«Narratives & Mechanisms Abstract Historical scientists are frequently concerned with narrative explanations targeting single cases. I show that two ...»
Narratives & Mechanisms
Historical scientists are frequently concerned with narrative explanations targeting single cases. I show
that two distinct explanatory strategies are employed in narratives, simple and complex. A simple
narrative has minimal causal detail and is embedded in a general regularity, whereas a complex
narrative is more detailed and not embedded. This distinction’s importance is illustrated in reference to
mechanistic explanation. I consider ‘liberal’ accounts of mechanistic explanation, which expand the traditional picture to accommodate less mechanistic sciences. Simple narratives warrant a mechanistic treatment, while some complex narratives do not.
Introduction Scientists examining the past are taken to be primarily concerned with narrative explanations which account for single events1. A meteor exterminated the dinosaurs; New Zealand’s lake Taupo was formed by an enormous volcanic eruption; the introduction of small-pox killed millions in the Americas. Of course, historical scientists are not narrowly concerned with narrative explanation. As Kosso (2001) and Jeffares (2008) discuss, they sometime target middle-range theories which connect contemporary phenomena to past events (see also Turner 2009). Moreover, much historical enquiry targets patterns and regularities in deep time. Paleobiological work covering the nature of mass extinction events (Raup
1991) the nature of speciation (Eldredge & Gould 1972), the role of selection and adaptationist explanations in macro-level patterns (Gould et al 1977, Huss 2009), are all concerned with regularities in life’s shape, not the explanation of a simple event. However, at least much of the time their explanatory interests are geared towards the particular rather than the general. This paper shows that historical explanation, understood as narrative, is disunified: at least two distinct explanatory strategies are employed. Simple narratives explain particular cases as instances of regularities – the explanandum is subsumed by a general model. Complex narratives do not account for explananda in terms of regularities or models.
I argue that simple narratives have more in common with the population-level explanations furnished by economists and ecologists than complex narratives. This is demonstrated by comparing narrative explanations with mechanistic models. Both population-level and simple narratives are amenable to For example, Kitcher 1993, Cleland 2011, Hempel 1965 and Hull 1975 appear to agree that historical enquiry is primarily narrative mechanistic gloss. However, in complex cases scientists are not typically mechanists. Faced with a complex world, they employ characteristically non-mechanistic explanations.
The paper is in three parts. In the first, two case-studies illustrate the distinction between simple and complex narratives. Part two discusses mechanistic explanation, sketching the view and introducing liberalism – the view that most or all scientific explanation is mechanistic. The third part examines narrative explanation in light of mechanistic explanation, arguing that simple narratives are characteristically mechanistic, while some complex narratives are not.
1. Narrative Explanations
Narrative explanations account for particular events2 via causal sequences concluding with the explanandum. The causal sequence makes the explanandum likely. Narrative explanations are taken to be distinctively historiographical (at least by Hempel and Hull) due to their ‘story-like’ structure and lack of appeal to laws. The treaties at the close of the First World War led inevitably to the Second; the extraterrestrial impact which caused the Chicxulub crater was sufficient to exterminate the dinosaurs;
and so on. There is more than one way to account for an event, however. Some causal sequences stand alone: even if only one extinction event was caused by an impact, we can be convinced of the impact’s causal sufficiency. Or we might explain an event as an instance of a general model: perhaps all wars have common causes, and the Second World War can be explained in terms of those commonalities.
I will be agnostic as to whether all narratives in fact reference regularities, and whether this is problematic. Hempel’s primary concern about historiographic explanation is the lack of nomological appeals and I (in part) share the suspicion that particular events can be satisfactorily explained without recourse to regularities (c.f Tucker 1998) but my claim of the disjunctive nature of narratives holds regardless of this.
Hopefully it is clear that narrative explanations are surely not restricted to historiographical inquiry – there is nothing stopping a chemist explaining a single event in terms of some causal sequence (perhaps even without explicit mention of laws) - and therefore the claims I make about narrative explanation will most likely not be restricted to the geological and paleontological cases I focus on. Whether the distinctions and lessons I draw are extendable to other sciences I leave for future work: given that I will speak in terms of past events, but historical enquiry also covers historical processes, entities and states of affairs. The claims made about events carry over to those other types of targets.
narrative explanation is paradigmatically the business of historical inquiry, it is the obvious place to center philosophical investigations And so narrative explanations (1) account for some particular explanandum in terms of some causal sequence; (2) may or may not appeal explicitly to laws or generalizations; (3) are paradigmatically, but not exclusively, historical. I argue that there are two explanatory strategies which historical scientists employ in providing narratives.
1.1 Snowball Earth
There were glaciers in the tropics at least twice during the Neoproterozoic (roughly 1000 – 542 million years ago). Towards the end of the period there was synchronous, ubiquitous glaciation: the entire earth covered in permafrost cut through by rivers of ice. This presents a series of geological and palaeoclimatological challenges. What could have caused this scenario? How did it thaw? Why are such events rare? The most popular explanation of these glacial events is Joseph Kirschvink’s Snowball Earth Theory (Schopf & Klein 1992, Hoffman & Schrag 2002).
The late Neoproterozoic was a time of continental dispersal: the supercontinent Rodinia broke up and the megacontinent Gondwana began to form. During glacial periods most continents clustered at the middle and lower latitudes. Kirshvink proposed that this clustering was responsible for the global freeze.
Both land and ice-caps have high albedo – they reflect more of the sun’s energy than water. Tropical landmasses have high albedo because more sunlight reaches the equator. Their warm, moist climate also increases silicate weathering (the absorption of C02). Land clustering around the tropics, then, increases albedo and decreases greenhouse gases. This would lower the earth’s temperature –
particularly at the poles where the growth of ice sheets would lead to a freezing feedback loop:
If more than about half of the Earth’s surface area were to become ice covered, the albedo feedback would be unstoppable… surface temperatures would plummet, and pack ice would quickly envelope the tropical oceans (Hoffman & Schrag pp 135)
The explanation can be presented in a simple flowchart:
Landmass clustering in the tropics lowers temperature by increasing albedo and thinning the atmosphere. Lower temperatures increase icepack cover, creating a feedback loop between lowering temperatures, larger icecaps, and higher albedo. Earth freezes over. As we shall see, Snowball Earth is a paradigm ‘simple’ narrative: an event is explained by a general model with reference to minimal causal factors.
1.2 Sauropod Gigantism
Despite public perception, most dinosaurs fit comfortably in the familiar mammalian size-range. The sauropods were different: not merely big, but puzzlingly so. Some were the largest land animals to have ever lived: Sauroposeidon and Argentinosaurus are estimated to have weighed between 50 and 70 tons, rivaling baleen whales in length. By contrast, the largest known terrestrial mammal was Paraceratherium, thought to be 12 meters long and weighing 20 tons at most. How did sauropods manage such sizes? Why was it unique? How was gigantism physiologically and evolutionarily possible?
As Sander, Christian et al (2011) review, sauropod gigantism was the result of myriad causes (see also Klein et al 2011). Sauropods were the right lineage, in the right place, at the right time. They had specific primitive characteristics which removed size limitations. Early sauropods were oviparous – egg-laying allows for fast population recovery, mitigating the small population size engendered by gigantism. They did not masticate, increasing food intake. They had a distinctive small-head-and-long-neck morphological structure, which maximizes grazing range while minimizing movement.
These primitive characteristics were supplemented by new adaptations. Gigantism itself protected against the increasingly sophisticated predators of the Jurassic, and accommodated the enormous digestive system mitigating the lack of mastication and gastric mill. Their basal metabolic rate increased to accommodate the speedy growth required. Sauropods evolved a distinctive pneumatized skeleton, a signal of a bird-like respiratory system, which increases the efficiency of oxygen dispersal and accommodates the growth rate required to reach gigantic size.
The road to gigantism was open to sauropods due to their distinctive primitive characteristics. The road was followed due to the evolution of particular adaptations in response to particular evolutionary pressures. The explanation of sauropod gigantism is a complex narrative: there is no appeal to a general model in explanation, but rather a unique, detailed causal sequence is employed.
1.3 Simple & Complex Narratives
In explaining snowball earth and sauropod gigantism historical scientists follow two distinct explanatory strategies. Both are narrative explanations: their explananda are individual cases, accounted for via particular causal sequences. However, snowball earth is explained as an extreme case of a general model. Sauropod gigantism is not. Moreover, the Snowball Earth contains less causal detail than sauropod gigantism. The geological case is simple, while the paleobiological case is complex3. Two features, an explanation’s detail and embeddedness, are characteristic of simple and complex narratives.
It is worth reiterating that these distinctions may well illuminate sciences not typically considered historical, or dealing with narratives. It is beyond this paper’s scope to discuss such cases, but I take it that if simple and complex explanations of particular events occur in ahistorical sciences, this only strengthens the importance of the distinction.
Detail A striking difference between the two explanations is the level of detail required. Detail is a measure of the specificity, complexity and diffusion of the explanans required for explanatory adequacy. Snowball The distinction between complex and simple is similar in spirit to ‘actual sequence’ and ‘robust process’ explanations (Sterelny 1996, Jackson & Pettit 1992), although is not cashed out in overtly modal terms.
earth is low- detail: few factors and a single difference-maker are required. General facts about global albedo, temperature, atmosphere and icepack work in tandem with particular facts about landmass clustering to produce the explanandum. Sauropod gigantism, by contrast, requires a more detailed explanation. Adequacy requires many explanans, quite disparate in nature. Important explanatory details are spread through time: from deeply primitive characteristics such as oviparity, to highly derived ones like pneumatization. Explanans are also spread across grain: oviparity is important because it mitigates evolutionary, population-level concerns while pneumatization solves individual-level, physiological concerns.
Detail, then, tracks the complexity required for explanatory adequacy, and its nature depends in part on the explanandum. In the snowball earth case, the world cooperates in granting sufficiency to low detail explanations while for sauropod gigantism the distended, messy nature of the explanandum demands a more detailed, messy explanation.
A narrative explanation is embedded when the explanandum is accounted for as a token of a type of process; an instance of a regularity. The relative simplicity of the snowball earth explanation allows it to be represented by a single climatological model. The hypothesis is an extreme case of run of the mill dynamics between ice cover, geography, climate and atmosphere. In explaining why the earth froze, I tell you about those general dynamics and how the scenario would arise given particular states of affairs. Sauropod gigantism, by contrast, is an exquisite corpse: birds provide a model for respiratory systems; giraffes, swans and structural morphology tell us something about possible sauropod stances;
elephants and large lizards about possible metabolism. There is no single unifying regularity which can be appealed to. In explaining gigantism, I refer to particular facts about the sauropod lineage and the environment in which it evolved.
I have mentioned that some philosophers take narrative explanations as problematic insofar as they do not appeal to regularities, and that I will not take a stance on this here. With embeddedness on the table, I can clarify this. Clearly embedded explanations appeal to regularities: the interesting question is whether non-embedded explanations do, or must. I am inclined to see non-embedded explanations as leaning on a patchwork of regularities. For instance, models of structural morphology, population genetics and metabolism are all appealed to in explanations of sauropod gigantism. However, it is open for others to argue that such appeals are not always required.
Embeddedness, then, tells us whether an explanandum is accounted for as an instance of a general model, or as an individual event.