Knowledge is generally associated with an ability to provide an explanation. In other words, if we know something we can express that knowing in terms of explaining it. For example, if I claim to know what is responsible for a disease I might offer an explanation in terms of a bacterium as an etiological agent. An explanation, traditionally, is an argument that provides a reason to account for a phenomenon, event, or action. It makes clear or intelligible what is obscure or unknown. It is more than just a simple description or statement of an event or a listing of its features or characteristics that answer questions of how or what. Rather, an explanation is an answer to a question of why. For example, an explanation may offer a cogent answer as to why cancer affects one segment of a population but not another. In explaining an event, an attempt is made to get behind or underneath it, in order to provide knowledge or an understanding of the event that can often be used to control or manipulate it.
There are several types of explanations (Ladyman, 2002). A historical explanation provides the antecedent events responsible for a past event. For example, one nation may cite previous grievances for invading another country. A psychological explanation accounts for organismic behavior. For example, a person may exhibit violent behavior because of being abused earlier as a child. An astrological explanation relies on the alignment of stars and constellations to account for an event or action. For example, a person may exhibit specific behavior patterns because of their astrological sign. A theological explanation invokes divine action. For example, a person may suffer from a disease because the divine is punishing that person for a trespass or sin. A teleological explanation invokes purpose to account for an event. According to Aristotle, for example, fire rises because its natural location is up or away from the center of the earth.
Although these various explanations are commonly used in everyday discourse, they are not as precise in terms of their explanatory power as a scientific explanation. Explanations in the natural sciences and hence in the biomedical sciences are intimately associated with natural laws and scientific theories. Theories generally function to enable us to explain natural phenomena, with universal and statistical laws providing the basis for them.' Theories, then, under gird powerful explanatory systems that can account for natural phenomena and events through the application of natural laws. By enabling explanatory systems, they capture the way the world is and reveal the mechanisms underlying the observable and unobservable dimensions of it. The explanatory power enabled by theories generally depends on the power and precision of the laws used to account for natural phenomena. Besides laws, scientific explanations as enabled by theories may also rely on other factors such as identifiable causes and functions. For example, diabetes can be explained as the absence of insulin.
In this chapter various explanatory schemes are examined, especially in terms of scientific and medical knowledge. Beginning with covering law explanation that dominated philosophical understanding of explanation for decades, the discussion turns to explanatory schemes proposed to resolve the problems associated with the covering law model. The first and most powerful explanatory scheme is causal explanation, especially Paul Thagard's causal network instantiation scheme for explaining disease. Kenneth Schaffner champions a "six component" explanatory scheme for the biomedical model that incorporates many of the preceding explanatory schemes, especially the causal mechanistic approach. Inference to the best explanation and functional explanation also represent two important explanatory schemes that are relevant for the natural sciences.' And, these explanatory schemes are utilized in the biomedical model to varying degrees of success to account for medical phenomena. Finally, although humanistic or humane practitioners appreciate the power of these explanatory schemes they often utilize a narrative explanatory scheme that includes the patient's personal knowledge, especially in terms of the patient's illness story.
8.1 Covering Law Explanations
In the late 1940s, Carl Hempel and Paul Oppenheim introduced an explanatory model using the laws of science. They claimed that explanations are like arguments in which the conclusion or explanandum, a statement used to describe the phenomenon to be explained, is logically deduced from the premises or explanans, those statements used to do the explaining. The explanans consist of at least one scientific law and of the initial or antecedent conditions. For example, why a patient's cardiac output (CO) is 4.91 is explained by the initial conditions of a stroke volume (SV) of 70ml and of a heart rate (HR) of 70 beats per minute and by the law: CO = SV x HR. Since laws are an integral component of the Hempel and Oppenheim's explanatory model, it eventually became known as the "covering-law model" (Dray, 1954, 1957). In other words, what is to be explained is covered or accounted for by a scientific law.
Hempel and Oppenheim (1948) provided the following scheme for their explanatory model:
The explanans appears above the line and consists of the initial conditions (C) and the laws (L) while the explanandum (E) appears below the line and consists of the explanation. Because the relationship between the explanans and the explanandum is deductive and the laws are universal or deterministic generalizations, the scheme is generally called a "deductive-nomological" (DN) explanation (Hempel, 1965).
Although DN explanations can be used to account for many natural phenomena in which the natural laws are universal or deterministic generalizations, not all phenomena can or could be explained using such laws. Some phenomena are better explained using probabilistic or statistical laws or generalizations. Moreover, the relationship between the explanans and the explanandum is not deductive but inductive. Consequently, Hempel (1965) called this scheme "inductive-statistical" (IS) explanation. For example, whether a person contracts lung cancer from smoking a pack of cigarettes a day for 20-30 years depends upon many factors and can only be assigned a probability. In other words, contingencies exist that may prohibit a person from contracting lung cancer from cigarette smoke-contingencies that are often unexplainable in deterministic terms.
Although the covering law model of explanation was influential, several problems arose that signaled its demise, especially for application to disciplines such as biology and medicine. First, these disciplines have few natural laws that are universal or even statistical generalizations and yet they do provide adequate explanations of biological phenomena (Thompson, 1989).3 Next, there are many counterexamples that are explanatory but do not conform to covering law explanation or that conform to covering law explanation but are not explanatory (Okasha, 2002).
Sylvain Bromberger (1966) identified another problem in terms of the symmetrical relationship between the explanans and the explanandum. For some phenomena, this symmetry does not hold. For example, Bromberger conceded that the covering law model explains a pendulum's period in terms of its length and the law of simple periodic motion; however, the pendulum's length can be accounted for by explanations other than the pendulum's period and the law of simple periodic motion. Wesley Salmon (1971) also pointed out the problem of relevance. For example, Salmon noted that although a person takes birth control pills to avoid pregnancy and even though these pills are relatively effective in preventing pregnancy, if the person is a male then the explanation, although satisfying the requirements of the covering law model, is irrelevant.'
8.2 Causal Explanation
Many commentators on covering law explanation account for its problems, especially in terms of symmetry and relevance, in terms of Hempel's antirealist position towards unobservable objects and his shunning the notion of mechanistic causation (Okasha, 2002). However, the notions of causation and explanation are traditionally connected. Aristotle's four causes are often presented as explanations. Thus, an explanation involves citing the various causes responsible for a phenomenon under consideration. For example, Aristotle (2001) explained the existence of a bed in terms of its material cause, wood, its formal cause, the bed's shape, its efficient cause, the bed's maker, and its final cause, for sleeping. During the scientific revolution, as noted in Chapter 2, the Aristotelian causes were reduced to two: material and efficient causation. Scientists adopted the notion of material causation to explain natural phenomenon, while philosophers continued to debate efficient causation to determine its precise nature for explaining events.
Causal explanation traditionally depends on the regularity of sequential events, either temporally or spatially. An event is considered the cause of another event if it precedes the caused event and is connected with it in a regular and consistent fashion. As such then the event is explained in terms of its antecedent cause. For example, if an organism is exposed consistently to a bacterium before developing a specific disease then the bacterium is said to cause the disease and serves as the principal etiological agent for explaining the disease. For causal explanation, the cause may be either necessary, i.e. the absence of the cause guarantees the absence of the effect, or sufficient, i.e. the presence of the cause guarantees the presence of the effect, or both. In other words, causal explanation traditionally assumes an ontological relationship between the cause and its effect. Finally, simple or singular causation is generally inadequate for explaining the nature of many natural phenomena since the events underlying them are complex and involve multiple interacting antecedent events. In these cases, causal explanation cannot be captured by simply stating a single cause but only by dissecting the causal matrix.
As noted earlier, Hume's critique of causation revolves around the notion of "constant conjunction" Since he denied any knowledge of causation is possible other than the regularity of causal sequences, his notion of causation became known as the regularity view (Beebee, 2006). Causal explanation is nothing more than describing the association of events, with no ontological basis for the association. In other words, there is no substantive causal structure that underlies a causal explanation. So, for example, to explain that a disease is caused by a bacterium requires only a regular concurrence of the bacteria's presence and the disease's expression. But this minimalist criterion for causal explanation fails to capture the extent to which causal explanations allow scientists both to develop theoretical accounts of phenomena and to manipulate them based on those accounts.
To address the apparent problem associated with the regularity scheme of causal explanation, Salmon (1984) proposed an alternative scheme called "causal mechanical" explanation. In this scheme, he claimed that causation refers to actual causal mechanisms and that an explanation depends upon explicating those mechanisms. His scheme, along with others similar to it, like Phil Dowe's theory (2000), is called process explanation, since the relationship between cause and effect involves the existence of specific sequential interactions or processes.
An important component of Salmon's scheme is the notion of causal interaction, by which two causal processes intersect spatially and temporally. The result of this interaction is a modification of or change in the properties or features of one or both of the causal processes. For example, an infectious disease is explained mechanically or mechanistically in terms of the intersection of the bacterial and organism. The issue, however, is how to distinguish the explanatorily relevant processes from those that are irrelevant, especially for complex systems like biological organisms.
James Woodward (2003) introduces a manipulability causal scheme of explanation to address this issue of relevance. The scheme is predicated on the variability of both cause and effect in terms of their values or properties, such that one event causes another if and only if the values or properties of the effect are altered upon intervention of the cause. Importantly, the other variables within a causal nexus must be held constant in order to determine the causal contribution of an antecedent event to an effect. Experimental trials best represent this scheme, in which scientists manipulate the causes and observe the changes in the effects. To that end, Woodward illustrates his causal scheme with randomized, controlled clinical trials. As discussed in chapter 10, these types of trials allow biomedical and clinical scientists to control for placebo and other effects not pertinent to a drug's action in order to determine its efficacy.
Medical knowledge according to the biomedical model is generally explained in terms of mechanistic causes, which represent the objective reasons for medical phenomena-whether disease or therapy. Physicians are interested in explanations that involve only the physical causes or mechanistic entities and forces responsible for the patient's disease and recovery. Just as scientists explain natural phenomena in terms of material components and mechanisms, so biomedical clinicians explain disease phenomena in terms of material entities and mechanisms. For example, biomedical clinicians reduce the cause of cancer to the mutated forms of around a half-dozen genes (Hanahan and Weinberg, 2000). Once the causal mechanism is identified, treatment or therapy, then, is generally based on chemical or physical intervention, either in the form of a pharmaceutical drug or a surgical procedure.
Thagard (1999) proposes a causal network instantiation explanatory scheme that combines both epidemiological and biological research. His explanatory scheme is a system of causal interactions, in which correlations, alternative causes, and mechanisms, along with conditional causal probabilities, factor into explaining why a person contracts a disease. "Explanation of why people get a particular disease," for Thagard, "begins by noticing associations between the disease and possible causal factors" (1999, p. 101). In order for these associations to count as correlations that lead to causation, the association must be statistical or probabilistic. The probability is more than just conditional but also causal, in that the probability is a measure of the disease causing the effect rather than simply being associated with the effect. Such a probability measures causal power. In addition, causal power must also take into consideration alternative possible causes of the disease. Elimination of these causes enhances causal power. Finally, knowledge of underlying mechanisms supports a causal relationship but is not necessary to infer it.
Thagard (1999) uses a duodenal ulcer case study to illustrate the causal network instantiation explanatory scheme. He begins with a patient who is taking a nonsteroidal anti-inflammatory drug such as aspirin for arthritic pain. Overuse of the drug can lead to increase acid secretion and rapid gastric emptying among other effects that expose the gastric lining to possible injury. In addition, genetic predisposition to any of these effects may exacerbate the patient's condition. Besides these factors, environmental conditions such as stress or cigarette smoking may contribute to increased acid secretion and gastric emptying. The patient is now susceptible to Helicobacter pylori infection, which Barry Marshall demonstrated can cause gastritis, duodenitis, and ultimately a duodenal ulcer. Importantly, Thagard's explanatory scheme attempts to account for complex disease processes in which no single cause is responsible for the disease.
Schaffner (1993) also introduces a multi-component explanatory scheme that welds causal mechanistic features with a variety of explanatory schemes proposed by others. According to Schaffner, "the Hempel model of scientific explanation and Salmon's earlier S-R [Statistical Relevance] account suffered from defects that an appeal to causation could remedy" (1993, p. 262). To that end he develops an explanatory scheme composed of six components, by explicating the metaphysical, epistemological, and logical elements of causation. The first is a semantic component that consists of a series of generalizations composing the biomedical system (BMS). The next is a causal component, which can be either deterministic or probabilistic. The third is a unificatory component, in which domains of a BMS are unified. The next is a logical component, in which the explanandum is a causal conclusion from a set of premises. The fifth is a comparative evaluation inductive component, in which the Bayesian inductive support of an explanandum is compared to and evaluated with that of others. The final is an ideal explanatory text background component, in which an explanation is selected from a range of "ideal" explanatory text background through pragmatic concerns.
Schaffner provides an example of the explanatory power of this scheme, especially the causal component, with short-term and long-term memory learning in the sea hare, Aplysia californicum. For short-term memory learning, a stimulus, such as an electrical shock, to the head or tail triggers the retraction of the gill and siphon into the mantle. The duration of the retraction can be lengthened by sensitizing the hare through a regime of shocks prior to a test shock. The molecular mechanism for this sensitization involves the release of a neurotransmitter from a facilitating interneuron that closes the potassium channels in the presynaptic membrane, through a cyclic AMP-protein kinase cascade. The result is an increased calcium flow in the presynaptic membrane, with an increased release of the neurotransmitter from the presynaptic bulb and, hence, longer retraction times. The mechanism for long-term memory learning is analogous to short-term memory learning but includes the regulation of genes and additional complexity such as "parallel processing."
Schaffner discusses the significance of this neurophysiological example for his "six component" explanatory scheme, especially in terms of the causal component. According to Schaffner, general laws equivalent to those found in the physical sciences are not constitutive of the biomedical sciences. "What we appear to have," claims Schaffner, "are rather intricate systems to which apply both broad and narrow causal generalizations that are typically framed not in purely biochemical terminology but rather in terminology that is characteristically interlevel and interfield" (1993, p. 285). In other words, a BMS is a complex system of interacting levels of varying scopes. For example, the sea hare's siphon-gill retraction behavior is explained in terms of macromolecules (neurotransmitters) interacting with cells (neurons), which in turn interact with tissues (muscle). The result is a complex interlevel system of causal generalizations that are idealized in schematic form. Importantly, "the explanations that are characteristically biological (as well as biomedical) will be `causal/ mechanical' more frequently than not" (Schaffner, 1993, p. 296).
8.3 Inference to the Best Explanations
In the mid 1960s, Gilbert Harman introduced an explanatory scheme called "inference to the best explanation" (IBE).5 According to Harman, "one infers, from the premise that a given hypothesis would provide a `better' explanation of the evidence than would any other hypothesis, to the conclusion that the given hypothesis is true" (1965, p. 89). In other words, to claim that one hypothesis explains a phenomenon more adequately than competing hypotheses and therefore is true the competing hypotheses must be eliminated as being inadequate to explain the phenomenon. The elimination of competing hypotheses is based on the totality of available evidence: "Now, in practice we always know more about a situation than that all observed A's are B's, and before we make the inference [that all A's are B's], it is good inductive practice for us to consider the total evidence" (Harman, 1965, p. 90). Thus, only after considering the burden of all the evidence is one warranted to choose one hypothesis as a better explanation of a phenomenon than its competitors.'
Philosophers of science, especially of the realist stripe, use IBE to account for scientific explanation and theory confirmation and claim that IBE is the preferred means by which scientists formulate theories and hypotheses about the world that capture its reality. For example, in discussing Darwin's theory of evolution, Richard Miller insists that Darwin's concern was "whether the best available account of the data, however vague or incomplete, entails the superiority of the natural selection hypothesis over its current rivals" (1987, p. 165).
Other philosophers of science are less than sanguine about IBE. For example, IBE is not deductive, according to critics, but ampliative. Consequently, the best explanation could eventually be demonstrated as false. In other words, IBE suffers from the disadvantages of the induction problem. The best explanation may simply be the best of a "bad lot" of explanations (van Fraassen, 1989). The proponents of IBE assume a privileged position with respect to determining the best explanation and do not develop adequate criteria to establish that the best explanation is present in the competing set of explanations.'
Certainly the main issue with IBE is the development of a set of criteria that establishes one hypothesis as the best while the others are eliminated as not being the best. "By what criteria," demands Thagard, "is one hypothesis judged to provide a better explanation than another hypothesis?" (1978, p. 76). Although Harman (1965) acknowledges this problem, he does not discuss it other than to list several features of a good explanation, including simplicity, plausibility, explains more, and less ad hocness. Subsequent analysis by Harman and others, according to Thagard (1978), is insufficient to provide a commonly accepted set of criteria.
Thagard draws upon several scientific case studies in which a theory choice is made among competitors, to identify three criteria for determining the best explanatory hypothesis.' The first is consilience, which is "a measure of how much a theory explains, so that we can use it to tell when one theory explains more of the evidence than another theory" (Thagard, 1978, p. 79). The next criterion is simplicity. Here the better explanation is less complicated or more economical and does not employ ad hoc modifications to account for additional experimental and observational evidence, especially evidence that disconfirms the preferred theory. The final criterion is analogy, in which the best explanation is analogous to other similar explanations. For example, Darwin's notion of natural selection shared similarities with the widely accepted explanatory account of artificial selection.
Alexander Bird (1998) provides several additional criteria for determining the best explanation, besides the ones listed above. The first is that the explanation may supply a precise mechanism that accounts for how a phenomenon works. As noted earlier, mechanisms are important for explaining how a phenomenon operates. Another important feature of a good explanation is its generality, akin to Thagard's consilience. In other words the hypothetical explanation has the ability to incorporate or unify a number of disparate facts and observations. A final feature is coherence. The best theory is the one that has "the ability to integrate or combine with other explanations" (Bird, 1998, p. 89). Certainly no one feature is adequate to account for the best explanation, but the explanation that exhibits the majority of these features is the most likely candidate for being the best. For example, Howard Temin's DNA provirus hypothesis was not accepted until the discovery of reverse transcriptase, which provided a mechanism for retrovirus replication, even though the hypothesis was not the most simple and certainly did not cohere with the central dogma of molecular biology (Marcum, 2002).
Peter Lipton (2004) proposes the most comprehensive notion of IBE to date. He focuses on contrastive explanations, in which there is a contrastive difference between the best explanation and its competitors. In other words, Lipton is concerned not with the question simply why this explanation is best but with the question why this explanation is best and another is not. He bases his notion on what he calls the "Difference Condition." This condition states that there is a causal difference between the acceptance of one hypothetical explanation and the rejection of another, i.e. the best explanation contains a causal factor absent from the other explanations. In other words, the best explanation has a causal contrastive edge over other explanations.
Lipton's version of IBE shares certain similarities with John Stuart Mill's "Method of Difference." According to Mill, a causal agent can be identified as the difference between two situations. In other words, if a person comes down with food poisoning and was the only one among a group of people who ate lobster bisque, with everything else equal, then the lobster bisque explains why the person came down with food poisoning. Lipton claims that his version of causal contrastive explanation accounts for inferring unobserved causes and the selection of one hypothetical explanation in the face of multiple differences.9
The method of IBE, according to Lipton, involves two steps. First, there is the generation of potential inferential hypotheses to explain a phenomenon. Only a limited number of hypotheses can be generated, because of an "epistemic filter" that selects only for the plausible explanations. Lipton is not so much concerned that the best explanation accounts for all the evidence, although this is important, but that it has a contrastive edge over the others. That edge is often obtained when the best explanation accounts for novel predictions, while its competitors do not. The second step then involves the selection of the best explanation. Selection of the best explanation depends on its "loveliness" and not necessarily on its "likeliness." The loveliest explanation is the one that provides the "most understanding," while the likeliest explanation is the "most warranted" (2004, p. 59). What makes one explanation lovelier than another is its explanatory virtues of simplicity and unifying power, and elucidation of causal mechanisms.
Lipton illustrates his version of IBE with Ignaz Semmelweis' research during 1844 to 1848 to explain an increased mortality of woman from childbed fever in one maternity ward of the hospital as compared to another. At first Semmelweis considered hypotheses based on current notions of "epidemic influences" and other plausible hypotheses concerning diet or general care, but rejected them for one reason or another. Comparison of the two wards, however, revealed that medical students attended patients in the high mortality ward. Semmelweis proposed several hypotheses to account for this observation, such as medical students' rougher handling of patients. Again, he eliminated these hypotheses. The chance occurrence of a colleague's death from an illness similar to childbed fever, after puncturing himself during an autopsy, led Semmelweis to hypothesize that medical students are contaminating birthing mothers with cadaver material. Simply having the medical students wash their hands before examining patients reduced the mortality rate to that of the midwife ward.
According to Lipton, the Semmelweis case study is a "gold mine" for supporting IBE. "By tailoring his explanatory interests (and his observational and experimental procedures) to contrasts that would help to discriminate between competing hypotheses," argues Lipton, "Semmelweis was able to judge which hypothesis would provide the best overall explanation of the wide variety of contrasts (and absences of contrasts) he observes, and so to judge which hypothesis he ought to infer" (2004, p. 81).
8.4 Functional Explanations
According to Larry Wright (1973), function is an ambiguous term with a "spectrum of meanings." However, function is usually defined philosophically in teleological terms as an activity or action that fulfills a specific goal or purpose. Hence, function is a performance concept oriented to the execution of an objective. Berent Enc symbolizes functional sentences accordingly: "the function of X is to do Y' (1979, p. 344).1° For example, the function of the heart (X) is to pump blood (Y). In this example, the heart's purpose or goal is to circulate blood throughout the body by pumping it. Moreover, the heart's structure is such that it contributes to its function as a pump. Functional statements, such as the heart pumps blood, raise additional questions, such as why does the heart pump the blood or how does the heart pump blood, and such questions require an explanation.
"Functional explanation," according to Huib de Jong, "often takes the form of decomposition of complex systems. This consists," he continues, "in describing a system in terms of what it does, and then explaining its behavior in terms of what it is" (2003, p. 292). In other words, a particular function is explained in terms of its structure. For the heart pumping blood example, the heart circulates blood because it is a pump. The how question is answered by detailing the structure of the heart and its muscular and nervous composition. Enc provides another formulation for functional explanations: "X does S in order to (so as to) do Y' (1979, p. 344). For the example of the heart pumping blood, the heart (X) pumps blood (S) in order to feed the body's tissues (Y)." It is the "in order to" that serves as the linguistic feature to answer the why question.
Functional explanations are particularly common in biology and psychology, especially given the complexity of biological organisms and their behavior. Ernest Nagel (1977) identified four types of function that inform functional explanations in these disciplines. The first is a teleologically neutral function, in which the notion of function has no connotation of purposeful action. Rather, function is expressed simply in terms of "biological role" and represents a property of an organism given its structure. Explanations based on this notion of function are expressed in terms of the structure-function relationship and are akin to those found in physics and chemistry. For example, the kidney filters the blood to remove metabolic waste (biological role) because of the glomerulus' capillary fenestrations (structure).
The second type is "selective agency" function, in which activities "are directed by purposive agents toward achieving selected ends" (Nagel, 1977, p. 280). This type of function is based on an analogy between human and nonhuman activity or behavior. In other words, a particular activity or behavior is selected because it performs a particular function in the organism's economy. This represents a "metaphorical extension" from human, conscious functioning to nonhuman, unconscious functioning. Explanations based on selective agency have the same pattern as those for conscious ones. Thus, a function is selected in an organism "for the sake of' the effects that function carries out for the organism. For example, the filtering function of the kidney was selected in vertebrates for the sake of removing metabolic wastes from the organism's blood.
The third type is "heuristic" function, in which function is perceived "as if' it were a product of design. According to its proponents, "a process cannot properly be characterized as purposive, if it can be explained on the basis of physicochemical laws, and that the effect of an organic process can count as one of its biological functions only if that process was intended or designed to produce the stated effect" (Nagel, 1977, p. 290). In other words, the ascription of function to nonhuman organisms is not to be taken literally but as a "regulative" principle or maxim in guiding research. Functional explanations represent statements about the particular activity of an organ or organism in terms of design function. For example, the kidney functions to filter blood of metabolic waste as if designed to accomplish this function.
Because each of the preceding types of function has a fatal defect that renders the explanatory scheme based on it suspect, Nagel championed a fourth type called "goalsupporting" or "welfare" function, in which "functional statements not only presuppose that the systems under discussion are goal-directed, but also that the function ascribed to an item contributes to the realization or maintenance of some goal for which the system is directly organized" (1977, p. 296). This type of function lends itself to a more general explanation that incorporates the overall goal or advantage of the function vis-a-vis the organism's flourishing. For example, the function of an organism's kidney within an environment of limited diffusion capacity is to remove metabolic wastes, so that the organism can maintain blood chemistry conducive to life.
Alex Rosenberg challenges the notion of functional explanations in biology. "The apparent generalizations of functional biology," according to Rosenberg, "are really spatio-temporally restricted statements about trends and the co-occurrence of finite sets of events, states and processes" (2001a, p. 148). In other words, functional explanations do not represent natural laws but rather descriptions that are contingent upon local conditions and the Darwinian law of natural selection.12 For example, the functional explanation traditionally given for the buckeye butterfly's eyespots is that they detract potential predators. However, the "functional individuation of biological kinds reflects the vagaries and vicissitudes of natural selection, since biological kinds are the result of selection over variation in order to solve design problems set by the environment" (Rosenberg, 2001a, p. 148).
Marc Lange (2004) takes issue with Rosenberg and claims that functional explanations cannot be simply reduced to local contingencies and the natural selection law. For example, he cites that "medicine does not take human evolutionary history as a variable" when explaining why a patient who smoked died of lung cancer (2004, p. 107).
8.5 Narrative Explanation
Compared to the logical, scientific explanatory schemes, the notion of narrative explanation seems problematic. "Our common sense notions of narrative and explanation," according to Jon-K Adams, "are so far apart that they appear incompatible: narrative tells what happened; explanation makes plain or comprehensible" (1996, p. 110). However, narrative represents a primitive or basic form of explanation compared either to commonsense or scientific intuitions. Through stories, events are structured cohesively so as to covey meaning, purpose, significance, and understanding to them. In other words, the events are made intelligible. Narrative often represents a powerful way by which to answer why questions, especially why an event has occurred, through the purposeful and intentional configuring of preceding events by a narrator such that the event to be explained appears to be a natural consequence of the narrative.
Based on the connection of preceding events with the event to be explained, Adams configures the structure of narrative explanation in Hempelian terms of an explanandum, the event to be explained, and the explanans, the sequence of events that precede the explanandum. "The logic of narrative explanation," claims Adams, "lies in the assumption that a sequence of events explains a single event by leading up to it" (1996, p. 110). Thus, the narrator assembles the events that precede the event to be explained in order to bring about an understanding of it. Importantly, explanatory stories can be told not only by an individual narrator but also by a society in which its individual narrators reside. Indeed, the stories that society often narrates are important in terms of defining its members and accounting for their positions and functions within it. Often these stories can be healthful, but they can also be destructive and harmful-not only to the health of a society's members, but also to the society as a whole.
Narrative explanations are especially important in history. Whereas scientific explanations are abstract in nature, in which a story's historical details are bracketed, narrative explanations take into account full historical details. Without these details, narrative explanations are sterile and fail to make adequate sense of the events being told or examined. According to the historian Paul Roth, narrative explanations supply "an account of the linkages among the events as a process leading to the outcome one seeks to explain" (1988, p. 1). An issue surfaces as to what constitutes the connection among the events to be explained. For scientific explanations, that connection or linkage is supplied by the invocation of universal natural laws. However, historians traditionally do not invoke universal laws.
In an attempt to shore up the soundness of historical explanations, Hempel proposed an explanatory scheme for historical events that includes general laws or universal hypotheses. In contrast to "the method of empathetic understanding," in which a historian "imagines himself in the place of the persons involved in the events which he wants to explain," Hempel proposed a logical structure for historical explanation analogous to that for explanation in the natural sciences that include general laws and initial conditions (1942, p. 44). Although Hempel's proposal did influence some historians, others objected to it. For example, William Dray argued that "in history, the demand for explanation is very often interpreted in such a way that the proper answer assumes narrative form" (1954, p. 17). At issue for Dray is the possibility of the event's occurrence and not its necessity.
Roth (1988) addresses two general objections, especially in terms of positivism, to narrative explanation in history. The first is methodological in nature and claims that narrative explanation is concerned with particular not universal events and thereby cannot invoke laws to justify or legitimate the explanation. Roth believes that this criticism is "misguided" and fails to account for other means of explicating explanatory schemes. Rather, he challenges this criticism and argues that narrative explanation could provide an alternative to the standard scheme if "enough formal properties of narrative accounts [are discovered] to establish how such explanations are viable candidates for objective evaluation" (1988, p. 4).
The second objection is epistemological in nature and concerns "how to verify a narrative" (Roth, 1988, p. 2). Roth claims the objection that narrative explanations are not verifiable in the traditional sense and cannot distinguish between fiction and nonfiction is founded on a correspondence theory of historical knowledge as true. Although he acknowledges that such knowledge is constrained by the facts, he challenges whether categories of truth or falsity are appropriate for evaluating narrative explanations. Roth argues that the notion of an objective narrative is not a coherent notion since "there are no ideal events to chronicle" (1988, p. 8). Rather, an objective, ideal account of history implodes from the fact that the human narrator's perspective cannot be excluded from the narrative.
Since truth is an inappropriate category for assessing narrative explanations, the question arises as how best to assess them. David Velleman (2003) proposes that goodness of the story is the relevant criterion. What makes for a good story or narrative is its ability to organize the various seemingly disconnected events into an intelligible whole. This involves practical reasoning, which connects what one can do with what one understands. The reasons supplied by practical reason vis-a-vis narrative operate explanatorily by setting the context, especially an emotional rather than a causal context. According to Velleman, understanding often occurs at a visceral or bodily level. A good story then taps into an emotional understanding that gives narrative its explanatory power. However, the emotional understanding achieved through narrative is not in contrast to causal understanding but is complementary to it, in an attempt to understand events both emotionally and causally at the same time.13
Mark Bevir (2000) also defends narrative explanation from positivistic criticism, which attempts to assimilate historical explanations into scientific explanations. The general scheme of narrative explanation consists of relating beliefs to proattitudes: "an action X was done because the agent held beliefs Y according to which doing X would fulfill his pro-attitude Z" (Bevir, 2000, p. 13). Two types of connecting relate beliefs to pro-attitudes in narrative explanations. This first is a conditional connecting, which "relate agents" beliefs and pro-attitudes to one another so as to make sense of the fact that they thought an action would fulfill one or more of their pro-attitudes" (Bevir, 2000, p. 14). This connecting is not causally necessary but it is also not arbitrary in that a theme or idea of a principal actor based on certain beliefs and pro-attitudes does preside in the historical event which the historian can identify. The second connecting is volitional, which "enables us to make sense of the fact that agents moved from having pro-attitudes to states of affairs to intending to perform actions and then on to acting as they did" (Bevir, 2000, p. 15). This connecting depends not on a reduction of behavior of mental and volitional states to brain states but to folk psychology.
Historical narrative explanations are distinct from fiction, because historians can deliver the facts (Bevir, 2000). In other words, historians can offer epistemically legitimate narratives that unpack themes revolving around beliefs, actions, and pro-attitudes because facts are not simply given through pure perception but are always embedded in prior concepts supplied by folk psychology. Even scientific explanations must rely on facts that are embedded in prior concepts supplied by what consensus declares are reasonable theories and concepts. "A rejection of native positivism," according to Bevir, "implies that the past does not present itself to historians as a series of isolated facts upon which they then impose a narrative so as to bring the facts to order. Rather," he continues, "the past, like all experience, presents itself as an already structured set of facts" (2000, p. 18). In other words, a historical event already exhibits a narrative structure.
Narrative explanation has also been used in humanistic approaches to clinical medicine. "For all the science that underpins clinical practice," observes Glyn Elwyn and Richard Gwyn, "practitioners and patients make sense of the world [of illness] by way of stories" (1999, p. 186). The question is "whether fitting symptomatic behaviors into a life-story adds to the understanding gained by fitting them into diagnostic categories" (Velleman, 2003, p. 1). For many humanistic practitioners, the answer to that question is a resounding yes. The patient's personal and historical information or story is imperative for a full understanding and explanation of the patient's illness experience, in order to make an accurate diagnosis and to provide an effective therapy.
Whereas explanations within the biomedical model are in terms of abstractions, within the humanistic or humane models explanations are made in terms of instantiating the abstract with historical and personal details. Moreover, the scope of explanations for the biomedical model is the physical (even the mental is reduced to the physical) while for the humanistic models it must include the nonreductive mental and the social and even the spiritual. The humane clinician's task, then, is to obtain the patient's story or narrative of the illness experience, in order to explain it fully.
Explanatory schemes are an important component of medicine in general, but particularly to the biomedical model. For the biomedical practitioner these schemes represent a variety of approaches to account for the intelligibility of medical phenomena, which are dependent upon the explanatory schemes developed for the natural sciences. Objective or brute facts and theories are critical for explaining a phenomenon, such as a patient's disease state.
Although biomedical explanatory schemes are important for designing intelligent and effective therapeutic modalities, they have fueled the quality-of-care crisis. Patients feel that biomedical practitioners are only interested in diagnosing the physical disease and prescribing a treatment plan to cure it, but they do not feel that physicians are concerned about the existential impact of the disease upon their lives. Humane practitioners champion narrative explanation to address this complaint. Through narrative, the physician can access information not only of the patient's existential dimension of the illness experience but also information about the unfolding of the disease state. Utilizing narrative explanation, humane practitioners search for a comprehensive account of the patient's illness in order to bring wholeness back to the patient's life. This is particularly important if the patient is suffering from a chronic or terminal illness.