledge; rather, prior to such representations, narrative-making plays a wider role in the sciences as a means of sense-making. In contrast with Crombie’s (1988) historically situated categorization of ways of doing science, an account developed further in Hacking’s (1992) philosophical analysis, narrative-making is not mainly about how scientists investigate the world but rather about how they make sense out of those investigations. Narrative-making does not satisfy epistemic questions and worries in the way the interventionist and observational modes of doing science described by Hacking (and others) – such as experimenting, category making, statistical work and case-making – can do. Narrative-making and -using, by contrast, are more closely aligned with ontological questions, or, rather, scientists’ claims in their ‘narratives of nature’ are ontological claims about the way the world is and works. The role of narratives piggybacks onto the epistemology of those other, more interventionist modes of practising science. So, while narrative usage may overlap in places with Pomata’s notion of ‘epistemic genres’ and can be an accompaniment to Hacking’s modes of doing science – narrative-making and -using fulfil other distinct roles for scientists, roles that need separate recognition.

Narrative emerges from this volume as having three functions for scientists: narrative-making operates as a means of making sense of their puzzling phenomena; it provides a means of representing that scientific knowledge; and it provides resources for reasoning about those phenomena. These three functions are related: it is because scientists often make sense of their world by making narratives that they then use those narratives to represent what they believe they know, and thence to reason with them. I propose we think of narrative as a ‘technology of sense-making’ that enables scientists to bridge between their interventionist activities of exploring the world and their knowledge claims about the world, that is, between their epistemic and ontological realms. To label narrative a technology may seem rather strange, but we are in some interesting company here. The philosopher John Dewey argued that the notion of technology was not just about how to make things in the economy, but equally attributable to the abstract and intangible work of enquiry and deliberation involving cognitive work – just as we find for narrative in science.² His contemporary, the sociologist-economist Thorstein Veblen, insisted on the priority of the human element in designing, making and using a technology. While narrative-making, -using, and -reasoning start with the scientist and their community, it is worth remarking that narrative also embeds its own technical elements and attributes. These three separate but related functions of narrative, broadly understood as a technology of sense-making for scientists, may be recognized in the chapters of this book by observing whether narrative is being used as a noun, verb or adjective.

All those nouns of scientific practice – experiments, models, theories, descriptions, observations, categories – hide actions and activities: experimenting, modelling, theorizing, describing, observing, categorizing. Other elements that scientists use don’t immediately convert between nouns and verbs – data has to be given its own multiple verbs (‘to gather, clean, assemble and prepare’), just as laws have ‘to be discovered or made’. Narrative is akin to laws and data: easily understood and effective as a noun, its scope as an activity is not quite so obvious; yet appreciating that scope is critical for understanding the broader role of narrative as a technology for scientists. The quintessential feature of narrative is that it shows how things relate together, so that constructing a narrative account in science involves figuring out how the elements of a phenomenon are related to each other. This is why narrative-making and -using are conceived here as a technology, one that enables scientists to make sense of their phenomena.

These basic usages of narrative in noun and verb forms are important, of course, but they might be still awkward, and limited, if we want to go one step further and conceive of narrative as flourishing in the knowledge-claiming activity of the sciences. In this respect, the adjectival form is more immediately useful: so, ‘narrative account’ and ‘narrative description’ might both be taken for granted. And, while ‘narrative inference’, ‘narrative argument’ and ‘narrative explanation’ might initially sound strange (even perhaps contradictory), it will turn out that we need these terms, for the narrative form does overlap in usage into these scientific activities of reasoning and knowledge-making. Thus, narrative as an adjective works as an attribute of a certain form of reasoning: giving a satisfactory narrative account may go beyond sense-making into the kinds of reasoning associated with inference and explanation.

None of these uses of the term narrative – in noun, verb or adjectival forms – should be problematic if we can find ways to appreciate the active work that narrative does in our sciences, particularly if we can figure out its features and its functions, just as we have for data and laws. These grammatical labels give clues, but only clues, to the ways in which scientists develop, create and use narratives in their various fields, for various purposes and in conjunction with various other forms of scientific representation and knowledge-making activities. These language terminologies need to be filled in with examples and hardened through analyses to reveal the active work we attribute to narrative in science, and so to appreciate how narrative operates as a technology for scientists in doing science.

There are, of course, many commentaries about narrative in other domains, especially in the fields of literature, narratology and legal studies. Narrative scholars from the domain of literature typically focus on the narrative as text: its plots, its structure, temporal and spatial organization, its eventfulness and cognitive function, as well as its rhetorical and aesthetic components, and terms of affect. Narratologists tend to focus on the narrators, readers, what constructions narratives follow, and their requirements for narrative tellability. It is fair to say that with few notable exceptions, neither group focuses especially on connections of narrative with knowledge-making.³ So, in an important chapter, Kim Hajek explores what is narrative about ‘narrative science’, and thus extends the relevant intersections of those fields with our agenda (Chapter 2). Discussions in the field of law about narrative range over matters of rhetoric and affect, but have an equal interest in the putting together of evidence, and the role of ‘theory’ – meaning both the hypothesis about what happened in a particular case, but also the concepts from law that need to be taken into account.⁴ As such, these latter interests fit closest to those of this chapter. But rather than work comparatively with this legal literature I treat narrative in science on its own terms – in order to examine how it makes itself ‘at home’ in the scientific knowledge environment.

Narrative is a broad, expansive term (with many definitions in narrative theory), and the challenge has been to develop an analysis which is insightful for scientists’ creation and use of narrative. Our research shows narrative to be an enabling, general-purpose technology, widely used by scientists within their own different communities to fulfil certain functions in their scientific work – even when they don’t use the word or recognize that label for their activity. It is important to note the limits of this claim: narratives are not found in all aspects of all sciences. Rather, they fulfil certain kinds of function with some regularity in some sciences, or some sites of science, and in conjunction with some methods of doing science. By tracing this (sometimes hidden) narrative activity, and its locations, we can understand both what is different and what is generic in these usages in different sites, and so develop an understanding of narrative in the domains of science.

1.2 Narratives of the Field

The first challenge we address in this book is to see and locate the narratives that appear in our sciences. The most obvious narratives found in science may be those wrap-up accounts in publications resulting from the activities of scientists. In modern science, these are usually impersonal narratives,⁵ cut down to the essential actions that scientists tell of how they went about their research: their ‘research narratives’. Less recognizable, but still apparent, as Robert Meunier argues (Chapter 12), are their ‘narratives of nature’: the narratives – ‘as if told’ by natural, human and social life – that those scientists have tried to reveal, recover and make sensible. And, as he points out, scientists’ research narratives often twine in symbiosis around their narratives of nature.⁶ This has fruitful consequences: the researcher–author, in guiding the scientist–reader along the path of their activities, enables the latter to gain practical familiarity with the former’s narratives of nature, particularly with any new elements and concept set in use.

A broader category of narratives can be found that seek to define and lay boundaries to new approaches for a whole field, or maybe to delineate a new interstitial field. These field-making narratives might be more or less reticent in their agenda. Grand ones are epitomized in the self-proclaimed narratives of those seeking to automate and computerize the whole of mathematics. Stephanie Dick (Chapter 15) discusses two such competing self-narratives in late twentieth-century American mathematics: one group sought to reformulate all mathematical knowledge into one single form, and the other to enable all mathematicians to contribute elements in their own format.⁷ Their politics of control vs. pluralism were explicit. Other field narratives may be more opaque, evident only in their alignments and commitments, to be discovered by an outside reader, as Dominic Berry (Chapter 16) does in looking at how ‘synthetic biologists’ positioned themselves between engineering and biology in defining and growing their own field. He uses longue durée changes in history writing – from chronicles through genealogies to narratives – to argue his case. These are important categories. Chronicles report events solely based on their place in a time sequence without paying attention to any relationships between those events; genealogies focus on the ‘family’ (broadly construed) relationships between the events or objects; narratives provide an account of the relationships between events or objects (whether or not these relationships are tied together in a time sequence or by family connections). Among narratologists, there is a widespread view that a chronicle does not count as a proper narrative because the relational content is absent, while genealogies are just a subset form of narratives. Anne Teather (Chapter 6) adopts the same categories to show how new technologies of dating in archaeology have effectively changed narrative practices in that field. Whereas archaeologists used to tell genealogical accounts to frame the periods of prehistory (e.g., the Neolithic period), more recent technologies of investigation have created the more limited chronologies or chronicles.

Certainly, the narratives of nature – narratives of how the world is and how it works (whether it be the natural, human or social world) – are sometimes much harder to see than these research and field-making narratives. Narratives of nature are more likely to be found implicated with, or inside, other accounts of scientific activity. Like those sherds and trenches of Teather’s archaeological sites, these traces of narrative point to the scientific activities that created them, and from which we must reconstruct the power that narrative-making and narrative-using have in such spheres.

1.3 Narrative: A Means of Scientific Representation

The core function of what narrative does is to bring and bind elements in a subject field together. Narrative-making in the sciences can be found in theorizing, in creating an adequate description of empirical materials or in marrying them to each other in ways that embed ideas and concepts, that is, in activities of sense-making and knowledge-making (examined in sections 1.4 and 1.5). Since the narratives that result from these activities express, or make evident, these connections between elements in a scientific domain, narratives can be treated as a form of scientific representation akin to other forms of representation. What are the characteristic aspects of such representations, and the implications of this way of understanding the role of narratives in science?

First, narrative representations found in science may appear as free-standing or separate pieces of verbal text – in ordinary or natural language. They might be embedded in visual representations (drawn into schemas such as diagrams of mechanisms or detailed representations of empirical matters in graphs), or even expressed in the completely formal languages of abstraction and mathematics. Wise (Chapter 22) contrasts the possibilities of natural and formal languages, and the extent to which they do different kinds of work, and say different things, and thus why narratives in the two forms are not simple translations or transpositions of each other. Depending on the science in question, the narrative form of representation will be more or less formalized, more or less abstract, and may have more or less dimensionality of elements compared to other representational forms of diagrams, equations and so forth. But, whatever their form and language, it is typically the case that they are ‘community narratives’, to be understood without further explanation or accompanying text only by those in the expert community who use them. Mat Paskins (Chapter 13) translates/explains, for us lay-readers, the ‘chemese’ of chemical reaction diagrams depicting the synthesis of particular molecules. He points out that early twentieth-century versions told a different narrative from early twenty-first-century versions of essentially the same representation: in early years, the ‘equation’ expressed the sequence of steps taken to synthesize a certain chemical, but in later years, such diagrams came to narrate the chemical reactions that took place. The ‘cartoon’ narrative shown in Andrew Hopkins’s chapter (Chapter 4) relates what happens in a meteorite impact as material explodes, flows out and gradually builds up deposits on the ground. This requires, for the lay reader, a lengthy verbal narrative that lets us follow the combinations of interacting processes and outcomes from these geological events.⁸ In other cases, indirect representations of nature (such as mathematical models) are manipulated to show the narratives implicit in visual schematic representations. We find such narratives in the computer visualizations from simulating snowflake growth and the processes of chemical reactions, as shown by Wise (2017); the latter offers an alternative free-standing, time-stepped, visualization of the chemical reaction ‘equations’ found in Paskins’s paper (Chapter 13). Such narratives give clues to the density of knowledge that typically lies behind formal language representations.

Second, more often than free-standing independent forms, textual narratives are strongly co-dependent with other forms of scientific representation, such as charts, graphs, drawings, maps, matrices, models, formulae and so forth. Such textual narrative accompaniments might well be an essential part of the identity of those representations, whether of the evidential diagrams in graphs or of the theory-based representations found in models. The classic well-known example is Darwin’s pictorial ‘tree of life’, which – when read alongside textual information – offers a shorthand depiction showing how evolving species branch, or die out, or survive. It is a kind of genealogy – but a conceptual tree not a report of observations. Greg Priest describes this as a ‘scaffold’ on which we as readers can stand to ‘create narratives that enable us to understand’ Darwin’s account of natural evolution.⁹ The infamous ‘prisoner’s dilemma’ model from economics (which was soon transferred to other social-science and biological domains) consists of a mathematical matrix, a set of inequality conditions on those numbers, and a narrative text of the possible behaviours of the ‘prisoners’ given the ‘dilemma’ of their situation (termed by economists, ‘the rules of the game’). The narrative is an essential element in identifying the game and differentiating it from others that may look similar, for the matrix and inequalities are both insufficient (see Morgan 2007). Combinations of text and drawings (keyed with numbers to each other) are found as essential partners in communicating narrative accounts of metamorphic changes in the insect world (from egg to caterpillar, larva to butterfly), as seen in Mary Terrall’s (2017) discussion of eighteenth-century accounts of this phenomenon. Such matching media of visual and text narratives, in which neither is primary but each depends on the other, are also used to explore possibilities of hypothetical events as we see, for example, in D’Onofrio’s account of eighteenth-century generals re-running historical battles according to geometrical lines (in Anthology II).

There is often a kind of bonding here, rather than co-dependency, of forms and functions. Narratives embedded in formal languages and visual representations often provide a highly efficient rendering of the materials of events. The phylogenetic trees of the evolution of the kangaroo and other marsupials discussed by Nina Kranke (Chapter 10) express a travel saga that charts their geographical and biological evolution over time and space as the species evolved while members of its ancestral population ‘journeyed’ from South America to Australia. As she shows us, such ‘trees’ exist in multiple formats – showing in succinct ways, but with distinctly different variants, the narratives of different kinds of family trees or genealogies. Some of these are for professional audiences, some for museums; some are plain, some ‘filigreed’; some read upwards, some downwards, some sideways. There is no one convention despite the related kinds of narratives that are told by these related kinds of trees. There is surely a family tree of such trees, a genealogy of trees, going forward in evolutionary biology from Darwin’s tree of life, and going back in time in a long tradition of drawing human dynastic trees.

This complementarity, and bonding, of narratives alongside and inside alternative representations show how narratives fulfil their representing functions in the sciences and how narratives do the kind of representing work they do. These kinds of co-dependency also suggest there are no strong reasons to privilege narratives as a text form when narratives can find their primary expression in other forms of representation. Narratives in the visual, schematic or even mathematical forms of representation may perform by showing as much as by telling; they are designed to be ‘seen’ by others in the same community of scientists who know how to ‘read’ them. For example, Martina Merz (2011) recounts how readers of a scientific paper in a particular field of physics will automatically follow the diagrams that are arranged in a clockwise fashion at the beginning of the paper – these ‘show not tell’ the research narrative of the salient activities, and readers follow that visual narrative before bothering to read the text of the paper. In some cases, nature’s entities show their own narratives directly. Devin Griffiths (Chapter 7) tells how the Darwins set up plants so that their roots traced out their own growth narratives in scientific experiments. Starting from these visual autobiographies, Darwin constructed narratives at three different genre (i.e., generic) levels: ‘micro-narratives’ of individual plant life, the ‘novella’ of the life history of plants and the saga of biological evolution.

In sum, I argue two points: first, that narratives (like models, diagrams, equations, graphs, etc.) can be understood as a mode of representing scientific things (ideas, theories, processes, evidential records, relations, etc.); and second, that such narrative forms are quite likely to hybridize or be co-dependent with, or even entirely embedded within, those other media of scientific representation.

1.4 Narrativizing: A Means of Sense-Making

Narratives in science are not given by God, or by some other external authority, but designed and made by scientists in their research communities. Attention has to be given to the ways that they create narratives as a means of sense-making – to the active work of narrative formation in the practices of scientists, especially with respect to their narratives of nature.

There are two points here:

1. 1. I take it that the quintessential function of narratives of nature in science lies in making, or unravelling and remaking, connections between things. The world presents many puzzles and scientists seeking to understand their phenomena in their ‘narratives of nature’ have to figure out how that part of their world works, and to give an account of how the bits of it fit together that makes sense. And, like other ways of making other kinds of scientific representation (such as models, schemas, diagrams, tables and category descriptions), narratives have to be developed, tried out, calibrated against other information, reconfigured and re-thought-out to fit the materials that need to be understood. I have in mind something like ‘narrativizing’ – which has the primary and distinctive aim of ‘bringing and binding’ together the heterogeneous elements associated with the phenomena in a field.

2. 2. Following this, I ask: what relational ‘grids’ and scaffolds do scientists use to build their narratives? This is not a question about the structure of the final narrative (whether it has to have a beginning, middle and end with a change of state (see Carrier, Mertens and Reinhardt 2021), nor whether it is primarily ‘tellable’ in terms of sufficiently interesting events (see Ryan 1986), nor on its ‘affect’ and how it facilities multiple connections (see Jajdelska, Chapter 18). Rather – this is a question about the basic dimensions of relations that scientists use in building or creating or supporting their narratives; it is about the lines of relationship on which narrativizing goes on.

First: what happens in narrativizing? The basic role of narrative and its special function for scientists is to put diverse materials into relation with each other through time, or across space, or through other conceptual dimensions (such as classes in society, or elements in an ecology) in order to form a coherent account of a phenomenon. Narrativizing is a way for scientists to organize their bits of scientific knowledge to create sense out of their relations. Narrativizing serves to join things up, glue them together, express them in conjunction, triangulate, splice/integrate them together (and so forth). Yet, the need to clarify relations between things means that narrativizing sometimes means scientists have to sort things out so that their interrelations can be seen more clearly.

One term that captures the challenge that scientists face when they make narratives – explicitly or implicitly – to help them order and relate the separate elements of their scientific knowledge into coherent accounts is configuring. This term comes from Mink (1970 and 1978), writing about the philosophy of history, but he remained opaque about the processes involved. Two other terms, discussed in Morgan (2017), offer recipes with more content for science narrativizing. Colligating comes from William Whewell, who used it to refer to the process of fitting together, under an idea, items primarily from the empirical domains of science (see Cristalli 2019; Kuukkanen 2015; and Swedberg 2018). Juxtaposing was the other term I used then – to refer to the activity of pulling together separate elements known about a phenomenon, but that did not initially make sense together. Narrativizing, constructing a narrative, was a way to make sense of them and resolve initial puzzlement. (This followed a lead from Paul Roth (1989), again for philosophy of history, rather than for science.) Both recipes offer the possibility for creating wider narrative-based understanding or even explanations (as we will see in section 1.5). I want to press the use of Whewell’s terminology of colligation for two reasons. First as a process (in verb form), colligating involves bringing elements together and binding them together just as narrative-making does; the outcome (its noun form) is equally appealing, for a narrative can be understood as a colligation. (A little care is needed here: while narrative-making in science can be understood as a process of colligating (the verb), not all colligations (nouns) necessarily come from narrative-making; for example, the elements brought together could be similar things, bound together in creating a category.)¹⁰ Second, with these two insights of bringing and binding, it is easy to see how the process of colligation can cover the many varied ways in which scientists use narrative to bring together all sorts of different kinds of elements: empirical elements, theoretical arguments and speculative claims. These practices of colligation vary from site to site, and from science to science. Thus configuring elements into a narrative that explores a time-based, path-dependent system in nineteenth-century biology mobilizes a different mode of ordering and relating, both from the juxtaposing narratives of mid-twentieth century case studies in sociology and from the ‘how possibly’ puzzle narratives of modern mathematical and computer-based simulation.

These examples take us to the second point: how does this narrativizing go on? We find two main sense-making strategies, two main relational ‘grids’ for colligating: one based on taking a possible network of relationships as the main device for ordering materials, the other by ordering elements along space or time lines. ‘Grids’ are not to be understood as rigid measuring rods, but rather as a shorthand way to express the main domain upon which the process of colligation – the ordering and relating, the bringing and binding together of elements – happen. These different kinds of grid are not straightforward in use, and often they are used conjointly, because the materials that need to be knitted together in science narratives are not going to be simple connections between elements as if lined up along an individual piece of string, whether that string is a time or space string or a causal path string.

A ‘cat’s cradle’ offers an analogy for narrative-making for the first, causal/associational, version (a term offered in Anne Teather’s chapter). It is a net made from one joined-up piece of string that can be fashioned into several different network patterns. Each network pattern will be different, for it uses the same elements of the string arranged in different ways. Each network pattern can be understood as depicting a set of relationships; the nodes and spokes of the elements may denote an ambiguous relationship, or be causal in a mechanistic kind of way, or they may indicate a much looser association. Indeed, the benefit of colligating or narrativizing on a network grid is that the resulting narrative can be opaque about the exact nature of those relations; it can allow knowledge to be uncertain; it can allow for multiple perspectives; it can enable complexity to be maintained; and it can embrace context where the cut between content and context is unclear.

Time-line and spatial relations seem to offer simpler grids for narrativizing. But, in practice, scientists don’t rest content with creating narratives just by moving along a chronological time-line or arranging items across a spatial grid. Their use of time is not straightforward: they might use relative or absolute time; will cut time up into different units; work backwards and forwards over time, etc. And while time-based accounts in sciences may find narrative necessary, it is important to remember that time-based relations are neither a necessity for narrative, nor sufficient in themselves.¹¹ Of course, things happen in time, and across space, but these may not be the domains in which relationships matter. And even where either time or space may be understood as the dominant dimension for observing change (as in fields such as geology, palaeontology, evolutionary biology and parts of anthropology, sociology and social science history), there is rarely any simple time or space sequencing of events. And often, these two major kinds of grids – relational and spatio-temporal – will mix together in the narrative and will interact. Sometimes, the time–space line enables the scientist to infer subject-matter connections, at other times the subject-matter connections enable the scientist to infer the time or space relations.

It perhaps helps to draw some comparisons. The main feature of the narrative form – that it fits elements together and reveals the connections between them – contrasts with other forms and modes of making and expressing scientific knowledge. The comparison here is particularly with those activities that list or rank elements of knowledge, those focused on activities of separating out similarities and differences between things, and the consequent listing, labelling and describing of such taxa and types. A list of fossil remains, or the table of chemical elements, or the species of natural history, organize or ‘order’ our knowledge according to weights, or categories.¹² They produce nuggets of knowledge, orderings of knowledge, families of like and dislike things, and whole classification systems.

So, on the one hand, the configuring and colligating of narrative-making sit in contrast to the making of tables and categories: the former stitching together relations between things, the latter separating out different kinds of things according to their particular characteristics. On the other hand, those alternative forms of ordering and expressing scientific knowledge are also, like narrative, more than description, for they too are a means of organizing and representing our knowledge. Both narrative-making and category-making develop our scientific knowledge and facilitate our expression of our knowledge about the world rather than being primarily technologies of intervention in the world. And, as usual, there is a caveat: the sense-making quality of narratives (the attempt to find narratives of nature) can also work co-dependently with those other contrasting activities and forms in science (category-making, case-making, statistical thinking and so forth). Narrative ordering and relating do not always substitute, or replace, but may complement other modes of developing knowledge in science. Narrative-making is one potential element, often an essential one, in a multifaceted network of practices that enables scientists to develop ideas and accounts of their domains.

1.4.1 Joining Things Up

The two main relational grids of narrative-making in science, as suggested above, are the spatial- or time-line and the relationship net. Narratives are widely accepted to provide the kind of glue that helps us to ‘follow’ a set of events through time, or across space. Free-standing, time- or space-sequenced narrative representations are found most readily in the historical sciences – natural and human/social – for these deal in matters of time and space and where such dimensions of ordering really matter.

Where time and/or space does matter, scales, measures of time and space, and ways of dating and locating events, may be critical to the kind of narrative made. John Huss (Chapter 3) analyses the competing narratives of the set of major mass extinctions in the natural history domain. The mass extinctions in species evident in the fossil records were recorded in graphs which then had to be ‘explained’ by the palaeontologists – either by a narrative of a periodic event (that might possibly have an unknown astronomical cause) that repeats every 26 million years, or by individual causal narrative accounts for each individual episode. Either way, periodic or individual causes, there was a desire for ‘narrative closure’, the satisfaction of closing the evidential/explanatory gap between the time charts of those visual artefactual fragments of fossils and understanding the causes of the timing of these enormous events. Narrative-making here required getting a satisfactory, plausible and convincing – i.e., narratively closed – alignment of evidential remains with major events, whatever the ultimate explanation might be.

In sites such as evolutionary biology, archaeology and geology, evidential requirements from both time and space typically create narrative density and narrative complexity, as found, for example, in Anne Teather’s account of narrative changes in recent archaeology. Previously, the recognition of familial relations between artefacts and their spatial distribution were used to determine the relative time datings of cultures, transitions and migrations, and so determine the genealogical periods of prehistory (the bronze age, the Neolithic period, etc.). Now, the more recent methods of dating the absolute age of archaeological remains (by technologies of tree ring and radio-carbon dating) determine time relations, namely the chronologies of those civilizations, and so have changed the nature of explanations in that field.

Narrativizing (or narrative-making) in science often relies on a kind of ‘tellability’ that stems, as in Ryan’s analysis of ‘embedded narratives’ in literature, from the intersections and inter-relations of characters that prompt the events or actions that happen. That is, time and space may not be the dominant dimensions needed to follow the sequence or set of events; other relationship factors may be much more important. For example, Morgan (2017) gives an account of the narrative-making habits of social anthropologists working in American cities, where the relationships between a street gang with the police, with the political machine, and with rival gangs are all drawn through the use of narrative accounts. In such contexts where existing social/class relationships are primary, time or space as a grid has almost no value. Narrative-making does especially well in enabling accounts where causal claims contain contingency, doubt and choices, as in John Beatty’s discussion of the tellability requirements for evolutionary change where the order of events is not well evidenced in a time sequence, even though they must have happened in time (2017 and Chapter 20 in this volume).

The intersection of time–space and relationship grids becomes clearer in the notion of ‘process tracing’, an activity discussed by Sharon Crasnow ( Chapter 11) and found in many scientific fields, which involves tracing the evidence of certain relationships (in processes and between events) through time and space and other dimensions, and putting them together into a narrative account. A closely detailed narrative following political changes may be the only way to open out a full understanding of a political science phenomenon which had previously only been accounted for in a spare theorizing or model format, as she argued in studying how political scientists unravelled cases of interactions between democracies to substantiate their ‘democratic peace hypothesis’ (see Crasnow, 2017). We can see in such process tracing how narrative-making actually depends on both kinds of grid relationality – time–space relations and causal relations. It seems in her cases that it is the causal links between events which enable the process to be traced through the time–space events, rather than the other way around. By contrast, in Huss’s mass-extinction events, the time domain is the predominant medium for tracing causes.

Relational grids sometimes function more like bridges in joining up other dimensions. For example, the genetic history in Kranke’s chapter involves following materials that bridge different levels of both time and space in the processes of evolution (Chapter 10). A narrative bridge might provide the link that joins over other gaps, such as it did between different accounts of evolution by R. A. Fisher and Sewall Wright, accounts which nevertheless shared the same mathematical formulation (Rosales 2017). A bridge could offer a methodological joining up, where the research life of the scientist and their narrative of nature intersect in a joint account by the scientist, as Griffiths’s chapter shows for the Darwin family’s investigations into plant growth (Chapter 7). A narrative bridge could be a vehicle for familiarizing the community with the research done, by overcoming the mismatch between actual research events and the given record of events (as in Meunier’s account). Or it could be the way that scientists place their own particular bit of research into a longer or wider research trajectory through ‘narrative positioning’ (see Berry’s 2019 working paper, published in 2021).

1.4.2 Sorting Things Out to Join Them Up

It is one of the paradoxes of narrativizing that it sometimes only succeeds in joining things up by first sorting things out, perhaps in order to join them up in a different order or set of relations than they first appear. The world presents phenomena in puzzling and myriad forms. For example, the massive data sets that come from modern earthquakes have to be sorted out and re-aligned before they can be joined up into any narrative. As Teru Miyake’s account (Chapter 5) of the Tohoku earthquake of 2011 tells us, each measuring instrument at each geographical point produces a data series that tells its own individual story, scaled second by second. These need to be colligated: they need to be sorted out, juxtaposed, aligned and somehow melded back together to produce an integrated full narrative of that quake. Miyake shows how the visual representations of such earthquake data enable the scientists to sort and depict the complexity of an earthquake in narratives that require one to follow time evidence at different geographical points. Narrative-making does the work of both filtering and unifying these multiple records of nature. This is a time–space-rich narrative, in which absolute time matters absolutely, but its narrative focus may be less evident than in Huss’s mass-extinction case because it is so strictly controlled by the technical scientific language of the field. At the same time, in both Miyake’s account of earthquake science, and Andrew Hopkins’s of geology, their analyses show us how narrative sense-making works underground, within and through professional accounts.

Such categorizing, sorting out and putting back together could involve a set of more heterogeneous observations, coming in different forms from different observers in different places, contributing diverse information in the empirical domain. Here the technology of colligating is more like jigsaw-making – where grids of time or space or cause are each separately insufficient, as we see in Lukas Engelmann’s ‘plague narratives’ from the late nineteenth-century(Chapter 14). As in all pandemic diseases, there are many elements that matter, and have to be sorted out for each local account of the causes of the spread of the disease. Here, space and time may be at least as important as the multitude of possible causes that might be ‘traced’ and blamed for such disease transmission. This points us to how narrative-making proves a useful way to deal with complex phenomena that don’t divide well, don’t separate well and don’t simplify or abstract easily but that have multiple elements and agencies. Just as narrative is good for following the connection of events through time, across space and through causal relations, narratives are good for taking all the elements into account without trying to separate them out on the grounds that they don’t exist as separate independent elements – and the scientists’ problem is to understand their interactions. That this entanglement problem can be ‘solved’ through narrative-making is well shown by examples in anthropology (e.g., Geertz, and du Bois, in Anthology II). Narrative-making is even at home where there are conflicting accounts of a phenomenon, which are resolved by understanding how these conflicts are inherent in the phenomena rather than in the scientists’ understanding, as in Hajek’s examples of multiple personality and memory confusion.¹³

1.4.3 Narrative Levels

There is one other important dimension in narrativizing – almost perhaps the first decision for the scientist: what is the focus of the narrative gaze; at what level of interest is the scientific phenomenon; and where is the narrative perspective? The relevant ‘level’¹⁴ may range from narratives about small atoms to the whole universe, from the single individual’s preferences to the market economy, from the smallest ant to the planet’s ecology. Scientists’ narrative-making is a reflection of these interests and decisions. We remarked earlier how Darwin constructed micro-, meso- and macro- narratives of plant life. Teather’s account shows how narrative-making works on two different levels in archaeology – at the broad epoch level of the bronze age or iron age, and at the small local level of the shape of flints to create fires, and in between in the styles of causeways. So we can think of narrative-making in that field as a process of erecting scaffolds on the basis of time-datings, relative or absolute, and then using these to understand both big cultural shifts and, equally important, really specific cultural habits.

Narrative-making can operate under a kind of umbrella for understanding a general approach within a science, or even across sciences: thus narratives of complexity theory, of catastrophe theory, and so forth. Mathematicians (as we have seen) sometimes like to frame their fields in broad and deep terms – a grand narrative of ‘everything’ that should fit under an approach or new form of theorizing (Dick, Chapter 15).

At the other end of the scale are ‘nutshell narratives’. Some of these are ‘anecdotes’ that capture telling examples in very particular short narratives that point to something atypical, extraordinary, unusual or exemplary.¹⁵ They are based on individualized observations and circulate just because they pick up things that don’t seem to fit together. Such juxtapositions are critical, for it is this detection of oddity that sets up the ‘epistemic switch’ that makes the scientist think anew about something. In one of Hurwitz’s cases (Chapter 17), it is the sudden recognition that a baby being observed is ‘well’ which surprises the medic. In another switch, it is from ‘seeing’ something as just a technical fault in a lung X-ray to the removal of a bike-spoke left from a long-ago accident to the patient! In Meunier and Böhert (Anthology II), it is the anecdotes of dogs learning to exchange small coins for buns at the baker’s door that creates new reflections about the natures of animals vs humans. Hurwitz’s epistemic switches are also ontological ones.

Anecdotes come from surprising observations, but other ‘small stories’ come from the scientist’s imagination to prompt theory-making. Stephan Hartmann (1999) tells how a small imaginative story used to launch the ‘MIT Bag model’ lay behind certain theoretical developments in hadron physics. Marcel Boumans (1999) tells how the little story of a child hitting a rocking horse at random intervals with random force motivated a new model of the business cycle in the 1930s. Both of these come from metaphors that were then extended and explored through narrative – a feature discussed by Gillian Beer (1983) in a literary examination of ‘Darwin’s plots’. In these two cases of narrative prompts to scientific model-building, the metaphor-narratives in natural language are extended into theories expressed in formal language (again, see Wise, Chapter 22).

1.5 Narrative Reasoning and Knowledge-Making

So far, we have examined the ways in which narrative provides a means, an enabling technology, for scientists to make sense of their investigations, rather than being a means of those investigations. Yet, we have also seen that narrative-making is not a passive part of science, nor an add on at the end of work, but rather (as noted by Meunier) that scientists’ research narratives are symbiotic with their narratives of nature. Our cases in this volume suggest a more ambitious claim, namely that such narrative-making and -using activate scientific understanding and explanation. Narratives appear in chains and forms of reasoning associated with direct knowledge claims, which can best be expressed in terms of ‘narrative argument’, ‘narrative explanation’ and ‘narrative inference’. Once again, we see that these narrative usages do not provide a competitive path to other modes of scientific reasoning and knowledge claims, but a complementary one.

1.5.1 Narrative Inference

Narrative-making and -using act as go-betweens in inferential domains – offering the means to join together, or mediate between, theories/laws and speculations on the one hand, and data, facts and specific empirical elements on the other. Drawing inferences implies a thesis of some kind that the evidence is asked to speak to; it involves making the connection from evidence to thesis. But this is rarely (perhaps never) entirely rule-bound in any science. Rather, drawing inferences involves some leaps of commitment because the evidence rarely speaks clearly, or uniformly, or exactly, and often has gaps in the chain. Constructing plausible narratives here can play a bridging role to help scientists draw and express such inferences, sometimes preliminary ones that prompt the next step, or search for further evidence.¹⁶

Most often narrative comes into play in inference where the evidence is heterogeneous; and where qualitative or quantitative observations need to be joined up. Elizabeth Haines (Chapter 9) argues that narrative-formation offers a critical resource for picking out bits of heterogeneous evidence, fitting them together and drawing inferences from them, and constitutes a ‘reticulate practice’. As an example, she discusses how a scientist might go about picking out particulars from a crowded field of vision – for instance, in contexts such as the photographic evidence of terrains in order to figure out what is salient and what not in a problem of intelligence-gathering. Debjani Bhattacharyya (Chapter 8) gives an account of two sites of narrative inference. One, offering a similar kind of reticulate practice, is the legal site where the various records of shipwrecks during cyclones in the Indian Ocean – as told by captains and pilots, in ships’ logs and weather reports – are spliced into narratives that draw such evidence together to determine ‘the main cause’, and so apportion blame. Inferences depend here on the consideration of several different narrative accounts, each of which may point to a different cause.¹⁷ On the other side, narrative works to aggregate cases: she tells how taking the evidence from many such storms created the meteorological science of cyclones. This new scientific understanding of the behaviour of cyclonic winds and storms was then used to create ‘storm cards’ which contained little ‘recipe narratives’ telling how ships’ captains should steer their ships when they found themselves in such a cyclone.

Such inferential judgements and arguments may look informal and squashy, and of course such narratives may only be partially informative. But all inference has an element of informal connections to be made. Even statistical inference, which may be strongly supported by statistical rules and criteria, needs subject-matter analysis in order to make sensible claims to answer scientists’ questions. We see this in Lukas Engelmann’s plague narratives. Different narrative accounts of past and current episodes of plague and its treatment based on varied sources of observation were essential to make sensible inferences from facts on the ground. Equally for the scientists seeking inferences about the causes and pattern of mass extinctions. The point here is that knowing about the statistical characteristics of plague does not give automatic entry into knowing about the statistical characteristics of fossil records – the subject matter is so different that simple rules of statistical inference have limited grip; subject matter knowledge, sometimes in narrative form, is needed to draw informative inferences.

Narrative inference may be said to have its own set of criteria for inference. Following the legal literature,¹⁸ one might reasonably argue that the requirements for narrative inference lie in consistency (taking account of all the individual bits of evidence) and coherence (fitting all the elements together in a way that makes sense in the context). For legal cases, there is an additional requirement of ‘agency’ (e.g., of those committing a crime), which might be translated for narrative science in terms of an adequacy/plausibility and perhaps an implicit or explicit agency in the relational claims used in the inference. Surely the most formal inferential rules for ordering and categorizing, and so transforming heterogeneous evidence into a consistent and plausible narrative, is proposed in legal scholarship: namely the Wigmore Chart method, which is designed to take into account the conjunctions among the individually separate pieces of evidence that need to be combined into legal narrative accounts.¹⁹ It is not clear that lawyers follow such strict methods of evidence colligation, but for scientists, it is clear that the use of narratives in a scientific field comes with its own generic criteria for assessing plausibility. Andrew Hopkins recounts how geologists attempting to account for a particular rock formation in Scotland inferred, on the basis of deposited material, that the cause must have been volcanic and told a story of geological formation based on that cause. Some years later, finding a different kind of deposit, the narrative changed to blame the fall of a meteorite. In neither case was there obvious evidence of that particular cause in the presence of a volcanic vent or meteorite crater! In both changes of inference – one might argue – some crucial evidence of the ‘agency’, or cause, was missing when these specific event narrative accounts were constructed against an ongoing background narrative in geology of more gradual causes of erosion and deposition.

1.5.2 Narrative Argument

Narrative argument features in our volume where narratives are involved in making arguments about causes, and about sequences, and about causal sequences – for in practical terms, single causes are hard to come by in science. The philosophical literature arguing about causes is long-standing and wide-ranging. Narrative does not solve those arguments in any principled way. Once again it helps to return to the purpose of narrative – relational sense-making. If, long ago, the adult fish species was upright and then became flat, what evolutionary causal sequence could possibly account for this change (Beatty’s paper, Chapter 20)? (And if that fish species still now begins juvenile life in upright form, and becomes flat only in adulthood, how does this work?) Simple adaptionist stories of efficiency or optimality don’t work very sensibly – the argument does not grip. ‘Back-stories’ are needed to make sense of the adaptations, and of their order, but such arguments may still not be definitive, and it is an open question how far the narrative sequence needs to go back in order to make an explanation that counts as satisfactory with no questions left over.

Strangely, and despite assumptions among some narrative scholars that time is integral to narrative understanding, a given temporal sequence may be consistent with very different sets of adaptations in evolution, or very different causal relations, because, as Jajdelska (Chapter 18) makes clear, narratives have their own power to invest perceptions of causality. The aesthetic details of a narrative matter to our perception and acceptance of such causal claims as being plausible, such that the narrative must be ‘performative’ in this kind of sense. Jajdelska’s argument is paralleled in legal analyses of narrative accounts. This is obvious in court rooms, where the performative aspect of narrative is associated with a degree of rhetoric, but much more interesting for science is that the order in which elements of evidence are introduced into the legal narrative affects the degree of acceptance of the narrative conclusion, just as, probably, happened in those colonial courtroom narratives of shipwrecks during cyclones.

The textual details of narratives are not just performative, but, like the diagrams and schema discussed earlier, they also embed important signals of community expertise. Line Andersen (Chapter 19) analyses how mathematicians read mathematical proofs in terms of ‘scripts’, a literary term denoting slim chunks of text that provide shorthand access to a set/sequence of taken-for-granted background elements for the reader of fictional or everyday factual narratives. For mathematicians, such a script can point to a set of mathematical elements that would be habitual at that point in a proof (a set of proof steps in the background, very different from the kind of ‘back-story’ argument Beatty tells about going back in time). They can best be construed as the denser argument behind the shorthand maths, or the thickness of activity behind the ‘chemese’ found in Paskins’s paper (Chapter 13). As Andersen argues, mathematicians reading a proof expect to see standard habitual moves shorthanded into these ‘scripts’; they are accepted by the expert community without expanding them. But gaps in a series of such scripts, or unusual linking moves between them, alert the community to some strange move in the proof argument – a narrative gap to which they must pay attention.

1.5.3 Narrative Explanation

Traditional arguments from philosophers of history portray narrative as offering an explanation for a particular set of events. By contrast, almost in direct opposition, the traditional philosophy of science position was to understand scientific explanation as both general and valid only if it were ‘covered’ by ‘laws’ as a kind of umbrella. We can see the contrast between these two notions of explanation most vividly in Huss’s account of mass extinctions. The periodic narrative ‘explained’ (according to philosophy) mass extinctions as a regular pattern driven by law-based behaviour elsewhere in the system and was contrasted to the ‘historical narrative explanation’ given for each particular historical case of mass extinction in terms of the reasons why each one happened.

More recently, philosophers of science have settled on a looser or more generous account that portrays explanations as answers to ‘why?’ (and perhaps ‘how?’) questions, but still with a presumption that scientific explanation involves a high degree of generality in its scope (although the strict ‘law-based’ account is now regarded as old-fashioned).²⁰ If we concentrate on how scientists do explain things in narrative forms, we can recognize elements of all these recipes for explanation, sometimes used at the same time.

As we have already seen, scientific narratives often embed causes for things to have happened, that is, they answer ‘why’ questions – so, on that definition, they are readily set up to provide explanatory accounts. Especially this applies to narratives using relational networks, for, as Olmos (Chapter 21) points out, narratives that make sense of relations (causal, associational, etc.) will double as reason-givers in persuading the reader/listener of the knowledge claims embedded in the narrative. This may account for why narrative modes of ‘reason-giving explanation’ work more easily than general law-based accounts in some sciences. But Olmos goes further in claiming that ‘law-dependent explanations’ using time relations invoke narrative as soon as they are examined and unpacked in a way that shows how those ‘laws’ account for real particular events. Thus, taken as an argument form, such narratives of particular events embed law-type explanations.

Olmos’s analysis offers us a framework for understanding narrative explanation more broadly, for we can recognize that there are a number of ways in which narrative accounts in the sciences answer ‘why’ questions while making use of ‘generic’ claims (claims relevant for a class of phenomena) without a full-blown appeal to ‘laws’ (this is particularly so in mechanism-type explanations). Following Olmos’s point, we can find this conjunction happening first in the considerable gap between giving more general explanations and finding the particular ones that might be needed for any specific scientific problem, and to recognize how this gap may be filled by the narrative form. Why is this so common? The ‘laws’ of science are in many cases ‘straw men’ – they are supposed to provide umbrella explanations but often do not organize scientific materials very well – they lie at too general a level to connect immediately or practically with many of the scientific problems studied. For example, scientists from several disciplines with different perspectives have general knowledge about pandemics, but for answering questions about any particular disease-class pandemic, they need to fit together knowledge about the genetic form of a virus, its transmission and medical treatment, and the social behavioural responses that might be relevant to control or eradicate it. As Engelmann’s paper shows, despite widespread generic-level knowledge of the plague, all explanations in his late nineteenth-century cases had to be made local, so each area narrative was relevant to particular causes, transmission, controls and effects of the plague in that context (Chapter 14).

Another conjunction of the two bases for explanation – question-answering with an appeal to the generic level in some form – also works in reverse. It starts with narrative explanations of particulars, but then generates accounts that have more general claims. Thus, in Bhattacharyya’s paper (Chapter 8), we see how, studying and aggregating the narrative accounts of many examples of cyclones, her ‘hero’ Piddington was able to infer the stable characteristics of the behaviour of cyclonic winds, and so set out how ships’ captains should behave in such storms. An alternative mode of extending particular narratives to the generic level was explored in Morgan’s (2017) account of how puzzles thrown up by the juxtaposition of evidence were resolved to answer an important ‘why’ question. The particular case evidence showed firms exited a failing industry in the ‘wrong’ order, according to the theory. The narrative account answered that puzzling ‘why’ question with a narrative of reasons that could be (and was) extended by the community of scientists to ‘explain’ a set of similar cases in similar circumstances.²¹

More often, the narratives of scientific particulars don’t pretend to offer generality, or extend to a more general level, but they rarely work without some generic element (including, at the limit, the use of, or appeal to, general scientific ‘laws’). My earlier account of narrative explanation (Morgan 2017) showed how narratives use conceptual elements from a science to bring together a set of examples under one conceptual roof. Cristalli (2019) has urged that such colligation is the basis for a wider notion of narrative explanation, one consistent with both philosophy of science and philosophy of history. The engineering narratives of particular accident reports rely on general claims and knowledge of the behaviour of materials and people,²² just as the legal narratives of particular cases are set within a framework that uses the general concepts and claims of the legal system. The narratives of ant-lions catching their prey (Terrall 2017) can be understood as particular instantiations of a more generic predator–prey account. The narratives of marsupial evolution told by Kranke provide particular accounts of general versions of genetic evolution. Félida’s narrative of multiple personality is a one-off case, but can be used for broader understandings of such cases (Hajek 2020). These all rely on some kinds of generic or conceptual framing in the narrative accounts. Specific causes can also fit easily and well with laws in a narrative account as they do in geology, where the laws might be said to lurk or police, rather than be specifically determinate (Hopkins, Anthology I). The appeal to a general or generic level, or the use of the conceptual level, is found somewhere in most scientific narratives – in fact, it is difficult to conceive of a narrative in science that does not do so.²³ This characteristic of narrative explanation (like the other answers given here for how narrative arguments extend their reach), does not start from an appeal to something general in the nature of historical or philosophical explanation, but looks to the practices of how scientists do reason with narrative to make their knowledge stick together with theoretical and conceptual materials and so speak beyond particulars to (some) more general kinds of knowledge.

1.6 Conclusion

Narratives are made by people, perhaps mainly for enjoyment, perhaps for enlightenment, but also in the sciences for far more utilitarian purposes; thus my labelling them an enabling, sense-making, technology for scientists. I propose we go further than this, and think of narrative as a ‘general-purpose technology’ (GPT). This term comes from economists and economic historians who have focused on the use of a technology and its histories (rather than its invention or how it is reproduced), and on two particular attributes of such technologies, both of interest for this account of narrative as a general-purpose technology for science.²⁴ First, GPTs are, as the term suggests, technologies with usage that is both generic in its main purpose, but gradually expands across a range of unexpected sites, fulfilling that main function in different ways, and becoming co-dependent with other technologies in the process. Steam power, electricity and computing all offer supreme examples of such GPTs: each has a general-purpose use but is harnessed in different ways for different specific purposes, just as narrative is harnessed across the sciences. Second, and less obviously, those scholars have noted the important role of users, and user-innovation, in the spread and development of those technologies into those multiple sites, and charted how those innovations created changes in economic and social life. For scientific life, our chapters have analysed the multiple and varied usages of narrative and shown how its general purpose of sense-making can be traced into narrative representation, reasoning and knowledge claims.

Of course, narrative is not a new technology, nor invented by scientists just for their use! It would be equally unhelpful to argue that narrative was introduced into science in a particular era, and that it became a revolutionizing GPT as it spread through the sciences in the way that steam power, electricity or the computer did for our everyday lives. Not at all. I am not claiming, nor did our project suppose, that narrative was introduced into science in the way that historians of science have argued ‘the experimental method’ was. Arguably, we could treat that ‘method’ as another possible GPT: historians have tracked how it came into the sciences in the early modern period, taking over the means of investigation and mantle of experiential knowledge, and gradually morphed into the method of controlled laboratory experimentation on the one hand, and field experiments on the other. It has now appeared under many guises (in computer simulations, in medical randomized trials, in thought experiments with models, etc.) and has grown into conjunction with other modes of investigation such as statistical methods and modelling – two other modes of doing science we might also label GPTs – to create the kinds of hybrid modes that characterize modern scientific practices.

In contrast to the laboratory experiment, narrative was surely always in science, and narrative-making and -using in science is a human activity, and so could easily become a social habit in new environments. Thus it surely has a history. Neither this chapter (nor our book) offers any serious history of the place of narrative in the sciences, although we can see a number of points salient for investigating that history of the changing roles, sites and manifestations of narrative.²⁵ And we can suggest the kinds of materials that would be involved. For example, Dear (1991) points to the use of narratives of individual experience in reports of actual and thought experiments in the English tradition of the seventeenth century. Holmes (1991), in response, compares that tradition with French scholars’ narrative modes, which aggregated several experiments at once and which melded their experimental accounts with arguments about the nature of the materials. This is the kind of historical point when, for a particular site of science, narrative turns from being an epistemic genre (using Pomata’s terminology) into a technology that goes beyond simple reporting into something like narrative inference (while still relying on particular modes of intervention and reasoning, to use Hacking’s and Crombie’s ideas). Another hinge point in this historical account of narrative in science might be the one noted by Terrall (2017) in the eighteenth century, when natural historians turned from pictures, narrative texts, or one plus the other, to keying the text to the pictures. Perhaps this is one of the moments when narrative became diagrammatic? These brief remarks suggest that whereas narratives in science may have been found largely as textual free-standing accounts in earlier centuries, the production and usage of narratives appear to have become increasingly intertwined with, and adopted into, other modes of doing science and making scientific knowledge over the last two centuries. In doing so, narratives and narrative-making may have changed in form, but perhaps not changed in their fundamental knowledge-making functions.

Those three GPTs of economic life – steam, electricity and computing – are called so not just because of their flexibility in use, but because they have infiltrated the ways we humans do things in ways which add power to, and expand, our human resources. Narrative, and narrative-making, have expanded or enlivened our human abilities and intelligence as scientists – just as different modes of doing science and different epistemic genres of scientific representation have done. For narrative – as we have suggested – the general purpose that makes narrative as a technology so useful to scientists in doing science lies in narrative’s sense-making possibilities: the power of narrative is to colligate elements together under conceptual frames to make sense of the phenomena that exist in the world.

2 What Is Narrative in Narrative Science? The Narrative Science Approach

Kim M. Hajek

2.1 Introduction: Narrative and the Narrative Science Approach

What do we mean by ‘narrative’ in enquiry into narrative science? How does the Narrative Science (NS) Approach relate to other scholarly interest in narrative? In everyday English, we most often encounter ‘narrative’ used to refer to an overarching position, or set of positions, on some issue – for example, there are competing ‘narratives’ of climate change,¹ while marketers for a brand develop its ‘narrative’ to appeal to particular consumers (see, e.g., Salmon 2008). More basically, ‘narrative’ serves as a synonym for ‘story’. The two gather literature into their associative constellation, such that it could seem straightforward in 2010 for Laura Otis to claim a ‘close affinity’ between literary studies and work in the history (less so philosophy) of science, due to a ‘common focus on narrative’ (Otis 2010: 570). With the overlapping ‘linguistic’ and ‘narrative’ turns, historians have read scientific documents ‘like novels’ (Carroy 1991: 22),² and sometimes joined literary scholars in tracing patterns of influence, shared elements or dissonances between scientific and fictional texts. These approaches have been enormously fruitful, but they disperse their analytic gaze over a wide and highly varied field of view. On the one hand, most studies in literature and science have tended to concentrate their attention on one or the other kind of text – usually novels, since most work in this domain is undertaken by literature scholars.³ Much rarer are investigations which take full advantage of the potential for careful, detailed exploration of formal reciprocities and intersections between narrative fiction and scientific writing (Vila 1998 and Griffiths 2016 are two examples).⁴ On the other hand, when it comes to scientific texts, ‘narrative’ stands in too often for what is primarily an attention to language or metaphor, as in Otis’s 2010 reflections. When narrative appears in such broad terms, it loses its value as a distinct category of analysis. This chapter aims precisely to recover narrative as a discrete analytical category – of significance in its own right, and also as one mode of writing and thinking to be investigated alongside metaphor, themes, argument, genre, etc. in scientific texts and their literary counterparts. In promoting this ‘Narrative Science Approach’, I construe narrative in the specific technical terms of narratology.

Narrowing our perspective in this way has value, first, for understanding the histories and philosophies of science (HPS). As Kent Puckett (2016: 8) puts it, ‘looking at and naming different aspects of [narrative] gives us the ability to see what is weird about almost any narrative’. Narratology (or narrative theory)⁵ provides technical concepts and well-determined labels with which to discuss aspects of narrative; this chapter elucidates some fundamental narratological ideas for HPS scholars (my first set of readers) and demonstrates how these concepts help open up a peculiar set of features of scientific activity – ones we call ‘narrative’. Scientific texts are my priority, as they are in this volume and the wider NS Project; novels make few appearances in these pages. The formalized, technical framework of narrative theory lets us defamiliarize aspects of standard scientific texts like experimental research articles, but also to study how diagrams or computer-simulation movies function in story-like ways – I encompass all of these scientific outputs under the term ‘text’.⁶ With tools from narratology, we can also point to imaginative processes undergirding certain forms of scientific reasoning. My analysis draws together the narratological work done in this volume and unpacks its workings, with the aim of promoting further use of rigorous narrative theory by scholars in HPS.

Such a NS Approach, secondly, has benefits for narratology more broadly, as well as for interdisciplinary research into literature and science. I thus also address this chapter to scholars in literary and narrative studies. (Indeed, bringing together a dual readership follows readily from my own interdisciplinary interests, and accords with the multidisciplinarity of the NS Project.) My analysis offers these readers an exploration of particular ways that narrative analysis plays out in historically and scientifically detailed enquiry. The contextual and technical expertise of historians and philosophers leads to perhaps surprising insights, which can, in a reciprocal movement, feed back into the work of narrative scholars. Studies of the kind in this volume provide much-needed, fine-grained analyses of non-fiction narratives in their particular historical and disciplinary contexts, for instance.⁷ They also open up arenas for productive comparison of scientific and literary texts in strict formal terms. My argument, then, brings narratological endeavours – including the growing field of factual narrative – and HPS studies into dialogue, for the benefit of both areas of scholarship.⁸

What the chapter is not, is a comprehensive introduction to narratological concepts – there exist many handbooks and critical introductions for that purpose.⁹ Nor do I survey all the ways narratology could inform HPS scholarship. Rather, following Morgan and Wise (2017), I concentrate on how scientists use narrative when doing science – as opposed to when they popularize it, or formulate an argument for a wider audience – and what narratological concepts enable us to see and say about such uses. Analyses from the NS Project serve as my principal examples; indeed, even where narratological concepts do not appear explicitly, they wind through many chapters in this volume, providing more or less implicit support to contributors’ arguments. My purpose here is thus twofold. In serving as an introduction to this volume, this chapter sets out how the NS Project thinks about narrative qua narrative. One might say the chapter ‘translates’ commonalities in contributors’ approaches into (some of) the terms of narrative theory. But, like all translations, mine is not neutral; this is my analysis of how narratological concepts provide an angle of entry into this collection. At the same time, the chapter stands on its own as a proposal for what narratology and HPS have to offer each other as fields of enquiry, and where that kind of dialogue might lead. My argument both complements and sits as counterpoint to Mary Morgan’s introduction (Chapter 1) – quite deliberately; each of us offers ways of looking at this collection of essays and at wider scholarly themes that intersect them. We just do not take quite the same angle of vision. Commentaries by Sharon Crasnow and Norton Wise do similar kinds of work at the mid- and end-points of this volume (Chapters 11 and 22).

Use of narrative spans the sciences – mathematical, natural, human and social – as Morgan outlines in detail in Chapter 1. For the purposes of this introduction, I identify two major classes of narrative knowing, each of which is particularly susceptible to investigation using a particular kind of narratological tool. In the first place, there is the ‘mise en mots scientifique’ (after Acquier 2010), or the ‘mise en récit’ (putting into narrative): the (re)presentation of scientific activities or findings in textual form, be that written, visual or spoken. Such texts, as material expressions of scientific work, are at once a product of scientific activity (think of a research article) and an index to the active process of narrative-making. Seen as output, the substantive ‘mise en récit’ takes nominal (noun) form – as a narrative – and overlaps with what Morgan calls ‘narrative representations’. Activity, by contrast, is verbal; what I see as the active flip side of the same ‘mise en récit’ is Morgan’s ‘narrativizing’. But, where Morgan treats the two as separate but related functions of narrative in science, I argue that they are thoroughly, even necessarily, interdependent when seen through the lens of narrative theory. Noun and verb, narrative-as-made and narrative-making, are two sides of the same coin. Both lend themselves to analysis through the output form, the text. Concepts from classical narratology serve to unravel this doubled nature of scientific narratives, as well as to pull out ways in which the events/phenomena to be recounted might differ from the way they are represented – which plots are told, from whose perspective, whether there are flashbacks. Such questions ultimately relate back to the fundamental distinction in narrative theory between story and discourse; this distinction, and what it reveals about scientific activity, is the subject of section 2.3 of this chapter.

Before undertaking this work of unpacking, however, it is worth asking where scientific narratives – as output, noun, representation – sit in relation to the kinds of texts usually studied by narrative scholars. This question is the subject of section 2.2. Until recent interest in ‘factual narratology’ (see Fludernik and Ryan 2020), and even now, narratological categories have predominantly been applied to literary texts, which are readily accepted as being narrative in nature. The NS Approach, by contrast, does not formulate an a priori definition of what counts as a scientific narrative before asking whether we can productively employ narratological tools to unpack (some of) its functions.¹⁰ Rather, contributors to the NS Project have examined both scientific narratives in the uncontroversial sense (like medical anecdotes or psychological case histories), and also (and more frequently) portions or characteristics of texts that might more readily be called ‘reports’, ‘accounts’ or just ‘articles’.¹¹ (Indeed, the French term I have been using, ‘récit’, encompasses both forms.) The broad features of scientific narratives that I develop in section 2.2, using Ryan’s (2007) elements of narrativity, thus emerge a posteriori from the NS Approach.

This definitionally flexible approach becomes especially evident in my second class of narrative knowing in science. Similar to Morgan’s notion of ‘narrative reasoning’ (Chapter 1), I construe this form of knowing as something that a scientist does with a scientific text. Each of us places the emphasis on a different word in the pair, however. Reasoning is privileged by Morgan under her functional approach as something scientists do with and within narrative representations – a deliberate cognitive process, distinct from imagining or affective reactions. By contrast, I understand ‘narrative reasoning’ as cognitively broader, involving imagination, affect and reason, in variable combinations. What matters for me is the combined result of these cognitive processes: story-like representations constructed in the mind/imagination of scientist–readers as they undertake some scientific activity (reading mathematical proofs, interpreting diagrams, framing their field).¹² The attention here is on the reader’s reception of a scientific document, and how it might share cognitive features with the reading of (literary) narratives, without presuming that the document is itself a narrative (representation). Ideas from cognitive, or post-classical, narratology are notably helpful for examining reader responses; I discuss these in section 2.4.

Importantly, this interest in narrative modes of reasoning does not mean the NS Approach makes any broad claims about narrative as a mode of human cognition; even less do we claim epistemic priority for narrative knowing. For all our definitional flexibility, we therefore set aside the perspectives of thinkers like Paul Ricœur (e.g., Ricœur 1980) or Jerome Bruner, for whom narrative fundamentally structures one or more functions of human thought (see Crossley 2010). Asking how narrative modes might enter into human cognition in general is a valuable question; it is just not one that we find particularly helpful in the context of this project. As David Herman presciently remarked in a 1998 commentary, claiming primacy for narrative is to set up an ‘idyll of narrative’ (1998: 385), which essentially only reverses the epistemic hierarchy present in earlier philosophers’ ‘myth of science as univocal rationality’ (Herman 1998: 384).¹³ Either hierarchization precludes fine-grained attention to the contextual nuances of science and narrative studies as historically evolving activities.

It is the evolution and intricacies of scientific activity which concern us in this volume; concomitantly, we do not take account of the historicity of narrative theory as a field of study. Rather, we make flexible use of a range of concepts from narratology and use them to interrogate the doing of science in its active sense: what in science is about narrating, constructing narratives, reading narratives? The narratological tools we employ, the places we find narrative, thus expand and contract with the contingencies of our case studies, and tend to draw from varied perspectives within narratology as a field of enquiry.¹⁴ I reflect in my concluding remarks on what it might mean to look for narrative knowing in a historicized science of narratology.

2.2 Narrativity of Scientific Narratives

When asked, ‘what is a narrative?’, common usage, like some cognitive-science perspectives (Crossley 2010), holds that humans are innately able to recognize story-like configurations. Morgan, in Chapter 1, circumscribes the domain of scientific narrative along functional lines – what it does for scientists alongside or in place of tables, models, diagrams and so on. For their part, narrative scholars have long striven to develop a precise and logically coherent definition of narrative.¹⁵ But NS contributors rarely begin with these kind of definitions, or even ask explicitly, ‘is it a narrative?’, about the documents or actions they propose to analyse.¹⁶ Rather, as illustrated in this volume, contributors find it more immediately significant to plunge into examining a given document’s (or action’s) narrative characteristics and how those function.¹⁷ This notably allows attention to the fragmentary or lumpy ways that narratives can appear in scientific work, which might be overlooked under too stringent an initial categorization.¹⁸ Andrew Hopkins (Chapter 4), for instance, identifies sentence-level narrative chunks in geological research articles. These highly condensed narratives recount the transformations undergone by a rock formation, but are chiefly only recognizable as narrative by trained geologists. The narrative lies between the textual lines of the document,¹⁹ so to speak, a point which emerges secondarily from Hopkins’s study.

In this section, I explore several characteristics of scientific narratives that can be identified through NS enquiry, taking narratologists’ definitional frameworks and theories as a sensible starting-point. Such comparison is additionally essential to developing a genuine dialogue between narrative theory and science studies. My preference is for Marie-Laure Ryan’s (2007: esp. 28–31) manner of classifying narratives according to a ‘fuzzy set’ of conditions on their narrativity.²⁰ Ryan lucidly divides the degree of narrativity of a given text into a number of ‘dimensions’ and ‘conditions’ that span narratologists’ instincts and preoccupations regarding what narrative is. By using her scheme, we evaluate the degree of narrativity shown by a given document, not whether it should be ruled out (or in) as a narrative. Here, I work with three of Ryan’s conditions in order to interrogate some salient features of scientific narratives: whether characters in a story are individuals with a ‘mental life’; the importance of the ‘temporal dimension’; and the issue of narrative ‘closure’.²¹ My discussion, drawing iteratively on chapters in this volume, opens up a few intriguing narratological features of scientific narratives – which may, in turn, inform further categorization work on narrative.

2.2.1 Narrative Protagonists

One of Ryan’s conditions on narrativity that resonates with everyday experience and literary studies is the requirement for narratives to contain some ‘intelligent agents’, with mental or emotional responses (Ryan 2007: 29). That is, a text has lower narrativity if it lacks this kind of ‘mental dimension’. Hopkins’s mini rock-narratives are one example; rock formations as agents have no mental reactions. What is immediately evident from the NS Project, therefore, is the need for a capacious approach to characters in scientific documents, because otherwise many texts would be ruled out of consideration as narratives. Scientific narratives very often recount transformations undergone by protagonists (main characters) that are neither human nor necessarily anthropomorphized: in this volume, the Stac Fada Member (Hopkins, Chapter 4), the Tohoku earthquake (Miyake, Chapter 5), organic molecules (Paskins, Chapter 13) and substances in the fruit fly (Meunier, Chapter 12).²² The first two examples involve narratives about particular individualized protagonists; there is only one Stac Fada Member – a spatially localized rock formation – only one spatially and temporally circumscribed earth rupture process that was the Tohoku earthquake. Hopkins and Miyake do each nonetheless unpack ways that these particularized narratives inform or are informed by generalized knowledge in their fields. On the other hand, organic molecules and biological substances are already less individuated, more generic, narrative agents; even though the fruit fly narratives distinguish between particular substances (e.g., cn⁺ or v⁺), all the instances of cn⁺ are held to be identical (indistinguishable) and to behave in a uniform manner across all fruit flies. When cn⁺ is the protagonist in a fruit fly narrative, therefore, it stands in for a class of identical cn⁺ substances, to be distinguished only from other generic character-substances (such as v⁺).

Robert Meunier (Chapter 12) characterizes the narratives scientists tell about such entities as ‘narratives of nature’; they relate what ‘happen[s] […] when no researcher is intervening or even watching’. As narratives of nature are abstracted, and become part of the acquired knowledge in a scientific discipline, the phenomena they relate also tend necessarily to become stabilized. Their narrativity correspondingly decreases, according to Ryan’s schema; at the abstract, generic limit, narratives of nature tell of (what have come to be seen as) habitual physical events, undergone by generic protagonists without a mental life.²³ As such, these narratives tend archetypally to fulfil conditions of factuality (or posited factuality) in a given scientific field.²⁴

By contrast, mentally reacting protagonists act in particular situations in Meunier’s other category of narratives: scientists’ ‘research narratives’. Here, scientists appear as characters performing specific actions (like steps in an experiment), and their reasoning processes or emotional reactions are often revealed through focalized narration.²⁵ Ryan’s condition about ‘mental life’ in narrativity thus plays usefully into the distinction between Meunier’s two categories of scientific narratives. For again, the scientist–protagonist may either be individualized – like Charles Darwin (see Chapter 7) – or generic, standing in for all scientists in a field (see Chapter 12).

Examining Meunier’s categories in detail can provide insight into the way a given scientific activity functions. The prevalence of a research narrative in an experimental research article helps familiarize its reader with a new approach, especially when, as Meunier demonstrates, the scientist–protagonist becomes generic, allowing the reader to imagine herself in that place. Alternatively, that both categories of narrative are intrinsically bound together in archaeological dating practices is fundamental for Anne Teather’s (Chapter 6) proposal for archaeology to become more reflexive about how research questions influence the narratives it tells about the past. Across studies from the NS Project, we mostly see that, as a field of enquiry develops, its research narratives, with their individual actors and dimension of mental life, yield place to the telling of narratives of nature. This has even led contemporary chemists to call for ‘thin’ narratives of nature, like chemical reaction schemes, to be ‘thickened’ by reinsertion of the research story (Paskins, Chapter 13). Where the two categories of narrative are less distinct is in precisely those sciences which study the human, such as anthropology or psychology. Early psychological case histories, for example, weave together narration focalized on the mental processes of both individual subject and individual scientist–observer (Hajek 2020).²⁶ Can (or should) we distinguish the interplay of ‘research narratives’ and ‘narratives of nature’ as psychologists start to worry about the effect of their acts and thoughts on their subjects of study?

2.2.2 Time in Scientific Narratives

For the vast majority of narrative scholars, it is an essential condition of narrativity that a text deal with events that progress in time; an account of events occurring in a single moment could not be a narrative, for instance, nor could a series of instructions. This is largely taken for granted, such that questions of time in (especially classical) narratology are chiefly a matter of differences between the story and discourse in the ordering or duration of events.²⁷ Many scientific narratives similarly have what we might call a ‘fundamental linearity’²⁸ – a straightforward, and highly significant, temporal structure – particularly those of the so-called historical sciences (geology, evolutionary biology).²⁹ Other work in the NS Project, however, has opened up the question of the relative importance of time sequencing, in comparison with other kinds of ordering that make meaning in a narrative. Mary Morgan’s (Chapter 1; Morgan 2017) notion of colligation privileges relations between disparate items brought together by virtue of a single framing, which may then be woven into nets of similarities and differences; here, orderings other than time are the ‘grid’ by which narratives structure their meaning. Both Morgan (Chapter 1) and the recent work of Carrier and colleagues mark a clear separation between such ‘configurational or coherentist’ narratives (Carrier, Mertens and Reinhardt 2021: 20) and their time-ordered counterparts. Certainly, the two make sense of their subject matter in different ways – according to different ‘grids’, to use Morgan’s terms. Yet the gulf between them is precisely about differences in function, rather than in narrativity, and we should not assume that ‘configurational’ scientific narratives are not also situated in time. It is simply that the time dimension is more or less implicit in the length and order of their ‘events’, as we can see from examining how ‘configurational’ narratives are structured and are transposable.

Chemical reaction schemes provide one example of a scientific narrative that is structured by principles other than time. Each of the diagrams in Paskins’s chapter (Chapter 13) proposes to answer the puzzle of how the molecule tropinone might be synthesized from a combination of other organic molecules. The structural formulae on the diagram are ordered under a causal logic and selected according to whether they show key stages in the transformation of the starting molecules (such as proton transfer or a rearrangement of chemical bonds). If this causal ordering is also implicitly a sequence in time, the duration of each step (between the arrows) is subordinate to consideration of which transformations take place, and which chemical substances are added to or removed from the reaction vessel (see, e.g., notations above and below the arrows).³⁰ Transformations, not duration, are what matters for chemists. These configurations are also of principal import for NS scholars in analysing the function of the reaction scheme as a narrative. What I want to stress is that a progression in time still underlies this kind of narrative, if only in an implicit or latent form. We can see this in two different ways.

First, the reaction scheme is a thin ‘narrative of nature’, Paskins argues, in the sense that the actions of chemist-researchers have been flattened onto the plane of the molecules. If we ‘thicken’ the narrative by reintroducing elements of the research narrative, time re-enters the account explicitly as both ordering and duration, such as in the gloss provided by Pierre Laszlo: ‘let this mixture return to room temperature (rt) over four hours’ (quoted in Paskins, Chapter 13). By virtue of involving human agents, a research narrative will always have some basis in time – human actions are performed in time – and, as Meunier (Chapter 12) demonstrates, narratives of nature are often distilled out of accounts that begin by mixing human interventions and objects’ reactions.

The above logic relies on re-inserting a human actor (or at least a living agent) into the narrative; it is an external logic of time-relatedness, if you will. My second proposal for understanding ‘configurational’ narratives as situated in time proceeds by invoking an internal transposition of the narrative.³¹ I like to think of this as similar to parameterizing the narrative in time, borrowing a term from my training as a physicist. Parameterizing is what mathematicians or physicists do when they take the movement of an object in space, like a ball thrown in an arc, and instead of writing equations showing how its vertical movement relates to its horizontal position, they break both down into how they rely on time. Time order and duration thus become explicit in the latter form, where time is only implicit in the former set of equations (vertical vs horizontal position) – the physicist chooses between them depending on what she wants to examine. Similarly, physical chemists might take a chemical reaction – expressed in the transformation-based (non-temporal) logic of the reaction scheme – and create a simulation that steps in time through the process by which molecules come together, exchange protons or create different bonds (as in Wise 2017). In other words, they might transpose the ‘configurational’ narrative into explicitly temporal steps – for instance to investigate which parts of a reaction occur most rapidly.³² An analogous transposition is described by seminal French narratologist Gérard Genette (1972: 78) when he compares the temporal extension of an oral narrative – the time taken to tell the story – to that of a written narrative: the written text has an extension in space (words on a page), which we can conceive metonymically as an extension in time, in terms of the time it would take to read the text.³³ Moreover, in any number of literary texts studied routinely by narrative scholars, there is a greater symbolic or semantic significance to other linkages than the temporal (Schmid 2013). Some scientific narratives have just as low a degree of narrativity – measured along the time dimension – as many of their literary counterparts studied by narratologists, and the inverse. My point, again, is that both narrative scholars and historians and philosophers can (and do) pose more fertile questions than definitional points about time-situatedness. Chapters in this volume demonstrate other, richer analyses of time in scientific narratives: whether chronologies take a relative or absolute basis (Teather, Chapter 6), or the narrative implications of adopting a periodic temporal structure (Huss, Chapter 3).

2.2.3 Narrative Closure and Narrative Levels

The final element in my discussion of narrativity in scientific narratives is the question of closure, which falls under Ryan’s (2007: 29) ‘formal and pragmatic dimension’ of narrativity. Narrative closure is a matter of a reader’s reception of a text on a cognitive or affective level, and is usually held to occur when a reader’s expectations of the story are met, or their questions answered (Klauk, Köppe and Weskott 2016). To the extent that scientists report completed research actions or propose answers to puzzles, scientific narratives tend to be constructed explicitly as closed (or alternatively as unambiguously open – when a puzzle remains unsolved). When twentieth-century palaeobiologists proposed to account for extinction events in the fossil record (Huss, Chapter 3), their narrative of how such mass extinctions are caused by periodic extraterrestrial events comes in itself to a closed ending: it answers the puzzle question of how and why extinctions occurred.

If the periodic narrative itself, along with most scientific narratives, achieves closure in the basic sense of providing an answer, the concept remains worthy of note in narrative science for pointing to the imbrication of several narrative levels in scientific knowledge-making. Narrative closure is perhaps always a matter of multiple levels, as an individual reader’s affective ‘sense of an ending’ is informed by that reader’s cultural expectations (see Klauk, Köppe and Weskott 2016). In the case of scientific narratives, this multi-level nature of closure is additionally linked to the nature of the scientific enterprise, under which knowledge must be validated by the scholarly community. John Huss (Chapter 3) teases out these intertwined narratives with regard to the periodic extinction story. It was not sufficient for the palaeobiologists to propose this new periodic narrative as explanation; while it offered a closed answer to their question, the palaeobiologists were also impelled to search for evidence to support its claims.

On the individual level, we can consider this search as palaeobiologists’ striving to reach an affective sense of properly ‘scientific’ completeness, in accordance with prevailing scholarly virtues and community standards for knowledge: the extraterrestrial story had to be ‘filled in’ with a certain level of artefactual evidence, however plausibly it accounted for mass extinctions. The palaeobiologists’ search for evidence also arguably constituted a pursuit of narrative closure on the level of the story of their discipline.³⁴ Joseph Rouse (1990) terms this level one of narratives ‘in construction’, in the sense that actors in a field of enquiry conceive of its past and future trajectory in narrative terms, and subscribe to a shared view that knowledge proceeds by seeking evidence for hypotheses and remaining open to revising past accounts.³⁵ Scientific activity, then, interweaves this shared, always open-ended narrative (of science, of a discipline) with the various closed and coherent narratives developed by scientists about their objects of study; it comprises ‘an ongoing tension between narrative coherence and its threatened unravelling’, in Rouse’s terms (1990: 183).³⁶ Examining the narrative condition of closure thus brings into prominence the necessary interweaving of the social in scientific activity (through narratives of a field, or expectations about epistemic virtues) and particular scientific narrative-making by scientists. What remains to be elucidated is quite what might demarcate closure of a scientific narrative in the proper sense, linked as it is to scientists’ affective responses, from the more general tenets of scientific enquiry as it develops through time. For it is far from clear that we should follow Rouse in considering all scientific activity as a narrative in progress – that would be to turn away from our narrower conception of narrative in the NS Approach. Exploring the affective dimension of scientific narratives – why, for instance, some seem more ‘elegant’ or appealing – indeed comprises a vital next step in the study of narrative science. Elspeth Jajdelska’s contribution to this collection (Chapter 18) makes a start, and points the way towards the kind of collaboration between cognitive science, narrative scholarship and HPS that is needed for careful work on these borders between the formal, the affective and the social.

2.3 Formal Matters

2.3.1 Story/Discourse

Thus far, I have been using the rather unwieldy term ‘narrative’ – as noun, as adjective – in relation to conditions on narrativity. One of the fundamental tenets of narratology, however, provides us with the possibility of bypassing the multivalent ‘narrative’ (especially as we use it in English), and delineating different levels of narrative as at once both act and representation. Narratologists conceive narrative as a dynamic relation between a story – the events which are recounted – and a discourse – the way those events are recounted.³⁷ Faced with a given narrative, we only have immediate access to the discourse, that is, to the text of the document. Let us assume that we have a fixed set of events to relate, such as the sequence of actions needed to isolate a biochemical substance.³⁸ We could represent those events in discourse in many different ways. For example, the story of how to synthesize tropinone could be written as a chemical reaction scheme or written in words; it might include essentially no information about the chemist’s actions, or it might add in those actions and their historical context; it could pass quickly over certain steps and linger when telling others. The distinction Paskins draws (Chapter 13) between thin and thick chemical narratives therefore also emerges out of considering how much information, and of what kind, is contained in different discursive versions of a single chemical synthesis story.³⁹ Using terms from narrative theory adds rigour to such investigations, because we can precisely label different domains of narrative structure.

To dissect a scientific narrative into story and discourse also draws our attention to potential mismatches in the order and duration of events recounted, which in turn means we can unpack the temporal dynamics of the narrative in detail. Many scholars have noted, for instance, that scientists do not necessarily recount experiments in the same order in which they performed them in the lab (see Meunier, Chapter 12).⁴⁰ Narrative theorists like Gérard Genette (1972) have given us not only the story–discourse pair (histoire–récit, for Genette), but also a precise, neutral terminology for designating different temporal orderings and durations.⁴¹ As yet, detailed analysis of the temporal workings of scientific documents remains another area to be filled in by further NS studies: for example, how might differing order and pacing (between story and discourse) be used to persuade readers, generate suspense or achieve closure? Here, I develop only several possible strands of this temporal analysis.

We know from the work of scholars like Genette (1972: esp. 78–80) that it is rare in literature for the ordering of events in the story to coincide directly with that of events as recounted in discourse. Fairy tales are perhaps one exception (Puckett 2016: 184–185). In science, short narratives of nature also tend to have the ordering of story and discourse coincide – look at examples quoted at the beginning of Meunier’s and Miyake’s chapters (Chapters 12 and 5). More intriguing is the kind of temporal dynamic required cognitively and epistemically by historical sciences like evolutionary biology. Sharon Crasnow groups these kinds of scientific endeavours under the framework of ‘process tracing’ in her Interlude (Chapter 11) and elucidates their shared reliance on forms of evidence that intermix time and causality. These are phenomena best construed by following the effect of certain causal factors through time, through a process; what does this entail for the relative temporality of their story and discourse?

Let us take John Beatty’s example (Chapter 20) of the evolution of flatfish. The narrative constructed by a biologist to explain this evolution might begin with the observation that flatfish have their eyes offset on their heads – that is, the discourse begins with an observation, which is the end event of the story of how flatfish came to have the features they do. (For the investigating biologist, it is likely a middle-term event.) The discourse would then usually jump backwards to the selected starting point of the evolutionary story – i.e., a moment when flatfish swam upright and had eyes located symmetrically on their head.⁴² But, after this initial jump, for a biologist to provide a properly Darwinian account of the flatfish’s evolution, they must ensure that the story unfolds each of the incremental steps in time order, leading from the fish’s initial form to its form with offset eyes (Beatty, Chapter 20). Such a story is narrative worthy, according to Beatty (also 2016) precisely because of its contingency. Potential evolutionary ‘branches-not-taken’ might appear implicitly, embedded in the narrative,⁴³ but there would not be the kind of jumps backwards (or forward) in time to new sets of events that we see in a novel like Frankenstein, or a classic Freudian psychoanalytic case.⁴⁴ The discourse also compresses millions of years of incremental changes (in story time) into a narrative tellable in human timescales.

The epistemic conditions on such a (Darwinian) historical account require a careful temporal unfolding on the level of the story of evolution; by implication, we would expect this to be reflected in the discourse. That is, we would expect the coincidence in timing between story and reasoning about the fish’s evolution to mean events must follow in sequence when scientists put such a story into narrative (the discourse), such that the crucial time-ordering of events could be conveyed to the reader. Curiously, analogous examples in this volume suggest that this is not the case. Hopkins (Chapter 4) demonstrates that geologists write very few narrative discourses into their research articles about temporally unfolding geological transformations. Similarly, political scientists trace along processual pathways to examine, for example, whether the United States would have entered the Iraq War even had G. W. Bush not been elected president – yet their publications do not recount those processes in order from beginning to end.⁴⁵ Such a choice not to have scientific discourse recount events in their story order seems surprising. To use a frequent analogy between narratives in historical sciences and classic detective stories, it is as though Holmes never unveiled his solution to Watson, but left the reconstruction of steps in the murder to the reader. For now, I can only raise the question; it must be left to further narratological investigation to ascertain the dynamics of ordering and duration in such scientific narratives.

2.3.2 Narration and Focalization

Beyond a careful attention to relative timings, classic narratology also directs us to interrogate whose perspective is expressed and with what authority, at each of the story and discourse levels. It is here that we can most clearly mark the ways narrative – especially in extended, verbal format – is a complex, formal edifice, however ‘natural’ it might often appear.⁴⁶ Narrative theorists differentiate first between the author of a work and its narrator: the author (e.g., Mary Shelley) writes down (or draws, etc.) the narrative, while the narrator tells the story (e.g., Victor Frankenstein). Although author and narrator are often presumed to be one and the same in non-fictional (‘factual’) narratives, Robert Meunier (Chapter 12) argues cogently that we should consider them as separate entities, especially for multi-authored scientific texts. Having posited that distinction, what interests me here are the narrators, the tellers from whose point of view we receive some narrative element: whether they appear as a character in the story, and how directly they reveal their perspective. In a pure narrative of nature, for instance, the narrator tells the story, but is not a character in it; the perspective is an external one, and appears impersonal, as in the quotation which opens Chapter 5. Historians of science will be used to contrasting such an impersonal narrator with the strong, self-fashioned narrative voice typical of eighteenth-century natural science (e.g., Terrall 2017). Such an early natural-scientist narrator is also a character in his story, and often relates his actions and emotional responses in the first person.⁴⁷ But there are more than these two options present in scientific narratives, and that is precisely where using narratological tools reveals complexities we might not otherwise grasp.

We notably encounter more than one internal perspective in accounts from the human sciences, when the aim is to gain access to a human subject’s mental, cognitive or emotional state.⁴⁸ Such interior views can be accessed and portrayed in a variety of different ways. In the following extract, from an experiment involving hypnotic suggestion, there is a shift in the focus of the narrative – it begins with the narrator–experimenters’ point of view,⁴⁹ then shifts subtly to that of the hypnotized subject.

We take another coat and we pass it to M. F…, who puts it on; the subject, who gazes fixedly at this coat with a wondering look, sees it wave about in the air and take the form of a person. ‘It is, she says, like a mannequin with nothing inside it.’

(Binet and Féré 1887: 229)⁵⁰

The hypnotic suggestion in question is that Monsieur F. will be invisible to the subject. As the extract begins, we see the narrator also present as character(s) in the story, performing actions with the coat, and then observing the subject’s reaction. This reaction first consists of external features of the subject – her ‘wondering look’ – described from the narrator’s perspective, before the text moves to portray what the subject sees, and then relate the subject’s words about her vision. Throughout, the telling is done by the narrator–experimenters; they refer to the subject in the third person. But the narrative also relates information to which, logically, the narrator–experimenters do not have access, in the form of the subject’s interior view; there is a shift in who ‘sits behind’ the words of the text, with the narrator–experimenters and the subject ‘doubling up’ for this part. This is an example of shifting narrative focalization.⁵¹

What I want to emphasize are the kinds of questions we can ask after noticing such a shift (or, more often, repeated shifts) in focalization in a narrative. On the one hand, the subject’s perspective is stamped here with the authority of the narrator as (a pair of) scientists. The description of what the subject sees is an interpretation, based on or validated by the subject’s words (which are also reported). On the other hand, noticing the shift in perspective – and that it occurs before the subject speaks – draws our attention, as readers, to the representational surface of the text – to the fact that it is a presentation of the story, and that there might be others. There is notably a small temporal mismatch here, since the narrator–experimenters’ interpretation, which occurs first in the discourse, must logically follow the subject’s speech on the level of the story. We are reminded that the immediacy of this experimental report is constructed, that writing occurred after the activity of the experiment. Did, therefore, the subject say exactly what is reported, or are the words (also) a reconstruction by the narrator–experimenters to validate their interpretation? More fundamentally, when did knowledge-making occur here – during actions, or during writing, or both? I would stress that it cannot be fixed down; narrative, (even) in its textual form, is not only an output of scientific activity, but fully and necessarily participates in the activity of knowledge-making. This is narrative as ‘the expansion of a verb’ (Genette 1972: 75), or the binding together of ‘narrativizing’ and ‘narrative representation’, in Morgan’s terms (Chapter 1).

If, in a sense, this brings us back to the kind of arguments well known in history of science under the label of ‘constructivism’ (e.g., Golinski 2005), it does so from the distinct perspective of narrative. Formal narrative analysis can do more than signal that knowledge emerges from putting scientific activity into words. It can suggest different patterns of authority in narratives from sciences which study humans, compared to those which do not. My brief analysis above, for instance, points to the ways that shifting narrative focalization seems essential to the business of the human sciences around the turn of the twentieth century, but also to a concomitant trade-off in the form of a more unstable textual authority. Further work could study how textual dynamics of this kind articulate with scientists’ avowed theoretical orientations; for example, do behaviourist psychologists, who eschew internalized observations, nonetheless produce focalized narratives? How do these dynamics compare with narrative focalization in accounts involving anthropomorphized (non-human) protagonists, on the one hand, or multiple interacting humans, on the other hand (as in social sciences like anthropology)? Curiously, there is narrative focalization on plant growth at multiple narrative levels in the Darwins’ Power of Movement in Plants – not only when the Darwins narrate their story, but also when the plants themselves are (co-)narrators, as Devin Griffiths’s narratological reading reveals (Chapter 7).⁵²

2.3.3 Which Comes First?

Analysing shifts in narrator focalization prompted me to ask whether Binet and Féré’s subject spoke the exact words related, or whether the experimenters filled in a plausible comment while writing their text. In story-discourse terms, this is equivalent to asking whether Binet and Féré’s text – as discourse – reports a pre-existing story and just reorders the events, or, alternatively, whether portions of the story are only constructed (and re-constructable) through their inclusion in the discourse. As Kent Puckett (2016: 35) asks: ‘Do events precede their representation, or does a representation somehow produce events as significant and thus knowable?’ This ‘paradox’ (Puckett 2016: 215) points to a central tension in narrative theory over which of story or discourse comes first; it has been a productive force structuring the work of key narrative theorists, as Puckett sees it. NS studies also provide a particularly rich site through which to trace the dynamics of this tension, with conclusions that can feed back into theoretical work on narrative.

I am not advancing some radical constructivist view here, as if there were no reality outside of that which is ‘mise en récit’ in a narrative. But, when it comes to scientific narratives, it is not always straightforward to identify what counts as story, as against the discourse, especially when we are dealing with non-human, non-anthropomorphized protagonists. Hence the richness of scientific narrative. Indeed, Meunier (Chapter 12) enunciates how both discourse and story (as events and their ordering implied in discourse) can differ from the events that took place in the experimenter’s laboratory in ‘reality’, or the ‘practice-world’, as Meunier terms it – and this even for actions performed by and recorded by humans. When an archaeologist finds many Neolithic stone axes at some site, these can, on the one hand, serve as evidence or markers of story events – through some absolute dating method, for instance. On the other hand, the archaeologist might construct a narrative about popular stone quarrying sites, which might frame the axe find as a trace in a story about demand for felling trees.⁵³ Either way, story and discourse sit in a dynamic relation within the activity of scientific narrative-making.

The interplay of story and discourse is particularly clear in those scientific endeavours where narrative is not an end point, but where discourse-making and story-reconstruction occur iteratively.⁵⁴ In this volume, Teru Miyake’s study (Chapter 5) of seismological work on the Tohoku earthquake is a salient example. Miyake’s seismologists first take evidence from a single kind of sensor and configure it computationally into a time-stepped narrative simulation of how events in the earthquake occurred: the rupture narrative. Many rupture narratives are generated (e.g., from different types of sensors), and then compared by seismologists, who next extract details which are present in several rupture narratives; these details are treated as story-level events. Finally, ‘these distilled details are strung together into a model-independent rupture narrative, which [Miyake] call[s] an integrating narrative’. In narratological terms, successive steps in this scientific work take each of story and discourse, respectively, as pre-existing. Rupture narratives are first configured from story points (i.e., the sensor data evidence), before a switch in perspective, which construes the discourse of the rupture narrative as a source from which to reconstruct and extract a different set of story elements (Miyake’s ‘distilled details’). The final step flips perspective yet again, back to the work of constructing a narrative discourse (the integrating narrative) from (the new set of) pre-selected story details. Morgan (2017; and this volume) speaks of ‘narrative inference’ as unravelling and reknotting sets of evidential or conceptual elements.

If these iterative steps are clearly separated in Miyake’s account, we could speculate that such dynamic work of narrative configuration and reconfiguration is in play in scientific activity more widely, especially where phenomena are not directly observable. For instance, Elizabeth Haines (Chapter 9) points to a doubled way of working within visual narratives, when she shows how ‘neither evidence collection nor explanatory accounts were prior’ in Hugh Hamshaw Thomas’s botanical and intelligence-gathering practices. Opening out from this NS work, we might ask further whether scientific narrative-making (and re-making) of this kind could serve as a useful model for broader processes of narrative-writing and narrative-reading.

2.4 Narrative Reasoning

For now, I turn to existing narratological understandings of reading practice and how they can illuminate scientific reasoning. No telling is without its implied or actual readers, and they too perform important work in narrative-making, in an interplay with the narrative as textual or visual material. In a sense, therefore, I move now from considering narrative as the dynamic relation between story and discourse, to considering an interconnecting relation between discourse/narration and reader. It is a move which brings us into the domain of cognitive narratology – a field that combines findings from psychology and artificial intelligence to explore relations between story-text and -language, on the one hand, and human memory, perception and affect, on the other.⁵⁵ Concepts from cognitive narratology are well suited to tracing the kinds of processes occurring in a reader’s mind (or imagination) as they read a scientific text or diagram; notably, narratological concepts point us towards elements of scientific reading practices that might well be compared to ways people read fictional texts.

I construe such cognitive processes under the banner of ‘narrative reasoning’: they comprise story-like imaginative constructions which scientific readers generate when reading a research article or examining visual evidence.⁵⁶ If the scientific text in question has a clear narrative discursive form, narrative reasoning in the mind may not differ greatly from the logic of the narrative on the page, or it might be inferred using more classical narratological tools (of the kind discussed in the previous section).⁵⁷ Narrative reasoning is more distinctive as a component in scientific activity when story-like imaginative work is prompted by apparently non-narrative scientific texts – texts with very low narrativity (to link back to my earlier discussion). An example I have already evoked is the ‘implicit’ or ‘covert’ narratives of historical sciences like geology, which Hopkins argues only unfold as narratives to an informed reader. To interrogate narrative reasoning under my NS Approach is to examine the processes by which a scientist imaginatively replays such narratives, and, importantly, how these processes map onto particular textual elements. This explicitly adds a textual dimension to the narrative thought processes opened up by Morgan (Chapter 1). We might refer to tacit knowledge, scientists’ trained judgement, or their horizon of expectations – to invoke some concepts current in HPS and narrative studies. However important, these are not enough for rigorous narrative enquiry, since they operate on a more general level: they relate texts as a whole to broad-scale expectations or knowledge in a field. With the NS Approach, we can delve into the specifics of which particular elements in a research paper or diagram activate story-like imaginative responses, as opposed to other cognitive functions. Notions like narrative performativity and scripts allow contributors to this volume to begin this work.⁵⁸ I briefly outline their findings in what follows.

As Elspeth Jajdelska emphasizes in Chapter 18, the question of who narrates a story and in what circumstances matters for its reception. Jajdelska transfers the notion of narrative performativity from the spoken to the written domain and, in a recursive move, elucidates its workings in a research article about cognitive science. Performative language is what early narratologists might have called properly literary language, in that it draws attention precisely to its aesthetic qualities. It thus bears a greater affective force and implicitly cues a certain imaginative worldview. The worldview thus rendered can encode assumptions or perspectives which support a researcher’s explicit argumentative position, as in the article analysed by Jajdelska. Importantly, under this framework, particular textual passages, or even a few words, can be identified as corresponding to a story-like cognitive effect – one which plays a highly significant role in the knowledge claims of this scientific article.

A different kind of small-scale textual (or visual) element that produces story-like reading is the script (Herman 1997). In her chapter examining how mathematicians read proofs, Line Andersen deploys this concept from early cognitive narratology to argue that mathematicians read proofs similarly to how people read fictional narratives. That is, portions of the proof call up a sequence of events or actions that are expected or appropriate in the context in question. These proof-segments operate, in other words, like the scripts in literary texts for events such as ‘eating in a restaurant, riding a bus, watching and playing a football game, participating in a birthday party, and so on’. As the AI researchers who developed the notion go on to say, ‘These scripts are responsible for filling in the obvious information that has been left out of a story. Of course, it is obvious only to those understanders who actually know and can use the script’ (Schank and Abelson 1977: 41).⁵⁹ Andersen develops the correspondences between script-activating elements of a proof and steps in mathematical understanding. Like readers of novels, the mathematical reader performs the mental action of running through steps cued by a script, but since scripts deal with expected sequences of actions, the reader’s attention is particularly caught when a proof deviates from the expected background of mathematical scripts. By undertaking such narrative reasoning, mathematicians are prompted to focus on the novel, likely crucial, elements of a proof. Reciprocally, HPS analysts like Andersen can identify more precisely which elements count as most significant in mathematical reasoning and understanding, and for which kind of readers, since script-activation depends on a reader’s level of understanding of an expected situation. Notions such as scripts, narrative performativity and other ideas from cognitive narratology could similarly be applied to many domains studied by HPS scholars. Wise, for instance, broadens the notion of script to several areas of scientific knowing in his Finale to this collection (Chapter 22). But where such an approach might bear most fruit is in combined textual and ethnographic analysis, of the kind sketched by Andersen – specific elements of a scientific text can be connected to particular narrative-like reasoning, and that mapping contrasted with scientists’ own accounts, as well as analysts’ reconstructions, of scientific activity.

2.5 Conclusion

Narrative theory is an extensive and complex field and, in this chapter, I have only worked through some of its key concerns and ideas as they apply to scientific narratives. My aim in doing so has been twofold. On the one hand, I have sought to encourage HPS scholars to treat narrative in the focused, technical terms of narratology, by demonstrating the analytical productivity this promotes. Such analysis – as undertaken in the NS Project and chapters in this volume – reveals that a ‘mise en récit’ always involves an active component of knowledge-making or reasoning, even when a narrative (representation) is also the output of some scientific endeavour. Reciprocally, if narrative in science is always active, it is not an activity divorced from any concrete, material basis; a major part of the value of narratological tools is that they can serve to trace precise connections between narrative as text and narrative as mode of reasoning. What the NS Approach provides, then, is precision and rigour to an object of study – narrative – that otherwise risks overflowing its conceptual bounds to such an extent as to offer no meaningful basis for comparison or interpretation. NS offers exciting perspectives as an approach deployed alongside the usual epistemic resources of HPS.

On the other hand, this chapter elucidates the various ways in which work in the NS Project is informed by concepts from narratology, even where such concepts are not emphasized or delineated. As historians and philosophers of science, contributors to this volume bring a sensitivity to the theoretical and contextual constellations in which their case studies can be situated. Our studies thus bring a depth of detail to explorations of narrative in a non-literary domain – they can complement and complete narratologists’ investigations in this area with much-needed science-specific expertise. Just as I hope future HPS work will be open to narratological perspectives, I similarly encourage narrative scholars to draw upon HPS expertise, as showcased in this volume, in developing their field beyond the literary. This chapter has notably pointed to some distinctive characteristics of scientific narratives – their frequent non-human, even generic, protagonists; their iterations of story-making and discourse-configuration – as well as proposing that there is less of a divide between scientific and literary narratives than often assumed, when it comes to their situatedness in time – it is just that different questions of timing might arise. And, of course, there remain many areas of enquiry where collaboration between HPS and narrative studies would be fruitful: the affective charge of scientific narratives, forms of narrative focalization and the particular interplay of ordering and duration in work in the historical sciences, to mention just some I have signalled above.

But to conclude this chapter I would like to turn briefly to the ambitions held by narratology to be considered a science, from its pre-history in Russian formalism to its more recent cognitive turn.⁶⁰ Could we apply the NS Approach to narratology itself? As Puckett (2016) stresses, narratology as a domain of enquiry is not without its own history. Where he historicizes it in terms of key political and intellectual currents, we might ask how narratology is informed by other scientific fields and what role narrative-making plays in its endeavours. If we had to classify narratology, we could place it in the category of the human or social sciences, as taking a human product – narrative – and its cultural and social imbrications, as its object of study. We might then sketch a shift in perspective from a view of narratology influenced by the model of chemistry – with stories dissected into a fixed set of re-combinable elements – to one that enacts something of a convergence with cognitive science and some branches of psychology. Early structuralist Algirdas Greimas (1983: 65), for instance, praised the language of chemistry as ‘a semiotic form which must, across all kinds of language, serve to express its meaning’,⁶¹ while to read Manfred Jahn’s encyclopaedia entry (2010) on cognitive narratology is to be plunged into considerations of ‘preference rules and processing strategies’ that would not appear out of place in a research article in computational science. By analogy with chapters in this volume, we might speculate that early structuralist narratology mobilizes ‘thin narratives’ of the kind identified by Paskins (Chapter 13), or that recent cognitive theories enlist strategies of ‘narrative performativity’ to provide imaginative support for their claims (as in the article investigated by Jajdelska in Chapter 18). What might such a transition imply for understanding the evolution or limits of narratology as a ‘historically specific logic’, to use Puckett’s terms (2016)? When we apply a notion like the script to a scientific narrative, to what extent do we invoke distinctively narratological theorizing, as against ideas from the script’s origins in AI? Or, is to pose such questions to descend into a methodological spiral, where narrative and science turn circularly around each other?⁶²