Minding the gap: discovering the phenomenon of chemical transmission in the nervous system.

The neuron doctrine, according to which nerves consist of discontinuous neurons, presented investigators with the challenge of determining what activities occurred between them or between them and muscles. One group of researchers, dubbed the sparks, viewed the electrical current in one neuron as inducing a current in the next neuron or in muscles. For them there was no gap between the activities of neurons or neurons and muscles that required filling with a new type of activity. A competing group, the soups, came to argue for chemicals, subsequently referred to neurotransmitters, as carrying out the activities between neurons or between neurons and muscles. But even for them the conclusion that chemicals performed this activity was only arrived over time. I examine the prolonged period in which proponents of chemical transmission developed their account and challenged the sparks. My goal is to illuminate the epistemic processes that led to the discovery of a new scientific phenomenon-chemical transmission between neurons.

Using neurons to maintain autonomy: Learning from C. elegans.

Understanding how biological organisms are autonomous-maintain themselves far from equilibrium through their own activities-requires understanding how they regulate those activities. In multicellular animals, such control can be exercised either via endocrine signaling through the vasculature or via neurons. In C. elegans this control is exercised by a well-delineated relatively small but distributed nervous system that relies on both chemical and electric transmission of signals. This system provides resources to integrate information from multiple sources as needed to maintain the organism. Especially important for the exercise of neural control are neuromodulators, which we present as setting agendas for control through more traditional electrical signaling. To illustrate how the C. elegans nervous system integrates multiple sources of information in controlling activities important for autonomy, we focus on feeding behavior and responses to adverse conditions. We conclude by considering how a distributed nervous system without a centralized controller is nonetheless adequate for autonomy.

Discovering autoinhibition as a design principle for the control of biological mechanisms.

Autoinhibition is a design principle realized in many molecular mechanisms in biology. After explicating the notion of a design principle and showing that autoinhibition is such a principle, we focus on how researchers discovered instances of autoinhibition, using research establishing the autoinhibition of the molecular motors kinesin and dynein as our case study. Research on kinesin and dynein began in the fashion described in accounts of mechanistic explanation but, once the mechanisms had been discovered, researchers discovered that they exhibited a second phenomenon, autoinhibition. The discovery of autoinhibition not only reverses the pattern in terms of which philosophers have understood mechanism discovery but runs counter to the one phenomenon-one mechanism principle assumed to relate mechanisms and the phenomena they explain. The ubiquity of autoinhibition as a design principle, therefore, necessitates a philosophical understanding of mechanisms that recognizes how they can participate in more than one phenomenon. Since mechanisms with this design are released from autoinhibition only when they are acted on by control mechanisms, we advance a revised account of mechanisms that accommodates attribution of multiple phenomena to the same mechanism and distinguishes them from other processes that control them.

Reductionistic Explanations of Cognitive Information Processing: Bottoming Out in Neurochemistry.

A common motivation for engaging in reductionistic research is to ground explanations in the most basic processes operative in the mechanism responsible for the phenomenon to be explained. I argue for a different motivation-directing inquiry to the level of organization at which the components of a mechanism enable the work that results in the phenomenon. In the context of reductionistic accounts of cognitive information processing I argue that this requires going down to a level that is largely overlooked in these discussions, that of chemistry. In discussions of cognitive information processing, the brain is often viewed as essentially an electrical switching system and many theorists treat electrical switching as the level at which mechanistic explanations should bottom out. I argue, drawing on examples of peptidergic and monoaminergic neurons, that how information is processed is determined by the specific chemical reactions occurring in individual neurons. Accordingly, mechanistic explanations of cognitive information processing need to take into account the chemical reactions involved.

Figuring out what is happening: the discovery of two electrophysiological phenomena.

Research devoted to characterizing phenomena is underappreciated in philosophical accounts of scientific inquiry. This paper develops a diachronic analysis of research over 100 years that led to the recognition of two related electrophysiological phenomena, the membrane potential and the action potential. A diachronic perspective allows for reconciliation of two threads in philosophical discussions of phenomena-Hacking's treatment of phenomena as manifest in laboratory settings and Bogen and Woodward's construal of phenomena as regularities in the world. The diachronic analysis also reveals the epistemic tasks that contribute to establishing phenomena, including the development of appropriate investigative techniques and concepts for characterizing them.

Organization needs organization: Understanding integrated control in living organisms.

Organization figures centrally in the understanding of biological systems advanced by both new mechanists and proponents of the autonomy framework. The new mechanists focus on how components of mechanisms are organized to produce a phenomenon and emphasize productive continuity between these components. The autonomy framework focuses on how the components of a biological system are organized in such a way that they contribute to the maintenance of the organisms that produce them. In this paper we analyze and compare these two accounts of organization and argue that understanding biological organisms as cohesively integrated systems benefits from insights from both. To bring together the two accounts, we focus on the notions of control and regulation as bridge concepts. We start from a characterization of biological mechanisms in terms of constraints and focus on a specific type of mechanism, control mechanisms, that operate on other mechanisms on the basis of measurements of variables in the system and its environment. Control mechanisms are characterized by their own set of constraints that enable them to sense conditions, convey signals, and effect changes on constraints in the controlled mechanism. They thereby allow living organisms to adapt to internal and external variations and to coordinate their parts in such a manner as to maintain viability. Because living organisms contain a vast number of control mechanisms, a central challenge is to understand how they are themselves organized. With the support of examples from both unicellular and multicellular systems we argue that control mechanisms are organized heterarchically, and we discuss how this type of control architecture can, without invoking top-down and centralized forms of organizations, succeed in coordinating internal activities of organisms.

Grounding cognition: heterarchical control mechanisms in biology.

We advance an account that grounds cognition, specifically decision-making, in an activity all organisms as autonomous systems must perform to keep themselves viable-controlling their production mechanisms. Production mechanisms, as we characterize them, perform activities such as procuring resources from their environment, putting these resources to use to construct and repair the organism's body and moving through the environment. Given the variable nature of the environment and the continual degradation of the organism, these production mechanisms must be regulated by control mechanisms that select when a production is required and how it should be carried out. To operate on production mechanisms, control mechanisms need to procure information through measurement processes and evaluate possible actions. They are making decisions. In all organisms, these decisions are made by multiple different control mechanisms that are organized not hierarchically but heterarchically. In many cases, they employ internal models of features of the environment with which the organism must deal. Cognition, in the form of decision-making, is thus fundamental to living systems which must control their production mechanisms. This article is part of the theme issue 'Basal cognition: conceptual tools and the view from the single cell'.

Hierarchy and levels: analysing networks to study mechanisms in molecular biology.

Network representations are flat while mechanisms are organized into a hierarchy of levels, suggesting that the two are fundamentally opposed. I challenge this opposition by focusing on two aspects of the ways in which large-scale networks constructed from high-throughput data are analysed in systems biology: identifying clusters of nodes that operate as modules or mechanisms and using bio-ontologies such as gene ontology (GO) to annotate nodes with information about where entities appear in cells and the biological functions in which they participate. Of particular importance, GO organizes biological knowledge about cell components and functions hierarchically. I illustrate how this supports mechanistic interpretation of networks with two examples of network studies, one using epistatic interactions among genes to identify mechanisms and their parts and the other using deep learning to predict phenotypes. As illustrated in these examples, when network research draws upon hierarchical information such as provided by GO, the results not only can be interpreted mechanistically but provide new mechanistic knowledge. This article is part of the theme issue 'Unifying the essential concepts of biological networks: biological insights and philosophical foundations'.

From parts to mechanisms: research heuristics for addressing heterogeneity in cancer genetics.

A major approach to cancer research in the late twentieth century was to search for genes that, when altered, initiated the development of a cell into a cancerous state (oncogenes) or failed to stop this development (tumor suppressor genes). But as researchers acquired the capacity to sequence tumors and incorporated the resulting data into databases, it became apparent that for many tumors no genes were frequently altered and that the genes altered in different tumors in the same tissue type were often distinct. To address this heterogeneity problem, many researchers looked to a higher level of organization-to mechanisms in which gene products (proteins) participated. They proposed to reduce heterogeneity by recognizing that multiple gene alterations affect the same mechanism and that it is the altered mechanism that is responsible for the cell developing one or more hallmarks of cancer. I examine how mechanisms figure in this research and focus on two heuristics researchers use to integrate proteins into mechanisms, one focusing on pathways and one focusing on clusters in networks.

Sketching Biological Phenomena and Mechanisms.

In many fields of biology, both the phenomena to be explained and the mechanisms proposed to explain them are commonly presented in diagrams. Our interest is in how scientists construct such diagrams. Researchers begin with evidence, typically developed experimentally and presented in data graphs. To arrive at a robust diagram of the phenomenon or the mechanism, they must integrate a variety of data to construct a single, coherent representation. This process often begins as the researchers create a first sketch, and it continues over an extended period as they revise the sketch until they arrive at a diagram they find acceptable. We illustrate this process by examining the sketches developed in the course of two research projects directed at understanding the generation of circadian rhythms in cyanobacteria. One identified a new aspect of the phenomenon itself, whereas the other aimed to develop a new mechanistic account. In both cases, the research resulted in a paper in which the conclusion was presented in a diagram that the authors deemed adequate to convey it. These diagrams violate some of the normative "cognitive design principles" advanced by cognitive scientists as constraints on successful visual communication. We suggest that scientists' sketching is instead governed by norms of success that are broadly explanatory: conveying the phenomenon or mechanism.

Using computational models to discover and understand mechanisms.

Areas of biology such as cell and molecular biology have been dominated by research directed at constructing mechanistic explanations that identify parts and operations that when organized appropriately are responsible for the various phenomena they investigate. Increasingly the mechanisms hypothesized involve non-sequential organization of non-linear operations and so exceed the ability of researchers to mentally rehearse their behavior. Accordingly, scientists rely on tools of computational modeling and dynamical systems theory in advancing dynamic mechanistic explanations. Using circadian rhythm research as an exemplar, this paper explores the variety of roles computational modeling is playing. They serve not just to determine whether the mechanism will produce the desired behavior, but in the discovery process of hypothesizing mechanisms and in understanding why proposed mechanisms behave as they do.

Mechanists Must be Holists Too! Perspectives from Circadian Biology.

The pursuit of mechanistic explanations in biology has produced a great deal of knowledge about the parts, operations, and organization of mechanisms taken to be responsible for biological phenomena. Holist critics have often raised important criticisms of proposed mechanistic explanations, but until recently holists have not had alternative research strategies through which to advance explanations. This paper argues both that the results of mechanistic strategies has forced mechanists to confront ways in which whole systems affect their components and that new representational and modeling strategies are providing tools for understanding these effects of whole systems upon components. Drawing from research on the mechanism responsible for circadian rhythms in mammals, I develop two examples in which mechanistic analysis is being integrated into a more holist perspective: research revealing intercellular integration of circadian mechanisms with those involved in cell metabolism and research revealing that stable␣rhythms are dependent on how individual cells in the suprachiasmatic nucleus synchronize with each other to generate regular rhythms. Tools such as network diagramming and computational modeling are providing means to integrate mechanistic models into accounts of whole systems.

Circadian Rhythms and Mood Disorders: Are the Phenomena and Mechanisms Causally Related?

This paper reviews some of the compelling evidence of disrupted circadian rhythms in individuals with mood disorders (major depressive disorder, seasonal affective disorder, and bipolar disorder) and that treatments such as bright light, designed to alter circadian rhythms, are effective in treating these disorders. Neurotransmitters in brain regions implicated in mood regulation exhibit circadian rhythms. A mouse model originally employed to identify a circadian gene has proven a potent model for mania. While this evidence is suggestive of an etiological role for altered circadian rhythms in mood disorders, it is compatible with other explanations, including that disrupted circadian rhythms and mood disorders are effects of a common cause and that genes and proteins implicated in both simply have pleiotropic effects. In light of this, the paper advances a proposal as to what evidence would be needed to establish a direct causal link between disruption of circadian rhythms and mood disorders.

Can mechanistic explanation be reconciled with scale-free constitution and dynamics?

This paper considers two objections to explanations that appeal to mechanisms to explain biological phenomena. Marom argues that the time-scale on which many phenomena occur is scale-free. There is also reason to suspect that the network of interacting entities is scale-free. The result is that mechanisms do not have well-delineated boundaries in nature. I argue that bounded mechanisms should be viewed as entities scientists posit in advancing scientific hypotheses. In positing such entities, scientists idealize. Such idealizations can be highly productive in developing and improving scientific explanations even if the hypothesized mechanisms never precisely correspond to bounded entities in nature. Mechanistic explanations can be reconciled with scale-free constitution and dynamics even if mechanisms as bounded entities don't exist.

The Non-Redundant Contributions of Marr's Three Levels of Analysis for Explaining Information-Processing Mechanisms.

Are all three of Marr's levels needed? Should they be kept distinct? We argue for the distinct contributions and methodologies of each level of analysis. It is important to maintain them because they provide three different perspectives required to understand mechanisms, especially information-processing mechanisms. The computational perspective provides an understanding of how a mechanism functions in broader environments that determines the computations it needs to perform (and may fail to perform). The representation and algorithmic perspective offers an understanding of how information about the environment is encoded within the mechanism and what are the patterns of organization that enable the parts of the mechanism to produce the phenomenon. The implementation perspective yields an understanding of the neural details of the mechanism and how they constrain function and algorithms. Once we adequately characterize the distinct role of each level of analysis, it is fairly straightforward to see how they relate.

From molecules to behavior and the clinic: Integration in chronobiology.

Chronobiology, especially the study of circadian rhythms, provides a model scientific field in which philosophers can study how investigators from a variety of disciplines working at different levels of organization are each contributing to a multi-level account of the responsible mechanism. I focus on how the framework of mechanistic explanation integrates research designed to decompose the mechanism with efforts directed at recomposition that relies especially on computation models. I also examine how recently the integration has extended beyond basic research to the processes through which the disruption of circadian rhythms contributes to disease, including various forms of cancer. Understanding these linkages has been facilitated by discoveries about how circadian mechanisms interact with mechanisms involved in other physiological processes, including the cell cycle and the immune system.

Dynamical models: an alternative or complement to mechanistic explanations?

While agreeing that dynamical models play a major role in cognitive science, we reject Stepp, Chemero, and Turvey's contention that they constitute an alternative to mechanistic explanations. We review several problems dynamical models face as putative explanations when they are not grounded in mechanisms. Further, we argue that the opposition of dynamical models and mechanisms is a false one and that those dynamical models that characterize the operations of mechanisms overcome these problems. By briefly considering examples involving the generation of action potentials and circadian rhythms, we show how decomposing a mechanism and modeling its dynamics are complementary endeavors.

The cell: locus or object of inquiry?

Research in many fields of biology has been extremely successful in decomposing biological mechanisms to discover their parts and operations. It often remains a significant challenge for scientists to recompose these mechanisms to understand how they function as wholes and interact with the environments around them. This is true of the eukaryotic cell. Although initially identified in nineteenth-century cell theory as the fundamental unit of organisms, researchers soon learned how to decompose it into its organelles and chemical constituents and have been highly successful in understanding how these carry out many operations important to life. The emphasis on decomposition is particularly evident in modern cell biology, which for the most part has viewed the cell as merely a locus of the mechanisms responsible for vital phenomena. The cell, however, is also an integrated system and for some explanatory purposes it is essential to recompose it and understand it as an organized whole. I illustrate both the virtues of decomposition (treating the cell as a locus) and recomposition (treating the cell as an object) with recent work on circadian rhythms. Circadian researchers have both identified critical intracellular operations that maintain endogenous oscillations and have also addressed the integration of cells into multicellular systems in which cells constitute units.

How can philosophy be a true cognitive science discipline?

Although philosophy has been only a minor contributor to cognitive science to date, this paper describes two projects in naturalistic philosophy of mind and one in naturalistic philosophy of science that have been pursued during the past 30 years and that can make theoretical and methodological contributions to cognitive science. First, stances on the mind-body problem (identity theory, functionalism, and heuristic identity theory) are relevant to cognitive science as it negotiates its relation to neuroscience and cognitive neuroscience. Second, analyses of mental representations address both their vehicles and their contents; new approaches to characterizing how representations have content are particularly relevant to understanding the relation of cognitive agents to their environments. Third, the recently formulated accounts of mechanistic explanation in philosophy of science both provide perspective on the explanatory project of cognitive science and may offer normative guidance to cognitive science (e.g., by providing perspective on how multiple disciplinary perspectives can be integrated in understanding a given mechanism).