Concepts as Epistemic Tools: A Comparative Approach

Recent work in the philosophy of scientific concepts has examined their productive roles in investigative practice (e.g., Feest & Steinle 2012). For example, rather than merely serving as vessels for previously accumulated information, concepts play a role in the empirical individuation of the phenomenon of interest, allowing for further generation of knowledge. This entails that concepts must be formed before scientists have a complete grasp on the entities in question (see, e.g., discussion in Feest 2010). This paper discusses the generation and use of such preliminary concepts in light of recent work on concepts and similarity in cognitive science.
Specifically, we ask: when does a new phenomenon warrant the formation of a new scientific concept, rather than an expansion of an existing one? Moreover, how does such a new category, which may lump together phenomena that turn out to be heterogenous at a more causally fundamental level of description, lead to the development of later, more homogenous categories? We address these questions by examining the formation and evolution of the concept "slow infection," which paved the way to the later concept "prion."
The category of "slow infections" was proposed in 1954 by Björn Sigurdsson. While Sigurdsson believed these diseases were caused by viruses, he suggested that they be distinguished from acute and chronic types of infections. Specifically, in forming this concept, he focused on attributes such as a long incubation period and a regular, predictable constellation of symptoms leading to death. These features had been known-and were associated with individual infectious diseases such as scrapie-long before Sigurdsson's proposal. What prompted him to suggest the formation of a new category?
Drawing on Rosch's work on similarity and conceptual taxonomy and on more recent work on the role of alignment in similarity comparisons and categorization, we discuss the interactive process that gives rise to the formation of a new category. Specifically, it is not enough that a phenomenon displays some anomalous features; rather, these features need to placed in correspondence with features of familiar phenomena that serve as contrast classes. The alignment, in turn, allows for further grasp of commonalities among the features of the members of the new categories.
Finally, we discuss the way in which the preliminary category of "slow infection" paved the way to narrower, more homogenic categories and the development of the "prion" concept. The paper thus draws on insight in cognitive science to shed light on the role of concepts in supporting their own evolution as part of what Chang (2004) terms "epistemic iteration."

Renormalization has many faces. It's a tool for removing infinities from quantum field theories; it's a constraint on which theories we should take seriously; it's a means for precisely characterizing the separation of scales we observe in physical systems; it's a coarse graining procedure; it's a window into new physics at scales that we have limited experimental access to. Over all the varied technical uses of renormalization, one concept is almost always seen at work in the background: the Wilsonian picture. Kenneth Wilson's work on critical phenomena is described as having "had enormous influence on almost all fields of science" (Lebowitz, 1995) and as being a primary source of the "effective field theory philosophy" (Zee, 2003). Inspired by Wilson's move, physicists are now in the business of constructing effective theories with highly constrained domains of applicability that isolate the relevant physics from the irrelevant. Yet much confusion about the conceptual details and the technical implementation of renormalization remains many decades after the advent of the Wilsonian picture. 
I argue that, with several concepts falling under the name of "renormalization," Wilsonian renormalization has played the role of a phantom concept. That is, even in contexts where the Wilsonian renormalization group is absent, the concept of Wilsonian renormalization is either explicitly invoked or taken for granted. It acts as a structuring force on the form and content of how we think about doing physics and how we renormalize theories. I aim to show that this role of Wilsonian renormalization follows as a consequence of a tension between the physical content that the Wilsonian picture provides and the intractability (or computational expense) of carrying out the Wilsonian calculations. In Wilsonian renormalization's place other renormalization concepts (such as the continuum renormalization group) are used, but Wilsonian ideas remain due to the need for its physical content as a justification of renormalization methods. Close scrutiny of Wilsonian renormalization as a phantom concept should be fruitful for understanding the role of concepts and conceptual confusion in scientific practice and in the dominant effective theory paradigm of modern physics. 

'Species' is both a central concept in biological science and a source of endless controversy. Currently, biologists work with an array of 'species' concepts, none of which has proven to be adequate to all the purposes for which biologists use the term. Philosophical responses to this situation vary, ranging from a continued search for a single unifying concept (e.g., de Queiroz's metapopulation lineage species concept), to an endorsement of some form of conceptual pluralism, to the suggestion (by Marc Ereshefsky, most prominently) that the concept be altogether eliminated: there are multiple disparate concepts hiding under the auspices of a single term, and little unity to be found among them.
This paper takes Ereshefsky's arguments for eliminative pluralism as a springboard to develop a novel account of the unity and function of the 'species' concept. Specifically, we argue that Ereshefsky's case for eliminative pluralism turns on a crucial assumption: that the primary purpose of the term 'species' is to delineate a certain kind of unit in nature, and that therefore there is a single 'species' concept only if there is a single kind of unit to which all 'species' concepts refer. Granting this assumption, Ereshefsky's arguments for eliminative pluralism succeed: there is not a single kind of unit captured by the myriad uses of the term 'species'. Both forms of life and biologists' interests in studying these forms of life are too diverse to allow for a single, all-purpose 'species' concept.
Our aim, however, is to challenge this assumption. Our argument involves two central claims. First, we argue that, while the term 'species' is sometimes used to delineate a certain kind of natural unit, that is not its only use. We show that the concept has important roles to play in managing inquiry into process of "speciation" (or diversification)-processes that can operate without necessarily resulting in the production of species (this is common in microbial domains especially). This allows us to relax the assumption that the only way in which the 'species' concept could be unified is by picking out a particular sort of unit. This leads, in turn, to our second core claim: that 'species' sub-concepts (e.g., the ecological and biological species concepts) have a patchwork structure (in the sense developed by Mark Wilson in the context of the physical sciences). This patchwork structure is the result of 'species' concepts being applied to domains of life for which they were not initially developed, in which different speciation processes dominate. We show how these patchwork structures interact in productive ways in the study of speciation, helping biologists to draw connections and transfer resources (e.g., methods and styles of reasoning) between domains. In this sense, there is a single overarching 'species' concept, albeit one with a highly complex internal structure.

I propose a novel model for analyzing synchronic and diachronic aspects of conceptual change. Some scientific concepts are essentially dynamic. Change and variation in their use follows naturally from the complexity of the systems to which they apply and from the epistemic complexity of science itself. Scientists initially adopt such concepts because they identify a phenomenon worth studying and consolidate promising epistemic tools, including methods and theories. However, newly formed concepts neither fully explain their target phenomenon nor fully prescribe the course of subsequent research. In order to understand the behavior of such concepts over time, it is necessary to examine the development of the research programs they support.
My model identifies processes responsible for cohesion as well as diversification of essentially dynamic concepts. It takes inspiration from W. B. Gallie's proposal that under certain conditions, conflict over the meaning of socially relevant concepts (e.g. art, democracy) becomes essential to their proper use. I show that scientific concepts exhibit analogous features, but argue that change (rather than conflict) is their most characteristic result. I focus on a critical but vague component in Gallie's original proposal-the role of exemplary achievements in anchoring the development of dynamic concepts. Many successful scientific concepts can be understood as complex epistemic achievements that set the stage for future research. They integrate multiple epistemic tools in order to make progress in understanding natural systems that extend across multiple scales of space and time. Conceptual change naturally results as scientists pursue the open-ended project of developing the original achievement. 
I demonstrate the power of this model by applying it to the case of homology, a biological concept that is widely acknowledged as central but notoriously hard to define and apply. The research program of Richard Owen (1804–1892) played a formative role as exemplar for this concept. Owen is famous for providing an influential early definition of 'homology' but I show that his epistemic achievement was far more substantial. It also included the articulation of criteria for applying the definition, the identification of key homology phenomena, the specification of nomenclature for describing these phenomena, and the development of theoretical framework for explaining them. Given this much fuller picture of Owen's research program, we can see how it provided cohesion to later streams of homology research while also encouraging their diversification.
My analysis challenges the common assumption that different contemporary versions of the homology concept only hang together if there is a single kind of metaphysical entity, or a single epistemic structure, in which they all participate. Instead, the variants of an essentially dynamic concept are loosely bound by their common historical relationship to one another, and to the exemplary research program. More generally, the processes identified in my model are likely to be prevalent across many fields of science. Therefore it offers inspiration for the open question of how to account for the widespread semantic variation of key concepts.