ISSA Proceedings 2002 – On The Argumentative Quality Of Explanatory Narratives
This paper tentatively draws together the three concepts of argumentation, narrative and explanation. The three concepts are all highly rich ones and denote complex areas. Some parts of each conception may have implications for or illuminate the other two – that will depend both on what one takes each of them to be, and on the perspective one chooses to employ. The existence of rival views within all three areas further adds to the complexity.
An exploration into the argumentative quality of explanatory narratives is a venture that requires great caution. Some explanations are arguments and some narratives are explanations, but it does not automatically follow that some narratives also are arguments. Again, it may depend on what one takes them to be. Should it emerge in the course of the analysis that narratives indeed are not arguments, I think that argumentation theory nevertheless can throw critical light on explanatory narratives. There is a significant overlap in vocabulary (e.g. use of such concepts as premise, antecedent, conclusion, warrant) that indicates the usefulness of argumentation theory, but equally evidently this overlap may cause confusion and mix-ups. Again, caution is called for, as well as precision.
My proposed exploration minimally requires that the notions of narrative and explanation be discussed such that the connections between them can be made clear. Furthermore, the connection between arguments and explanations must be discussed. Then we may find ourselves in a position to tentatively use argumentation theory to evaluate such narrative explanations; for example whether narratives distinguish between what is part of the narrative and what is evidence for the truth of its premises.
But first, the concept of a narrative, as it will be used here, must be made clear. My discussion will refer mainly to empirical narrative research done in the field of education, but it should be made clear that narrative theory is an interdisciplinary field, covering e.g. literary theory, history and education. Originally, narratives are fictional stories and belong to the domain of literary theory. I will not here discuss the wisdom in importing narratives, with all their connotations and presuppositions, to the educational field.
What Narratives Are
Despite the enormous existing body of literature, the notion of a narrative remains fairly elusive, and its uses in educational research largely contested. It seems that there exists no generally agreed-upon view of narratives, but rather a set of overlapping meanings from which advocates of narrative may choose the meanings that best suit their intentions. Narratives are discussed as a way of making sense of life, a phenomenon, a method and a result (product) of this method (e.g. Carter 1993, Casey 1995/96, Clandinin & Connelly 1991). I will focus on narratives as products, that is, as written texts. The educational literature on narratives by and large focuses on elements or items to be found in such narratives in its explications of what narratives are. Thus, we are told, narratives consist of events, actions, intentions, characters and plots. These items are connected in some way; frequently it is required that they be organized in causal sequences (Gudmundsdottir 1990). The causal sequence makes up a meaningful, coherent whole with a beginning, a middle and an end. The demand for coherent wholes with non-random beginnings, middles and non-random ends dates back to Aristotle’s Poetics (1982). His requirements concerned poetry and fiction, but have spilled over into narrative educational research – they are hardly left out of any accounts of narrative research. Events and actions should follow from one another, not just after one another. Things that happen by chance or are randomly present can hardly be fitted into this kind of coherent wholes.
Cheryl Mattingly (1991, p.242) defines narratives as follows:
Narratives, first of all, concern action. In stories people do things and as a result situations change, or things happen to people and as a result the people change. … stories do foreground intending, purposive agents in presenting how things have come about.
“How things came about” seems to be central to various definitions of narratives. Narratives may be conceived as representations of sequences that show, tell or explain how something came to be; how a certain result was achieved etc. It is the explanatory function of narratives that will be the focus of this paper.
Implicit in Mattingly’s definition is a broader claim made on behalf of narratives, namely that they deal with particular events, actions, characters or intentions. For example, narrative researchers within the teacher thinking tradition investigate, describe and explain particular events that happened to particular teachers and/or particular students in particular classrooms at particular times (e.g. Gudmundsdottir 1990). In a much-cited article, Jerome Bruner (1985) makes a distinction between what he terms paradigmatic and narrative modes of thought. The paradigmatic mode is the logic-scientific one. It deals in general causes, timelessness, universal and context-free explanations, empirical proof and consistency. The narrative mode, on the other hand, seeks explanations that are particular and context sensitive, it deals in human intentions and actions; it is essentially temporal and does not establish truth, but verisimilitude. Thus, in Bruner’s words, “The imaginative application of the paradigmatic mode leads to good theory, tight analysis, logical proof, and empirical discovery guided by reasoned hypothesis. The imaginative application of the narrative mode leads instead to good stories, gripping drama, believable historical accounts” (p.98). On Bruner’s view, the two modes are irreducible.
As a prelude to the subsequent analysis, let me cite in full a narrative as it appears in an article by Cheryl Mattingly (1991). It seems that this narrative is fairly representative of empirical educational narratives, although this inference is based on an admittedly small sample and therefore must be viewed as highly hypothetical. Narratives tend to be short parts of “ordinary” research texts. Often they do not satisfy the criteria for something’s being a narrative. How ever that may be, this is the story of a student therapist (1991, p.246):
When he [the patient] was on drugs he could do all the ADL [activities of daily living]. When he was off, he couldn’t do anything. He had a mask-like facial expression. His changing ability to function was frustrating for him and his wife. The only adaptive equipment I gave him as a shoe-horn, because it was difficult for him to put on his shoes. I suggested … [unclear] but he didn’t want that. He said that something would have to be changed because his bedroom was downstairs but finally agreed that he could have a bedroom in the living-room. He progressed rapidly and after a week and a half he was smiling, becoming more social. His wife told me, “He does nothing at home”. I don’t know if she could hear what we were telling her. We said, “He is not just sitting around. Many times he simply can’t do anything because of the disease”. When the wife heard that he would be on medication and that this would improve his functioning she said to him, “Good. There’s a lot of chores around the house you can do”. I don’t know how much she heard of what we were telling her.
This story has no obvious plot, there is no causal sequencing, no non-random beginning or ending and no obvious temporal order. It is not even obvious precisely what “came about”, or how. On the other hand, there are characters, actions and intentions, and I will, for the sake of the argument, accept the story as a narrative – it is after all presented as one. It illustrates what seems to be a paradox in narrative empirical research, namely that the actual narratives appear to be rather simple compared to the rich and sophisticated theory about narrative that exists.
Narratives As Explanations
Advocates of narratives in education do not explicitly state what they take an explanation to be. They seem to take for granted that it is immediately understood or in some sense self-evident. This appears to be quite common. Indeed, as Frederick Suppe (1989) points out, virtually all literature on explanation tacitly assumes that explanations are explanations why; that is, an explanation is equivalent to an answer to a why-question. Already at this point we run up against a possible problem, since Mattingly clearly thinks of explanatory narratives as answering how-questions. On the other hand, literary theorist Paul Ricoeur (1984), who has an extensive discussion of narratives as explanations, sees explanations as why-explanations. I shall return to this problem subsequently.
One universal characteristic of all explanations is that events are explained after the fact. It is of great significance here that we already have knowledge of the result, the event, the happening when we set out to explain it, namely that it did take place. I shall return to the question of whether this trait effectively bars narratives from being arguments. For the moment let us focus on narratives as a kind of genetic explanation: a story leading up to the event-to-be-explained. As suggested above, this must involve causally relevant antecedent events. Explanatory narratives thus are reminiscent of what Wesley Salmon (1990) has termed the “ontic” conception of explanation: the explanation of an event is what produced it. In narratives, this is usually cashed out in terms of causal chains hooked up by a hindsightful narrator.
Hindsight plays a crucial role in the configuration of narratives. Configuration or emplotment means the “grasping together” of all the items a narrative consists of into a coherent whole with a non-random beginning, a middle and a non-random end. Telling stories of “how things came to be” clearly presupposes hindsight, even though advocates of narrative research do not discuss what is principally involved in hindsight reasoning. According to Mattingly, we tell stories of how a thing came to be by returning to its origins and tracing a coherent story from origin to present. That is, we first reason backwards and then tell the story forwards again. The hindsight position means that we possess knowledge of the outcome; that is, we already know the “conclusion”, the end, for a fact. Narratives are thus explanations by retrodiction. As Paul Ricoeur puts it, in retrodiction we begin “… from the fact that something has happened, we infer, backward through time, that the antecedent necessary condition must have occurred and we look for its traces in the present, …” (1984, p.135). The implications of hindsight for configuration are most vividly described by David Carr (1991). Whereas a radio announcer who gives a live description of a baseball game must describe what happens in the order that it happens, a narrator’s position is entirely different. The narrative of the game “… is told afterwards and in full knowledge of who won. It will mention only the most important events, especially those that contributed to scoring points and thus to the outcome” (p.59).
What happens first of all in retrospective reasoning is a re-description of events and actions in terms of later events. Re-description is an act of configuration or emplotment in that is serves to tie events together by relating them causally. With knowledge of the outcome, the “conclusion”, earlier events can be re-described as causes of the outcome in question. Philosopher of history Louis Mink (1978) maintains that hindsightful redescriptions are necessary in the construction of narratives; without hindsight things hardly hang together. Thus, without knowledge of the end we cannot pinpoint the beginning. Drawing on the work of Arthur Danto, Mink says that the narrator accomplishes this re-description by using a certain class of typical historical statements, namely those that describe events by referring to subsequent events – so-called narrative sentences. Such sentences also abound in everyday language usage. For example, “The murder of Grand Duke Franz Ferdinand on the 28th of June 1914 started WW1”. When the shot was fired, nobody could have known it was the beginning of WW1. Common to all narrative sentences is that the original entity is described in a manner in which it could not have been described when it took place. The reason is that the description makes references to events that had not yet occurred at the time. The original entity gains its significance in the light of subsequent events, and a coherent narrative with beginning, middle and end may be produced. This kind of retrospective re-description surely must affect the explanatory power and goodness of narratives. I shall return to the issue.
Generality And Particularity
But first it is necessary to inquire into certain features of explanation theory, as they apply to narratives. To begin with, we should note that it seems to be agreed among philosophers of science today that there is no single logic of explanation. As William Dray pointed out as early as in 1957, the term “because” does not commit the following answer to any particular logical structure (Dray 1957). Wesley Salmon, in his overview of philosophical literature on scientific explanation, maintains that “explanation” is used in many ways that have little or nothing to do with scientific explanation (Salmon 1990). It is a question where that leaves narrative explanations.
As we have seen, scientific explanations have generally been seen as answers to why-questions, and Carl Hempel’s Covering Law model (hereafter CL-model) is generally seen as the first serious attempt to spell out what constitutes a correct answer to a why-question (Hempel 1965, 1966; the model was first published 1948 by Carl Hempel and Paul Oppenheimer). On the CL-conception of explanation, explanations are arguments, and all explanations contain and make use of a law or a (statistical) regularity. Hempel explicitly rejects the idea that causality plays any essential explanatory role (1965, p.352). In short, an event (explanandum) has been explained when it has been subsumed under a law or a regularity; that is, has been shown to be an instance of (be covered by) a law or a regularity. The deductive-nomological (DN) version is a valid, deductive argument in which the explanandum is deduced from premises consisting of a law and of initial conditions. The inductive-statistical (IS) version is an inductive argument. It includes among its premises a statistical regularity. The explanandum thus cannot be deduced, but the premises confer a high degree of probability on it.
This is not the place to delve into a discussion of the relative merits and demerits of the CL-model. Rather, I wish to apply some of the features of the model to throw some light on the nature of explanatory narratives. On the surface of things it would appear that if the CL-model represents the logic of scientific explanation, the explanatory narratives are indeed not scientific since they make no use of or reference to laws or regularities. In fact, most narrativists explicitly make a virtue out of not dealing in generalizations, witness the quote from Bruner above. Yet they see themselves as (social) scientists. But the issue points to deeper and more interesting problems in the configuration of explanatory narratives.
To begin with, it should be observed the CL-model and the narrative “conception” of explanation share the same framework, namely that of an opposition between particular and universal or general. The narrativists reside at the particularity end of this opposition, whereas the case is slightly more complex for the CL-model, since both the CL-model and narratives explain particular events. But whereas the CL-model explains by subsumption, narratives explain by rendering a causal sequence leading up to the event.
Narrativists have not used much space to discuss problems that may be involved in their insistence that narratives deal with particular people and particular events at particular times. One might speculate – rather maliciously, perhaps – that they are too preoccupied distancing themselves from what they take to be defining features of “traditional” research to spell out what they take “particular” to mean. We touch here upon deep philosophical and methodological problems in narrative configuration. What, for instance, is the unit of investigation in narrative research? Is it the characters? The events? How should an event be conceived? Hempel discusses the status of an event in his essay The function of general laws in history (1959). Here the notion of a historical event is subsumed under a general conception of event that puts it on a par with e.g. physical events. Then the individual event-to-be-explained is placed in a direct relationship to a law or a regularity to produce a DN- or an IS-explanation respectively. With suitable initial conditions added, the event can be deduced or inductively inferred and thus explained.
Evidently a number of things could be said (and have been said) about Hempel’s application of the CL-model to historical explanation. For my purposes here, I shall simply observe that in order for this to work, the event in question must be repeatable; hence it possesses a degree of generality that may be unacceptable to the narrativist. But this is not clear, since narrativists do not specify what they mean by “particular”. Does it, for example, imply that events are unique in the sense that they never repeat themselves? Paul Ricoeur (1984), in his lengthy discussion of French historiography, observes that the rejection of the CL-model seemed to imply a return to the conception of an event as unique. He goes on to make a point that also seems highly pertinent to empirical educational narrative research:
This assertion [about uniqueness] is false if we attach to the idea of uniqueness the metaphysical thesis that the world is made up of radically dissimilar particulars. Explanation then becomes impossible. The assertion is true, though, if we mean that, in contrast to the practitioners of the nomological sciences, historians want to describe and explain what actually happened in all its concrete details. But then what historians understand by “unique” means that nothing exists exactly like their object of inquiry (1984, p.124).
It seems to me that narrativists adopt the second sense of “particular” cited above, despite occasional uses of the term “unique” in narrative texts. In fact, narrativists have recourse to and naturally use general and classificatory terms in their texts. For example, Mattingly speaks of “[Parkinson] patients”, “therapists”, “medication”, “ADL”, “husband” and “wife” in her story – all of them general, classificatory terms that allows the characters, events and actions in question to be placed in broad, general categories thus making a wide range of beliefs and knowledge applicable to the particular cases. Other narratives contain such words as e.g. “teaching” and “dialogue”, both events that evidently repeat themselves. This represents a “push” toward the generality end of the continuum, and a few comments are in order. First, the level of “uniqueness” or particularity is relative to the level of precision chosen by the narrative researcher. Second, and closely connected to choice of precision level, is the problem of choice of reference class (Salmon 1990). Mattingly’s student therapist has chosen very broad classificatory terms in her story, and the reference classes are nowhere mentioned. A fair assumption would be the class of all Parkinson patients. A narrower and more precise reference class would add to the particularity of the case, and it would also narrow the range of considerations that people automatically bring to bear, suggested by the classificatory terms used. If the reference class for the Parkinson patient in question was the class of “retired, physically clumsy businessmen suffering from Parkinson’s disease”, maybe the student even would have considered other interventions. Third, both Ricoeur and Salmon suggest that the need for explanations may arise from perceived differences between the case in question and those which are grouped under the classificatory term. Again, the choice of reference class affects the differences one perceives; hence, the explanation of differences one can give; hence, which narrative one produces about the case. It should be noted that the placement of something in a reference class in and of itself explains nothing.
The Relation Between Premises And Conclusion
It emerges from the above discussion that despite the push toward generality by the use of general, classificatory terms, narrative explanations do not explicitly formulate or use laws or regularities. In passing, though, it is worth noting that generalities of some sort seem to underlie the idea that explanations arise from perceived differences between a particular case and the cases usually grouped under the classificatory term. Particular differences stand out against an assumed background concerning how things are “in general”, “usually” or “normally”. Every explanatory narrative has recourse to such generalizations. However, the kind of premises or antecedents explicitly employed in narratives differs from the premises of the CL-model, as well as the relation of antecedents (beginning and middle) to the conclusion (end). On the CL-model, the premises support the conclusion with certainty or near-certainty. The premises of a DN-explanation may be seen as constituting conclusive evidence for the conclusion, whereas the premises of an IS-explanation provide strong evidence (provided the evidence is relevant). On this view, the relation between premises and conclusion – the explanatory relation – may be construed as evidentiary (Salmon 1990).
As already stated, the premises of a narrative are made up of the causes that lead up to the conclusion or end. Narratives thus constitute a form of causal explanation; events should follow from one another, nor just after one another. A number of problems arise here. Should there be restrictions on choice of antecedent conditions? Should we distinguish between necessary and sufficient causes, and if so, what are the implications? And finally, how should we construe the relation between causes as premises and the conclusion?
The first problem points to the role of hindsight discussed above. To the best of my knowledge, narrative theory places no restrictions on the choice of antecedents, and as a consequence this choice is subject to well-documented hindsight effects. For example, outcome knowledge dramatically increases the perceived likelihood of the outcome in question (Fischhoff 1975, 1988). In fact, with hindsight the outcome frequently comes to be viewed as inevitable; expressed as e.g. “it couldn’t have happened otherwise” or “I do not see what I could have done differently”. Outcome knowledge also changes the judged relevance of data describing the situations that precede the event in question; as is clearly shown in Carr’s baseball example. Retrospective judges effortlessly make sense of what they know about past events by constructing coherent wholes. Such acts of configuration are so natural that we are largely unaware of hindsight effects on re-description. Narrative explanations of how things came to be a susceptible to hindsight biases. In addition, it is a well-documented empirical finding that causal inferences (from perceived effect or outcome to alleged cause) are highly unreliable. In narratives, it is not just a matter of establishing one antecedent – that would be what Ricoeur calls a truncated explanation – but an entire causal chain. With no restrictions on antecedents and virtually no harnessing of inferences, chances are fairly great of picking an incorrect cause at each step. But the result may be a coherent narrative with a beginning, proceeding through a causal sequence and ending with a closure of the plot.
The second problem, of necessary and sufficient causes, is obviously connected to the retracing of causal chains, but is more interesting when the story is told forwards again to explain how things came to be – how the causal chain leads up to the end. The issue is discussed by Ricoeur, and I shall return to it in the subsequent section.
The third problem concerns the relation between premises and conclusion in narratives. Drawing on the work of W.B. Gallie, Ricoeur here introduces the notion of “followability”. To follow a story is to “… understand the successive actions, thoughts and feelings in the story inasmuch as they present a particular ‘directedness’” (1984, p.150). The orientation in a certain direction that we find here, Ricoeur says, amounts to recognition of a teleological function in the conclusion or end. But when a story is told forwards, the storyline must be followed up to the conclusion – in no way can the conclusion be deduced or predicted from the premises. It is unclear whether Ricoeur himself endorses this view, but Louis Mink certainly does. There can be no “detachable” conclusion in a historian’s work, he claims, because the narrative as a whole supports the conclusion. The end is an integral part of the narrative order. Even though hindsight frequently makes us believe that we and other people could and should have known the result in advance, the story must be followed to its end so the “directedness” can be made visible, explicit. The relation between premises and conclusion is internal; the two cannot be viewed separately from each other. The implications of this for the argumentative quality of explanatory narratives will be further explored in the next chapter.
According to Frederick Suppe (1989), explanations as arguments do not capture the structure of explanations. With the assimilation of explanations and theories into laws, he claims, goes a failure to appreciate how theory structure radically affects the nature of scientific explanation.
A full-blown account of Suppe’s view of explanation requires an account of his view of theories, but this is beyond the scope of this paper (readers are referred to Suppe 1989). However, his account includes some points that are highly pertinent to the present discussion. First, he agrees with Wesley Salmon that scientific explanation concerns explanation of classes of events. Explanations of particular events may have practical, but not scientific, value. The implication of this view for narratives is plain, but will not be pursued here.
Second, and more interesting, he enlarges the class of explanatory-seeking questions by including who, where, how, which – questions that are not translatable into why-questions. The kind of explanation one can have, is determined by the structure of the theory in question. Some theories yield explanations how, but not why. For example, theories with statistical laws of succession (law applies not to empirical phenomenon, but to replica or model of phenomenon), yields a how-could explanation in that is shows how the model may assume a number of different subsequent states, and assigns probabilities to each state. This type of explanation makes no use of such notions as statistical relevance (Salmon), maximum specificity (Hempel) or causal notions. The laws have the in-built temporal asymmetry required for explanation, without explicit recourse to causal notions. This temporal quality is essential to explanatory narratives. A why-explanation can be supplied by a theory with a deterministic law of succession. Such a theory would require a development “path” of unique subsequent states, such that there is only one state in which the model could end up. An explanation why automatically provides a how-could explanation. With a deterministic law, it also provides a how-did explanation. The how-did question seems to be the narrative question of how things came to be.
It is now time to bring Ricoeur back into the picture. As we have seen, he thinks that explanations are causal whereas Suppe points out that there is no simple connection between causes and the ability to provide why-explanations. On Ricoeur’s view, causal explanation occurs in two major forms; in terms of sufficient and necessary conditions. They provide different types of explanations; why-did and how-possible, respectively. So despite his brief suggestion that explanations are answers to why-questions, he claims that necessary conditions provide a kind of how-explanation. The sufficient condition relation, Ricoeur says, governs manipulation: we bring things about, or things come about. The necessary condition relation governs prevention: in setting aside x we prevent everything from happening for which x is a necessary condition (cause). And he exemplifies the explanatory-seeking questions these different relations are answers to:
We respond to the question “Why did such a state necessarily happen?” in terms of a sufficient condition. On the other hand, we respond to the question “How was it possible for such a state to occur?” in terms of a necessary, but not sufficient, condition (1984, p.135).
A few observations should be made. To begin with, his why-did question seems reminiscent of Suppe’s why-question, with a unique state as the conclusion or end. Ricoeur does not expand on this issue, so it is hard to say whether he also thinks that a why-did explanation requires a unique path leading up to the conclusion. Since the sufficient condition governs the “came about”, it may seem that this is what Mattingly is after. It should be noted, though, that the how-possible question is entirely consistent with Mattingly’s view of narratives as explanations how. He further complicates things by maintaining that explanations in terms of sufficient conditions allow for prediction. This evidently runs against the views of Louis Mink; and, it seems, many narrativists who wish to distance themselves from “traditional” science and its use of laws. Explanation in terms of necessary conditions do not allow for prediction, but rather for retrodiction, as we have seen. We might want to say that Ricoeur makes things very complicated by using the word “necessarily” in the question that he answers in terms of sufficient conditions. Mattingly makes no distinction between sufficient and necessary conditions, so we cannot tell what kind of answer she thinks will provide an explanatory narrative. It is a standard view among narrativists, however, that narratives are configurated in hindsight, that is, with retrodiction. Ricoeur does not tell us how a prediction may be possible or what it may look like, or whether it requires a unique path as well as a unique end-state. The question is interesting, also because it points to the problem of restrictions on antecedents raised above. Suppe maintains that in how-did explanations only knowledge of prior states and the theory may figure in the explanans. But Suppe’s account of explanation is not causal, and narratives are not theories, nor do they employ theories in any conscious or explicit manner, although they have recourse to generalizations, especially if the explanandum is a deviation or a difference from “what usually is”. The only safe conclusion at this point is that a lot of work is needed to endow narrative theory with a tenable account of explanation, whether causal or not.
Narratives, Explanations And Arguments
Before applying (selected parts of) argumentation theory to the nexus of problems described above, a few comments should be made. First, there is the problem of explanation-seeking questions. Traditionally, answers to why-questions are viewed as yielding explanatory knowledge, whereas how and what are viewed as yielding descriptive knowledge. On Suppe’s view, however, many types of questions are explanation-seeking. Two points emerge from this: that different types of questions, their inter-translatability and their presuppositions should be inquired into, and that the distinction between descriptive and explanatory knowledge may not be as clear cut as we want it to be. Second, Ricoeur claims that narratives are “self-explanatory” in the sense that the what and the why coincide. To narrate, to grasp things together, already is to explain. Thus, a narrator explains by using the process of emplotment. To describe what happened is also to explain why it happened (or perhaps how, depending on whether sufficient or necessary causes are employed). Third, Suppe’s account of theoretical explanation seems useful for narrativists because it accommodates a wide range of explanation- seeking questions. However, Suppe states that scientifically relevant explanations explain classes of events, not just particular events. If we accept this view, which he shares with Wesley Salmon, where does that leave the scientific status of narratives? Does it turn narratives into a form of data rendering? Fourth, both Suppe and Ricoeur make the point that explanation theory leads a life separated from scientific theories and narratives, respectively. But their views are not parallel. Suppe thinks it is unfortunate that the structure of explanation should be discussed in terms of the structure of theories. Ricoeur thinks that science (in this case history) has removed explanation from the fabric of narratives and set it up as a separate problematic. History passes from descriptive to explanatory, he says, when why is freed from what and becomes a separate inquiry. It is unclear where this leaves narratives in relation to history, since he also claims that narratives answer both what- and why-questions at the same time.
In the subsequent discussion two problems will be highlighted; the question of the relation between premises and conclusion and the question of believability. Both are, it seems to me, at the heart of the application of argumentation theory to explanatory narratives, and both have the potential both to disentangle and to confuse matters.
The Relation Between Premises And Conclusion Revisited
According to Wesley Salmon (1984) an argument is a group of statements standing in relation to each other. Among the basic terms are conclusion, premise, (causal) inference and evidence – terms also found in explanation theory and highly pertinent to any evaluation of the quality of narratives.
Argumentation is a complex, interdisciplinary phenomenon, much the same as narratives. Different views emphasize different functions and different properties of arguments, and presumably the relation between premises and conclusions may also be construed in different ways. I believe that the nature of this relation is the core issue in deciding whether narratives can be construed as arguments or not. Whereas Hempel focuses primarily on the status of the conclusion, a number of theorists argue that it is the relation between premises and conclusion that should be the main focus (e.g. Biro & Siegel 1992, Salmon 1984). In the previous chapter we saw that the relation between premises and conclusion in a narrative is internal in the sense that the conclusion is undetachable from the premises. The conclusion is a part of the causal chain, not an independent “result”.
I seriously doubt that the premise-conclusion relation found in narratives is similar to that found in arguments, although caution is needed here because argumentation theories may construe the relation differently. However, I shall provide some premises for my (tentative) conclusion. First, there is the “status” or “position” of the conclusion. Two things seem important here. There is the independence of the conclusion and the premises that one finds in arguments; the two are independently knowable. This is denied in narratives, where the conclusion is an integral part of the narrative as a whole and in no way independent of the premises. Then there is the highly important point that in narratives and explanations, the conclusion is known for a fact. It exists as something that happened before we can tell a story about it or explain it. Second, if we follow Biro and Siegel (1992), arguments (premises) provide reasons to accept the conclusion. They base their argumentation theory on the claim that “… it is a conceptual truth about arguments that their central (…) purpose is to provide a bridge from known truths or justified beliefs to as yet unknown (…) truths or as yet unjustified beliefs” (p.92). This point is closely connected to the foregoing. A narrative conclusion or an explanandum is not an as yet unknown truth or an as yet unjustified belief. It is known for a fact, because it is something that already took place. The relation of premises and conclusion in arguments is one of justification; it is a matter of warranting our belief in the conclusion. But in narratives there is no need to warrant our belief in the conclusion, since we already know it for a fact. The problem of when reasons (evidence, premises) are good enough to warrant belief in the conclusion does therefore not arise in narratives as it does in arguments. If the point of arguments is to show that knowing the premises warrants knowing the conclusion, and if this justificatory relationship of premises to conclusion is at the heart of the very definition of an argument, then I conclude that narratives are not arguments. We have to look elsewhere or take a different perspective to make use of insights from argumentation theory to narratives.
Believability is at the outset mainly connected to one of the functions of arguments; namely their ability to persuade or convince. Recall Jerome Bruner’s claim that the use of the narrative mode would lead to believable stories. Given that the conclusion (end, explanandum) is already known for a fact, the believability does evidently not concern the conclusion. Rather, I suggest – and Bruner does not say this – it concerns the narrative as a whole; the causal sequencing leading up to, culminating in or producing the conclusion.
One controversial issue immediately arises when the focus is shifted to believability; namely what kind of argumentation theory is involved. It is perhaps not entirely clear what is meant or implied when something, e.g. a narrative, is described as believable, but it is tempting to suggest that convincing the audience is at least partly involved. Biro and Siegel point out that if conviction is the main function of arguments, we arrive ultimately at a purely psychological theory of argumentation. Their own preferred conception is epistemic and places argumentation in a network of such epistemic concepts as knowledge, proof, evidence and rationality of beliefs. I will side-step this discussion, but I think that when unpacked the notion of believability of narratives exhibits both psychological and epistemological factors, and that their relative strength must be dealt with contextually. Believability is influenced both by various factors such as level of descriptive detail and opportunities for the audience to recognize their own experiences and views in the story as well as tight and truthful causal reasoning (Kvernbekk 2002).
Let us briefly look at a distinction made by Wesley Salmon (1990), between explanation-seeking and evidence-seeking why-questions. Salmon sees the two types of why-questions as a possible source of confusion between explanation and believability (although he does not use the term believability). A question about why an event happened is an explanation-seeking why-question, whereas a question about why we should believe something is an evidence-seeking why-question. We must take great care, he says, not to confuse explanatory facts with evidential facts. Offering an explanation for a fact is different from providing reasons for believing something is the case. Now, what we find offered in narratives are explanatory facts. These facts – largely in the form of a causal chain – do not comprise evidence for believing in the conclusion of the narrative, since we know this conclusion for a fact already. But evidence may play a different role for the believability of a narrative. As said in the previous chapter, narratives are configurated with hindsight. The hindsight position influences both choice of causes, inferences made and judgments of the relevance and relative significance of data. Considerations of evidence should clearly be made by the narrator as he or she makes causal inferences to explain how things came about. Ultimately the believability problem concerns the narrative as a whole, since we are invited to believe the whole story and not just a part of it. In fact, we are specially invited to believe in the antecedent causal chain rather than in the conclusion, which we already know. This could, I suppose, be broken down into evidence for each causal inference that the chain consists of. We know that a great degree of detail seems to increase the believability of stories. As the truth of narratives is concerned, this is a double-edged sword, since probability theory tells us that the probability of a causal chain as a whole decreases the longer the chain becomes. Properties that make a story believable may thus counteract properties that concern the truth and probability of the story. Given genre demands, the evidence rarely (if ever) appears in the finished narrative, and I am not certain how conscious empirical narrative researchers are of the possibilities of making mistakes, picking out wrong causes etc. Evidence may this not appear in the product, but it should somehow be there in the process. Presumably this issue also ties in with the problem of restriction on antecedents. As far as I know and understand what is generally called the narrative method, no such restrictions are even discussed.
This paper is a first stab at a huge nexus of problems tying together the concepts of narratives, explanations and arguments. It seems to me that this nexus is largely an unexplored area. Unsurprisingly, conclusions should remain tentative.
Many narrativists think of narratives as being explanations, but the absence of a theory of explanation in narrative research makes it hard to judge to judge the quality of proffered explanatory narratives. Believability, as suggested by Bruner, seems an unsatisfactory criterion for various reasons. The notion is unclear, and unless we have a clear picture of what makes a narrative believable, the criterion is difficult to apply. Furthermore, a narrative can be believable and yet false or badly configurated. It seems to me that believability tends toward the psychological, and if one thinks – as I do – that narratives also should be judged by epistemic criteria, believability in itself is not sufficient. This is a highly contentious matter among narrativists and their critics. But to return to explanation: part of such a theory for narratives, I believe, awaits developments in general explanation theory and in our knowledge of different types of questions and their presuppositions. Much narrative research proceeds on simple, implicit notions of explanation.
When argumentation theory is introduced, the picture gets even more complicated. In narratives, the premises consist of causes tied together in a chain that culminates in and produces the conclusion. The chain is described in hindsight, after the fact, and the relation between premises and conclusion is internal. The causes do not constitute evidence for the conclusion, and no evidence is needed since we know the conclusion for a fact. The configuration of a narrative always begins with the conclusion, and then inferences are made backwards from observed effect to alleged causes. A good, tight causal reasoning does not warrant belief in the conclusion, but it may warrant (or produce) belief in the narrative as a whole. It seems to me that narrativists pay little attention to evidence for the truth or adequacy of the premises, and argumentation theory may help focus this point which should be a major concern to narrativists who take an interest in the epistemic as well as the explanatory qualities of their narratives, over and above mere believability.
Aristotle (1982): The Poetics. New York: W. W. Norton. Trans. James Hutton.
Biro, J. & Siegel, H. (1992): Normativity, argumentation and an epistemic theory of fallacies. In F. H. van Eemeren, R. Grootendorst, J. A. Blair & C. A. Willard (Eds), Argumentation illuminated (85-103). Amsterdam: ISSA.
Bruner, J. (1985): Narrative and paradigmatic modes of thought. In E. Eisner (Ed), Learning and teaching the ways of knowing (97-115). Chicago: University of Chicago Press.
Carr, D. (1991): Time, narrative and history. Bloomington: Indiana University Press.
Carter, K. (1993): The place of story in the study of teaching and teacher education. Educational Researcher, 22, 1, 5-12
Casey, K. (1995/96): The new narrative research in education. In M. W. Apple (Ed), Review of Research in Education, American Educational Research Association, Vol. 21, 211-253.
Clandinin, D. J. & Connelly, F. M. (1991): Narrative and story in practice and research. In D.A. Schön (Ed), The reflective turn (258-281). New York: Teachers College Press.
Dray, W. (1957): Laws and explanations in history. Oxford: Oxford University Press.
Fischhoff, B. (1975): Hindsight – Foresight: The effect of outcome knowledge on judgment under uncertainty. Journal of Experimental Psychology: Human Perception and Performance, 1, 3, 288-299.
Fischhoff, B. (1988): For those condemned to study the past: Heuristics and biases in hindsight. In D. Kahneman, P. Slovic & A. Tversky (Eds), Judgment under uncertainty: Heuristics and biases (335-351). Cambridge: Cambridge University Press.
Gudmundsdottir, S. (1990): Curriculum stories: Four case studies of social studies teaching. In C. Day, M. Pope & P. Denicolo (Eds), Insights into teachers’ thinking and practice (107-119). London: The Falmer Press.
Hempel, C. G. (1959): The function of general laws in history. In P. Gardiner (Ed), Theories of history (344-356): New York: The Free Press.
Hempel, C. G. (1965): Aspects of scientific explanation and other essays in the philosophy of science. New York: The Free Press.
Hempel, C. G. (1966): Philosophy of natural science. Englewood Cliffs: Prentice-Hall.
Kvernbekk, T. (2002): On the alleged superiority and convincing powers of narratives. Paper presented at the Annual Meeting of the Philosophy of Education Society of Great Britain, Oxford, April 5-7.
Mattingly, C. (1991): Narrative reflections on practical actions: Two learning experiments in reflective storytelling. In D. A. Schön (Ed), The reflective turn (235-257). New York: Teachers College Press.
Mink, L. (1978): Narrative form as a cognitive instrument. In R.H. Canary & H. Kozicki (Eds), The writing of history (129-149). Madison: University of Wisconsin Press Ricoeur, P. (1984): Time and narrative, Vol. I. Chicago: University of Chicago Press.
Salmon, W. C. (1984): Logic. Englewood Cliffs: Prentice-Hall.
Salmon, W. C. (1990): Four decades of scientific explanation. Minneapolis: University of Minnesota Press.
Suppe, F. (1989): The semantic conception of theories and scientific realism. Urbana: University of Illinois Press.