Scientific Speculation

I will begin my talk with the three views about the purpose, legitimacy, and constraints on, speculation formulated in the General Description of this proposal: the Newtonian position ("don't speculate"), the hypothetico-deductive view ("speculate without constraints, but soon test"), and the Feyerabendian view ("speculate like mad even if you can't test your speculation"). James Clerk Maxwell, one of the giants of physics, rejected all three views, in favor of a very different one. The first he regarded as stifling science. He characterized speculations as (potential) "advances being made on unknown ground," but advances nevertheless. They initiate and develop possible ways to explain known laws and regularities. In his ground-breaking 1860 paper on molecular theory, which he explicitly calls a speculation, he introduces the speculative assumption that gases are composed of spherical molecules that collide elastically with each other, exert only contact forces, and obey Newtonian mechanics. From these assumptions and others he derives and explains various known gas laws. What he is trying to show is that a mechanical theory is possible, not that the particular one he offers is true or probable. Thus he avoids Newton's objection to speculators who say, mistakenly, that because their speculation yields known laws it is true or probable.
Maxwell's view about speculation was also incompatible with the hypothetico-deductive and Feyerabendian views, both of which reject any constraints on a speculation, considered as such. Maxwell insisted on constraints, the most important of which are not universal or a priori but local, empirical, and pragmatic, making it possible to evaluate the speculation before testing it experimentally. In 1875, in a later paper on molecular theory, he employs what he calls a "method of physical speculation." A major constraint he imposes on speculative theories explaining gas laws (that relate properties such as pressure, temperature, and volume) is that such theories be mechanical ones involving bodies in motion obeying Newtonian mechanics. His defense for doing so includes the idea that mechanical theories in other areas have been successful, and that according to experiments of Joule and others, heat is a form of energy and hence motion, and not a fluid. The speculation he develops can then be given a defense an important part of which is empirical and local, and part pragmatic (a point I will develop).
In my talk I will defend the view that speculating, in ways such as Maxwell did in the 19th century about molecules, or some contemporary physicists do in string theory, is, or can be, a legitimate and potentially important scientific task, even in the absence of testing or the prospect thereof. And it is possible to evaluate such speculations in significant ways even if those speculations lack experimental confirmation.

The Darwinian Tree of Life Project: Heuristic and Aesthetic Roles for Speculation
Richard A. Richards
In an entry from his early notebooks Darwin speculated that the arrangement of species could be represented by a simple branching diagram. And then in the only diagram in his 1859 On the Origin of Species, he represented evolution by natural selection with an elaborate branching and diverging tree. Since then, tree thinking has been ubiquitous in phylogenetics - the reconstruction of the evolutionary past. We see it in the cladograms and trees that have been constructed for narrowly defined groups of species, but also in the great "Tree of Life" project dedicated to reconstructing the comprehensive tree of all living species on earth from a single origin - beginning with the initial branching of the Eubacteria, Archaea and Eukaryotes. (http://tolweb.org/tree/) But as those who are involved in this project recognize, this great Tree of Life is highly speculative. There may be good evidence for some of the smaller branches at the crown, and perhaps even some of the branches below the crown, but many of the branches, especially at the base of the tree, are highly speculative. The branching order of Eubacteria, Archaea and Eukaryotes, for instance, is mere guesswork. Moreover, since the connections are genetic, there are good reasons to believe that the branching, non-reticulating structure of the Tree of Life misrepresents the actual history of life, which is full of reticulation, and may be better represented by either a network - when we look at introgression and horizontal gene transfer, or a forest of trees - when we look at the many inconsistent gene trees.
So what is the value of this speculative, great Tree of Life project? Given that much of it seems to be unsupported by adequate evidence and it arguably conflicts with what is taken to be established about evolutionary processes, the Tree of Life doesn't have indisputable epistemic or theoretical value. There instead seem to be two related kinds of value to this Tree of Life project. The first is cognitive and heuristic, and this goes back to Darwin's original speculation in his notebooks. It is a way to represent in an easily processed manner an arrangement of species. This function is seen in the non-evolutionary tree diagrams that preceded Darwin's Origin. The second value seems to be narrative and aesthetic. The Tree of Life presents a narrative about the evolution of life that can be represented in aesthetically pleasing ways and that can engage our imaginations. In effect, it is an intellectually and aesthetically satisfying, yet speculative and misleading story about the great history of life. What is notable here is that this function of speculation doesn't require the eventual testing or confirmation of the speculation. And while there may be empirical constraints about how the Tree of Life can be represented, what is important to its use is not just empirical but psychological. This might be seen as extending Maxwell's views about the pragmatic value of speculation. 

Abstract: A central aspect of Newton's methodology is his sharp distinction, in his fourth rule for appropriate reasoning in natural philosophy, between what he regarded as propositions appropriately gathered from phenomena by induction and mere contrary hypotheses which carry no weight to undercut the provisional acceptance of such propositions. This paper will investigate the extent to which the rather negative attitude towards speculation suggested by this treatment of hypotheses is already evident in Newton's famous 1672 response to questions raised by Pardies about his investigation of light and colors. This response includes the following quote,
For hypotheses should be subservient only in explaining the properties of things, but not assumed in determining them; unless so far as they may furnish experiments,
which, though agreeing with the main point of Rule 4, suggests a positive role for hypotheses in finding experiments. This paper will also consider the even more famous hypotheses non fingo passage from the general scholium added to the second edition of his Principia. This passage includes the classic remark,
For whatever is not deduced from the phenomena must be called a hypothesis; and hypotheses, whether metaphysical or physical, or based on occult qualities, or mechanical, have no place in experimental philosophy.,
which strongly reinforces the negative aspect of the treatment of hypotheses recommended by Newton's methodology. This paper will assess the extent to which the negative treatment of hypotheses suggested by this remark is supported by the actual practice of Newton's methodology in his argument for universal gravitation. The step from gravity as inverse square fields of attraction towards stars and planets to gravity as a universal force of pairwise attraction between bodies is one that Huygens and other mechanical philosophers regarded as mere speculation, because they saw no way it could be explained by interaction by contact between bodies. In sharp contrast with Huygens, Newton put aside the generally accepted commitment to forces by contact action to allow what we might call the speculative reasoning that led to his theory of gravity. The context of the hypotheses non fingo passage suggests Newton regarded the achieved and reasonably expected potential successful applications of that theory as sufficient to count it as appropriately deduced from phenomena. The illumination by these investigations will be further refined by investigating passages where Newton, explicitly, allows himself to speculate. This will include the important scholium to Proposition 69 of book 1 of his Principia, and Queries from his Optics. One feature supported by these investigations is that hypotheses as speculations that suggest potentially fruitful issues to be investigated empirically are, indeed, part of Newton's methodology. The important negative treatment is that of his fourth rule: that mere contrary hypotheses carry no weight to undercut acceptance of propositions appropriately gathered from phenomena. An alternative proposal realizing sufficient potential empirical success to be counted as a serious rival would not be a mere contrary hypothesis.

In 1970, atmospheric scientist Ed Lorenz remarked: "In the matter of determining the response of the
[earth's] climate to specified influences, we are still in the speculative era" (p.325). Lorenz argued
that computer models had the potential to move climate science beyond speculation by allowing for
the testing of these hypotheses. This was not an experimental form of testing, however – as
recommended by the hypothetico-deductive (H-D) method – but more of a consistency check:
computer models could be used to check the quantitative plausibility of climate change hypotheses,
which previously had been formulated in qualitative (or at best rough quantitative) terms. In the
decades since, this sort of testing has indeed emerged an important role for computer simulation
within climate science (and in the study of complex systems more generally). This talk will discuss
both Lorenz's vision for how computer simulation models, as 'bookkeeping devices', could aid the
testing of hypotheses about earth's climate, as well as how these models are being used in this way
today. With the help of computer simulation, it is fair to say that climate science has now moved
beyond the speculative era, though perhaps not quite as far as Lorenz might have hoped. 

Debates about the role of speculation in science are relevant to whether one kind of defense of speculative metaphysics can succeed. "Metaphysics is continuous with the natural sciences"---this thesis was a core part of Quine's revival of metaphysics in the middle of the 20th century. Faced with the positivist-adjacent complaint that we could never have evidence that justified one metaphysical theory over another, Quine held that scientific theories were similarly underdetermined, but that this did not entail skepticism; and that the method of preferring the simpler theory, which he claimed was the method actually used in scientific practice, both delivered justification and was the right method to use in metaphysics. These conclusions, though, were not based on detailed observation of scientific practice, and some have doubted that the kinds of simplicity arguments Quine used in metaphysics and that his heirs continue to use today are argument-forms that scientific practice actually endorses. But if a closer look at scientific practice provides some data that threatens a Quinean defense of metaphysics it may also provide other data that can be used to shore it up. A lot of metaphysics is speculation ("speculative metaphysics" is almost a pleonasm), and this is often thought to be a defect. Philosophers dreaming up theories which disagree only about matters extremely far removed from observational evidence---how could that possibly be part of a method of getting at the truth? But that it is part of a good method follows from the basic Quinean idea that the methods the sciences actually use are good ones for metaphysicians to use, as long as speculation plays an important role in scientific practice. Peter Achinstein has recently drawn attention to the fact that it does. I will argue that Feyeranbend's view, that "speculating like mad" to generate many new candidate theories, even when there is little prospect of testing them, is an important part of scientific practice, and that both scientific realists and scientific anti-realists should agree. Scientific realists should agree, because without the generating of many candidates realists have little reason to believe that a theory (approximately) true about unobservable phenomena is among the theories scientists are choosing between; in which case there is little reason to believe that scientists' best theories are (approximately) true. Scientific anti-realists should agree because without the generating of many candidates they cannot explain why science has hit upon any empirically successful theories at all.

The nature and status of speculation in the life sciences around 1800


This paper analyzes past scientific practitioners' views and discussions about the roles that speculations play, or should play, in scientific research. Philosophical-historical analysis of scientists' attitudes to methodological and epistemological issues typically focuses on major scientific individuals, such as Newton, Helmholtz, Darwin, Maxwell, or Einstein. My paper takes a different approach. Instead of analyzing the views of a prominent figure about whether it is legitimate to speculate or use hypotheses in science, I examine a debate among several investigators about a puzzling phenomenon – the colored spots that sometimes appeared on fresh milk and made the milk unusable for consumption. Because the pressure to control the phenomenon was high and experimental evidence remained conflicting and inconclusive, the investigations engendered a lot of discussion about methodological questions.


In the late 18th and throughout the 19th century, practitioners and researchers in different fields, including pathology (veterinary and human), organic and agricultural chemistry, and natural history tried to figure out what caused these blue spots and how their appearance could be prevented. There were abundant speculations about the phenomenon, its hidden causes and the mechanism of its generation. Experimental investigation proved difficult because the phenomenon was variable and elusive and could not be reliably produced in experimental settings. In my paper, I draw on this episode to discuss the practitioners' approaches to speculation. What kinds of speculations were deemed acceptable? Was it legitimate and acceptable to speculate about hidden causes and mechanisms that might generate this phenomenon, according to the practitioners? What were the conditions and constraints for doing so? The debate about blue spots on milk brought out the philosophical commitments of the practitioners particularly well.


The analysis of the historical sources, arguments, and attitudes shows that in the decades around 1800 there were different understandings and interpretations of "speculation" and different attitudes to hypothetical and speculative thinking. Perhaps not surprisingly, these attitudes do not neatly map on the philosophical positions about speculation that modern philosophers of science have advanced and defended. Moreover, and more importantly, the analysis reveals new aspects of speculative thinking that are still worth considering today. Drawing on the historical materials, I show that the evaluation of hypothetical or speculative thinking has important pragmatic and social dimensions. What is at stake is not only what counts as legitimate speculation, but also who could be trusted to offer credible hypotheses, and how carefully the methodological rules regarding speculation should be, and have been followed. I want to argue that these two issues – whom to trust and how carefully methodological rules should be applied – need to be an integral part of our discussions about speculative thinking (and indeed about scientific methodologies more generally).