1.2Synopsis
Already in the 17th century, Francis Bacon realized that “the secrets of nature reveal themselves more readily under the vexations of art than when they go their own way.” (Bacon 1960 [1620], XCX) Merely observing an object or phenomenon, that is, is not as telling as inspection under manipulation. This principle is the very backbone of empirical research: when trying to understand and explain how a phenomenon comes about, scientists systematically manipulate it. But what exactly is the logic behind these manipulations? How do we get from manipulations to explanations? What inferences can we make based on which experiments? These are questions about the epistemology of scientific experiments. They fall within what I call philosophy of experimentation.
Philosophy of experimentation is a division of philosophy of science. Contemporary philosophy of science has focused on one particular kind of manipulation in the context of explanations: manipulations of some factor Y through interventions I on X. While such interventions have been discussed extensively (e.g. Spirtes et al. 1993; Pearl 2000; Woodward 2003, 2008a; Reutlinger 2012), different authors have presented slightly different conceptions. The most popular account is arguably James Woodward’s (2003; 2015). According to his celebrated interventionist theory of causation, interventions uncover causal (explanatory) relevance relations. Since Woodward nicely captures certain intuitive aspects of empirical research, many have adopted this approach to capture how scientists explain phenomena. Woodwardian interventions thus feature prominently in much of contemporary philosophy of science, specifically in mechanistic and even reductionist accounts of scientific explanation (e.g. Silva et al. 2013; Craver & Bechtel 2007; Craver 2007b; Leuridan 2012).
Yet, a note of caution is in order. First, although mechanistic explanations and interventionism share a common origin, importing interventionist manipulations into contemporary mechanistic accounts of scientific explanation raises a number of puzzles. These are mostly due to the problem of distinguishing between two different dependence relations: causal and constitutive relevance. Basically, the point is that constituents (parts) and mechanisms (wholes) stand in a (constitutive) non-causal dependence relation. Thus, they cannot be subjected to interventionist cause-effect analyses. For illustration consider the simple phenomenon of a clock showing the time. The clock’s showing the time is implemented by a mechanism that moves around cogwheels inside the clock as well as hands on the face of the clock. The cogwheels and the hands are component parts of the clock. Both their movements are constitutively relevant to the clock’s showing the time. The relation between them, by contrast is one of cause and effect: the motion of the cogwheels on the inside causes the motion of the hands on the clock’s face. But the motion of the cogwheels or the motion of the clock’s hands does not cause the clock to show the time—they constitute it. Mechanistic explanations laudably capture this difference between causal and constitutive dependence relations. Now the puzzle is this. Although causal and constitutive relations are clearly different in character, they are both dependence relations. Both the clocks’s showing the time (T) and the motion of the hands on the clock’s face (H) depend (in some sense) on the motion of the cogwheels (C). For instance, if we put a magnet right next to the clock (I) to interfere with C we will in turn interfere with both H (C’s effect) and T (the whole C is a part of). This is exploited by the mechanist who suggests using interventions to uncover both causal and constitutive relations. And indeed this suggestion seems empirically realistic given what we see in scientific practice. However, using interventions to uncover constitutive relations is not compatible with the interventionist view; for wholes (like T) non-causally depend on their constitutive parts (like C).
Second, even if we can modify interventionism such that we can use it to assess constitutive relations as well, problems remain for contemporary “mechanisms and interventions”-style philosophers of science. Most pressingly, there are in fact many different ways in which scientists manipulate when they try to figure out how something comes about; and intervening on a (cause-) variable to elicit a change in another (effect-) variable is only one method among many. The focus on Woodwardian interventions in contemporary philosophy of science therefore is rather short-sighted. If we want to give an account of scientific explanations, especially for heterogenous special sciences such as cognitive neuroscience, we need to pay attention to a wider variety of experimental and inferential strategies employed by practicing scientists.
By carefully examining different manipulative strategies used in cognitive neuroscience, it becomes clear that only a very small subset of empirical research actually employs Woodwardian interventions. And those that do, often combine Woodwardian interventions with what I call mere interactions. Mere interactions are manipulations that are not illuminating qua their effect but qua the information that their application makes available to us. If, for instance, we stain a sample of neural tissue the interesting experimental result is not that the tissue is stained after applying certain chemicals to it—we know it normally will be. This is why we use a staining technique in the first place. The result we are interested in is which neurons project to one another. Although they did so all along, independent of the staining manipulation, it is the staining that reveals this information.
Mere interactions are best understood as tools that allow for the assessment of certain features that are otherwise unobservable. They can be conceptualized as putting something under a metaphorical magnifying glass or fixing background conditions to enable scientific observation. While mere interactions may seem somewhat weaker than Woodwardian interventions, it is important to emphasize that they are in no sense less valuable for empirical practice. In fact, mere interactions might be a precondition for carrying out genuine Woodwardian interventions. Or they may be employed to simulate Woodward-style interventions where genuine interventions are unavailable for ethical, practical, or other reasons. In such cases, sets of mere interactions are interpreted as if there had been Woodwardian interventions, I speak of pseudo-interventions.
Experiments not only differ with respect to what manipulations (mere interactions, pseudo-interventions, or Woodwardian interventions) they employ. Different experimental strategies also permit different inferences with respect to (i) (causal) dependence relationships, (ii) features of (componential or temporal) organization, or (iii) both. All of this information, often integrated across different studies, contributes to explanations of specific phenomena or processes which typically cite some sort of generalization of an observed regularity. Understanding how scientists come up with such explanations takes careful consideration of the experimental designs and the materials and methods scientists employ. Pushing forward into this direction is the main goal of this book. In the end, I will present a catalog of experiments classifying different empirical studies with respect to which manipulations they employ and which research questions they can answer.
1.3“Materials & Methods” of This Book
This book engages with contemporary debates in the philosophy of science and introduces a project in philosophy of experimentation. It aims to draw philosophers’ attention to what precisely scientists do in their laboratories, and to demonstrate how understanding empirical research illuminates philosophical theorizing. At the same time, practicing scientists may find it a helpful introduction to current issues in philosophy of science. As I tailor my discussions mostly to cases from cognitive neuroscience, this book may also be considered a book in philosophy of neuroscience, or perhaps philosophy of cognitive science.
Cognitive neuroscience is a relatively young special science, and an ‘up and coming’ branch of cognitive science. It studies a wide range of cognitive or mental phenomena employing a large pool of different methodologies.2 The common conception among cognitive neuroscientists is that cognitive processes are somehow grounded in, implemented, or realized by neural processes in the brain. Thus, cognitive neuroscience faces some peculiar challenges about relating different levels or domains of investigation, such as psychological and neurophysiological. Relating mental and physical is a challenge familiar from philosophy of mind. However, this is not the place to delve into century-old debates about mind-brain relations. While my discussions do connect to these traditional debates in philosophy of mind, my focus is on scientific methodology, not metaphysics.
To learn about the principle logic behind scientific experiments, cognitive neuroscience is an ideal place to start. For one thing, the tools and phenomena are so diverse that insights from this field are likely to be applicable to many other special sciences. For another, the particular interlevel character of many experiments bears significant challenges for philosophy of science. Learning how we can handle these in cognitive neuroscience will likely inform philosophy of science more generally.
As said, contemporary philosophy of science is dominated by debates on mechanistic explanations (e.g. Andersen 2014; Bechtel & Abrahamsen 2008; Craver 2007b,a; Craver & Darden 2013; Craver & Tabery 2015; Fazekas & Kertész 2011; Franklin-Hall 2016b; Gebharter 2014; Glennan 1996, 2002, 2010b,a, 2013; Illari et al. 2011; Kästner 2015, 2016; Kuorikoski & Ylikoski 2013; Machamer et al. 2000; Woodward 2013) and interventions (e.g. Baumgartner 2013; Gebharter 2016; Hoffmann-Kolss 2014; List &Menzies 2009; Kästner 2011, 2017; Shapiro 2010; Woodward 2003, 2008a, 2015).3 Both theories are—to a certain extent—rooted in seventeenth century mechanistic philosophy, both present significant progress over logical positivist views, and both capture some important aspects of scientific practice. They form the core of contemporary “mechanisms and interventions”-style philosophy of science.
This is where my story begins. Indeed, mechanisms and interventions do some very good work for understanding empirical research, but they face several problems—problems that philosophers cannot solve from their (metaphysical) armchairs. But closely examining scientific practice can teach us how to do it. Once we examine different experiments and their materials and methods—if we engage in philosophy of experimentation that is—we will discover that there are different modes of scientific investigation, including passive observation and different kinds of manipulations as well as different combinations and interpretations of these manipulations. This gives us the information we are missing if we appeal to mechanisms and interventions only and thus helps us gain a better understanding of how scientists explain (not only) cognitive phenomena.