From Psychology to Neuroscience: A New Reductive Account 9783110322620, 9783110322255

Citation preview

Patrice Soom From Psychology to Neuroscience A New Reductive Account

EPISTEMISCHE STUDIEN Schriften zur Erkenntnis- und Wissenschaftstheorie Herausgegeben von / Edited by Michael Esfeld • Stephan Hartmann • Albert Newen Band 21 / Volume 21

Patrice Soom

From Psychology to Neuroscience A New Reductive Account

Bibliographic information published by Deutsche Nationalbibliothek The Deutsche Nastionalbibliothek lists this publication in the Deutsche Nationalbibliographie; detailed bibliographic data is available in the Internet at http://dnb.ddb.de

North and South America by Transaction Books Rutgers University Piscataway, NJ 08854-8042 [email protected] United Kingdom, Ire, Iceland, Turkey, Malta, Portugal by Gazelle Books Services Limited White Cross Mills Hightown LANCASTER, LA1 4XS [email protected]

Livraison pour la France et la Belgique: Librairie Philosophique J.Vrin 6, place de la Sorbonne ; F-75005 PARIS Tel. +33 (0)1 43 54 03 47 ; Fax +33 (0)1 43 54 48 18 www.vrin.fr

2011 ontos verlag P.O. Box 15 41, D-63133 Heusenstamm www.ontosverlag.com ISBN 978-3-86838-108-5 2011 No part of this book may be reproduced, stored in retrieval systems or transmitted in any form or by any means, electronic, mechanical, photocopying, microfilming, recording or otherwise without written permission from the Publisher, with the exception of any material supplied specifically for the purpose of being entered and executed on a computer system, for exclusive use of the purchaser of the work Printed on acid-free paper ISO-Norm 970-6 FSC-certified (Forest Stewardship Council) This hardcover binding meets the International Library standard Printed in Germany by CPI buch bücher.de

TABLE OF CONTENTS ACKNOWLEDGMENTS ...............................................................................3 1. INTRODUCTION .....................................................................................5

1.1 The mind-body problem in philosophy of mind ...........................5 1.2 The metaphysical issue of mental causation .................................7 1.3 The epistemological reducibility of psychology .........................10 1.4 Outline .........................................................................................12 2.

THE PROBLEM OF MENTAL CAUSATION ..............................................15 2.1 Objects, events and properties: preliminary remarks ..................15 2.2 Premises of the problem of mental causation..............................19 2.3 Inconsistency of the premises......................................................33 2.4 Typology of possible positions in philosophy of mind...............34 2.5 Summary and transition...............................................................51

3.

ONTOLOGICAL REDUCTIONISM ...........................................................53 3.1 Classical type-identity .................................................................54 3.2 Multiple realization in an ontological context.............................58 3.3 Supervenience..............................................................................63 3.4 The causal completeness: psychology and physics.....................79 3.5 Non-reductive physicalism ..........................................................83 3.6 The causal argument for the token-identity thesis ......................86 3.7 The token-identity thesis as ontological reductionism................90 3.8 Summary and transition.............................................................100

4.

PSYCHOLOGY AND NEUROSCIENCE ..................................................103 4.1 General background...................................................................104 4.2 Folks psychology as a functional theory of the mind ...............108 4.3 Neuroscience .............................................................................140 4.4 Summary and transition.............................................................167

ii

Contents

5.

EPISTEMOLOGICAL REDUCTIONISM ..................................................171 5.1 Why epistemological reductionism? .........................................172 5.2 Classical reductionism and the requirements of reduction........176 5.3 Multiple realization in an epistemological context ...................182 5.4 Overcoming multiple realization ...............................................193 5.5 The general dilemma of multiple realization ............................212 5.6 Summary and transition.............................................................214

6.

REDUCTION BY MEANS OF FUNCTIONAL SUB-TYPES ........................217 6.1 What should be expected from any account of epistemological reductionism?...................................................................................218 6.2 Starting point: an implication of multiple realization ...............220 6.3 Reduction by means of functionally defined sub-types ............224 6.4 Summary and transition.............................................................247

7.

REDUCTION OF PSYCHOLOGY TO NEUROSCIENCE: CASES STUDIES...251 7.1 How it works: guidelines to reducibility in principle................252 7.2 Finding critical conditions of manifestations ............................264 7.3 Down to neurobiology ...............................................................277 7.4 Summary and transition.............................................................288

8.

FINAL REMARKS ...............................................................................291 8.1 Complete reductionism..............................................................291 8.2 Conservative reductionism ........................................................294 8.3 Back to the mind-body problem ................................................297

BIBLIOGRAPHY .....................................................................................301 INDEX ...................................................................................................313

ACKNOWLEDGMENTS Let me express my gratefulness to the following people, who all have contributed to this dissertation. This includes firstly Michael Esfeld, for the quality of his supervision and his friendship, and Christian Sachse, whose work on functional reduction has been of first importance for the developments of the account exposed in this dissertation. I would like to thank John Bickle and Joëlle Proust for their valuable expertise. Laurent Cordonier, Carl Craver, Jens Harbecke, Michael Hertig, Michael Herzog, Vincent Lam, Michael Sollberger and Georg Sparber all have contributed to the development of my philosophical views through our intensive discussions. The Swiss National Science Foundation (SNF) has financially supported this PhD thesis (grant nr. PDFM1-118603/1), including my visit to Carl Craver at the Washington University in St. Louis in autumn 2009. Thanks to Carlos Fidalgo, Roger Gathman and Tom Tillemans for proofreading. Rafael Hüntelmann and the OntosVerlag made the publication of this book possible. Finally, I have to mention my dept to Jean-Marie Droz for his help in the redaction of this dissertation and for a long and fruitful friendship. Apart from this academic support, thanks are also due to my family, and to my friends in Bienne and Lausanne. University of Lausanne, January 2011

Patrice Soom

Chapter 1 INTRODUCTION 1.1 The mind-body problem in philosophy of mind The mind-body problem has become one of the most central issues of contemporary philosophy of mind in the XXth century. Several attempts to clarify the relation between the mind and the body have been ruled out while others are still in debate. Since none of these proposals seems to enjoy a decisive advantage, the philosophical community is still seeking new arguments and theoretical improvements to fix this issue. The mind-body problem has been for a long time confined to the field of pure philosophical problems. This period starts with Descartes’ (1596-1650) writings on the ontological distinction of two substances, the res cogitans and the res extensa. Giving the impulse for the contemporary opposition between the realm of the mental and the material world, Descartes’ argument, formulated in his Méditations métaphysiques (1641), encountered several objections. Among them, the most important one is certainly the question of the relation between the two substances, giving rise to a debate, of which the contemporary discussion that we know under the label “mind-body problem” is the continuation. The need to understand the nature of the relation between the mind and the body has increased continuously. Several influences contributed to this evolution, among which prominently figures the progressive elaboration of modern physics equipped with conservation principles as well as the increase of our knowledge of the brain contribution to the production of human behaviour. In a historical context characterized by the progressive application of scientific

6

From psychology to neuroscience

methods of investigations to the mechanisms underlying mental phenomena, the mind-body problem slowly started to exceed the limited field of the most prominent philosophical problems and entered the scope of the just-born community of the cognitive sciences.1 It is therefore not surprising that in the middle of the XXth century, some philosophers, in particular Herbert Feigl (1958) and John Smart (1959), came to develop the initial versions of the mind-brain identity theory in order to radically escape the Cartesian framework, which gave rise to so many difficulties with respect to the question of the causal interaction between the mind and the body. However, the type-identity theory has had a very short life, since it was contested only a few years after its initial promulgation. Without entering into detailed historical considerations, what matters here is the fact that the mind-brain identity theory created a clear rupture with the Cartesian framework and settled the parameter for the upcoming and still ongoing debate. Even if the identity theory in its initial formulations has been defeated by the multiple realization argument, advanced by Hilary Putnam (1967) and Jerry Fodor (1974), mainstream philosophy of mind took a clear physicalist direction by staying close to the assumptions that motivated the promulgation of the mind-brain identity theory in the first place. At this point, it has taken some distance from the Cartesian framework, favouring the theoretical background of nonreductive physicalism, which tries as far as possible to conciliate physicalist considerations with intuitions that are sensitive to the specific features of the mental. However, non-reductive physicalism, qua conserving the idea of an ontological distinction between the mental and the physical, is unable to account for mental causation, which is the first issue we shall address here. 1

The interdisciplinary field of research labelled “cognitive sciences” appeared during the 1950s. A crucial event for the institutionalization of this new area of knowledge is certainly the conference on artificial intelligence and its relation with psychology, which held at the MIT in September 1956. Were present at this occasion notably Allen Newell, John McCarthy and Marvin Minsky (computer sciences), Claude Shannon (mathematics), Herbert Simon (psychology), Noam Chomsky (linguistics), Herbert Simon, George Miller and John Swets (psychology), David Hubel and Torsten Wiesel (neuroscience).

Introduction

7

1.2 The metaphysical issue of mental causation Since the specific features of the mental confer to it a least four different dimensions, let us briefly set out our motivations to tackle the mind-body problem from the point of view of the causal issue. The most problematic dimension is certainly the one referred to by David Chalmers (1995: 201) as the “hard problem”, whose challenge is to account for the phenomenal character of our conscious experience in terms of its neurobiological substratum. It currently seems that a kind of extra-ingredient is required in order to account for this dimension of the mind-body problem, ingredient that still remains out of the scope of our current scientific schemes of explanation (see for instance Levine 1993). A second dimension of the mind-body problem ensues from the fact that at least certain types of mental properties, namely propositional attitudes, are intentional properties. From that point of view, the mind-body problem appeals for an account of the intentional “rays” connecting the representing properties to the represented properties and for an explanation of the role of the body in the exemplification of intentional properties. A third important aspect of the mind-body problem ensues from the fact that certain mental properties are supposed to be content bearers. Not only are the propositional attitudes intentional properties, but they also specify a certain mode of presentation of the object to which the properties in question refer, which parallels meaning in the case of language. Again, we have to account for how the body contributes to the fact that human beings are able to exemplify properties that are content bearers. The fourth dimension consists in the problem of mental causation, which is the issue addressed in the first part of this work. Since it will be clearly stated in the next chapter, let us here introduce two important motivations for taking the causal dimension of the mindbody problem as central. Firstly, the causal relation between the mind and the world is problematic per se, as shown by the central position this issue has occupied since the early Modern period. Indeed, in times of deep intellectual upheaval characterized by the first faltering steps of

8

From psychology to neuroscience

the modern sciences, the Cartesian thesis that non-extended minds could initiate freely new causal chains in the natural word appeared already suspicious. Leaving aside Descartes’ attempt to account for how a nonextended substance could interact with extended bodies, the interesting fact is the introduction of the question of the causal interaction as a central issue in philosophy of mind. The progress of empirical science let progressively emerge the idea that the physical world is causally closed. The tension between the claim that our mind is the ultimate cause of our actions and the claim that the physical world is causally closed certainly reached its highest point with the development of neuroscience, which is nothing but the vast research project of accounting for the genesis of human behaviour using the methodological apparatus of the natural sciences. Assuming the causal completeness of physics as a heuristic principle, neuroscience traces the causes of human behaviour back to the activity of the human brain rather than to mental properties, thus putting into jeopardy the status of the mind as being the ultimate non-physical cause of our behaviour. To put it more clearly, the problem of mental causation arises from the fact that a world, which is universally describable by natural science, seems to exclude non-physical entities from having any causal influence on physical bodies. Such a conclusion is problematic because it clashes with our practices as naïve psychologists in which, in our daily lives, at work, at home, or while shopping, we continuously interpret the behaviour of other agents as being the product of their mental activity, be it conscious or not, intentional or not, rational or not. Typically, we are inclined to say, when watching a soccer game, that the player decided to shoot because he judged that he had a good opportunity to score after having seen that the goalkeeper was off-centred. In other words, we ascribe mental properties to other agents precisely in order to explain their behaviour. The causal relation between our mental properties and our behaviour is important because it is a sine qua non condition for our pre-theoretic understanding of ourselves as rational agents. Therefore, rejecting the causal efficacy of mental properties with respect to behaviour would lead to insurmountable problems to

Introduction

9

understand agency, but also thinking, perception and action in general, since they inextricably involve causal relations. Secondly, this centrality of the causal efficacy of mental properties for our ordinary practices justifies tackling the mind-body problem starting from its causal dimension, because the problem of mental causation occurs for mental properties in general, including their phenomenal, intentional, or intensional character. For instance, any account of the phenomenal quality of our mental experiences finally faces the question of how it is possible for that phenomenal character to contribute to the production of our behaviour. After all, just as we explain certain behavioural patterns by referring to mental properties in general, we also account for other similar patterns by referring to the phenomenal character that is associated with the exemplification of these properties. The feeling of “redness” that characterizes my experience when I see a ripe red strawberry is supposed to causally contribute to my eating the strawberry. It is precisely for that reason that we account for these behavioural patterns by referring to the phenomenal character of certain mental properties. Similarly, we expect that any account of the intentional character of mental properties should explain how this character contributes to the production of a specific behaviour, since it is common to explain why people behave as they do in virtue of the intentional character of their mental properties. For instance, my action of slowing down my car can be explained by referring to my belief that the next signal on the road is red, and it is precisely in virtue of being intentionally directed towards this state of affairs that my belief is explanatory relative to my behaviour. An account of intentionality faces thereby the very same challenge, namely to explain in which sense intentionality improves behavioural responses to the environment. Finally, the same reasoning applies to the intensional character of mental properties. For instance, Oedipus’ omission to represent the driver of the chariot he bumped into after having run away from Corinth as his birth-father strongly contributes to his involuntarily fulfilling a part of the prophecy according to which he was predestined to kill his father and marry his mother. Again, accounting for the

10

From psychology to neuroscience

intensional character of mental properties supposes that we account for how such modes of presentation can impact the behaviour. Consequently, the first part of our developments will be addressed to the ontological question of the relation obtaining between mental and physical property tokens, in order to establish a position which can do justice to the causal efficacy of the mental in a world described by the natural sciences. This is not to say that the other dimensions of the mind-body problem do not matter, but rather that the accounts of consciousness, intentionality or content should take into account constraints hanging over the causal issue in order to be satisfactory. Tackling the mind-body problem from its causal dimension will bring us to propose an ontological monism, whose core claim is the token-identity thesis, according to which any causally efficient mental property tokens is identical to a complex configuration of physical property tokens. As we shall argue, if the physical world is causally closed, the only way to vindicate the causal efficacy of mental properties consists in identifying them to configurations of physical property tokens. The multiple realization argument plays here a key role, since it precludes isolating a physical correlate for each type of mental property. However, it does not prevent the identification of property tokens. Any mental property token is identical to a certain configuration of physical property tokens, although it is not the case that every token of a certain mental type is identical to a configuration of physical property tokens of the same type.

1.3 The epistemological reducibility of psychology The token-identity thesis implies that psychology and physics ultimately share, at least partly, their truthmakers, i.e., certain entities in the world that verify both a physical and a mental description. This obviously raises the question of the epistemological relation between the theories accounting for what appeared only prima facie as two ontologically different kinds of property tokens. On the one hand, psychology is concerned with the theorization of the relations obtaining between people, their environment, the mental properties they

Introduction

11

exemplify and their behaviour and it uses the interplay between environment and mental properties in order to explain the behaviour. On the other hand, neuroscience, being that sub-set of natural sciences devoted to the study of the nervous system, is concerned as well with behaviour. It aims to account for behaviour on the basis of the brain functioning. From that point of view, both descriptive modalities share their truthmakers, but also have the same explanandum, namely human behaviour, with the result that we can expect to be able to establish systematic links between the two theories. Inter-theoretic reduction is the primary candidate for such a systematic relationship. However, current accounts of epistemological reduction fail because of the argument that was already problematic in the context of the ontological question of the relation between mental and physical property tokens, namely, the multiple realization argument. This argument prevents the establishment of the bi-conditional bridging principles that are required in order to deduce psychological law-like generalizations from lowerlevel laws. We shall therefore, in the second part of our developments, establish a diagnosis of the problems raised by the multiple realization argument in order to propose a robust account of epistemological reductionism. Claiming that psychology is epistemologically reducible to neuroscience contradicts a widespread view in the philosophy of mind, according to which one can accept the token-identity thesis without endorsing the epistemological side of reduction (see for instance Fodor 1974). However, this position has as its drawback nothing less than the elimination of psychology. After all, if any causally efficacious mental property token is identical to something physical and if, as claimed by the multiple realization argument, there is nothing relevant in common from the physical point of view to entities verifying the same psychological description, one can follow the proponents of eliminative materialism (for instance Churchland 1985), who claim that psychology fails to grasp any objective feature of reality. This obviously leads to the uncomfortable results that psychological descriptions fail to refer to anything real and, as a result, that psychological property tokens do not exist. For that reason, the

12

From psychology to neuroscience

epistemological reducibility of psychology remains a central issue to secure the claim that mental property tokens are causally efficient. This is why we aim here to pursue a fully fledged reductionist but nonetheless conservative account of the relation between psychology and neuroscience, on both the ontological and the epistemological sides. This account will explain how it is possible for psychology to grasp objective patterns of similarities in spite of multiple realization. Thus, it vindicates the status of psychology as a legitimate part of our system of objective knowledge.

1.4 Outline Our discussion will proceed as follows. After defining a general ontological framework taking property tokens as primitive, the second chapter aims to formulate properly the metaphysical problem of mental causation. The problem results from four premises that are sound, independently taken, but together inconsistent, raising a quadrilemma that any position on the mind-body issue must face. This formulation will enable us to draw the logical space of possible solutions to the problem of mental causation and evaluate the philosophical costs of the abandon of each one of the constitutive premises. The third chapter examines the main alternatives resulting from the rejection of the premise stating the ontological distinctness of mental property tokens and physical property tokens. Provided that classical type-identity fails when confronted with multiple realization argument, we shall turn to the mainstream position in philosophy of mind, non-reductive physicalism, which aims to reconcile this failure with the requirements of a physicalist position by the recourse to supervenience. However, Jaegwon Kim’s supervenience argument shows that taking mental property tokens to be supervening on physical property tokens, while being nonetheless ontologically distinct from them, is not a conclusive move. This motivates the endorsement of a weaker version of the identity thesis, namely token-identity, which is immune to the multiple realization argument. The token-identity thesis is the claim that any causally efficacious mental property token is identical to a certain configuration

Introduction

13

of physical property tokens. Thus, certain entities in the world make true both a psychological description and a physical description, suggesting the epistemological reducibility of psychology. The fourth chapter aims therefore to clarify the status of these two descriptive modalities that uses two radically distinct bodies of conceptual resources to account for the genesis of the human behaviour. On the one hand, we shall argue that common-sense functionalism is the only position that can do justice to our practices as naïve psychologists. This supposes that we avoid the functionalism dilemma, whose horns are epiphenomenalism and eliminativism, by endorsing a causal theory of properties. On the other hand, neuroscience provides reductive explanations of the mechanisms by means of which the behaviour is produced. It thereby develops a hierarchical model of explanation accounting for higher level cognitive phenomena on the basis of the functional organization of brain components. The fifth chapter examines the main alternatives with respect to epistemological reductionism and the crucial role played by the multiple realization argument in this debate. On the one hand, the classical Nagelian account of epistemological reductionism fails when confronted to the multiple realization argument since the latter excludes the deducibility of psychological laws from lower-level laws. On the other hand, both New Wave Reductionism and functional reduction collapse into the elimination of psychology, failing to provide an account of the ability of psychology to abstract from lowerlevel differences distinguishing entities verifying the same psychological description. This highlights the general dilemma raised by the multiple realization argument, which drives the debate either towards the elimination of mental properties or towards property dualism. Against that background, we shall propose in the sixth chapter a conservative account of epistemological reduction which aims to turn the multiple realization issue into an intra-theoretical problem, thereby opening the way to an explanation of the ability of psychology to abstract from physical differences. Grounded in the lessons to be drawn from the multiple realization argument in the context of token-identity, the central claim of this account is that physical differences imply

14

From psychology to neuroscience

functional differences as well. The latter can be grasped from the psychological point of view insofar as that appropriate environmental circumstances are provided. This enables the construction of fully functionally defined sub-types of psychological types that are ultimately co-extensional with physical types. On this basis, psychological sub-types can be safely reduced to physical types while neglecting the functional specificities embedded in the sub-types of a given abstract type allows retrieving the original psychological types through an intra-theoretical process of abstraction. The last chapter aims to provide guidelines for the application of this proposal and to substantiate the claim that physical differences among truthmakers of a given psychological description imply functional differences as well. Firstly, the case of the disconnection syndrome will be discussed in order to illustrate the different features of our account of epistemological reduction by means of functionally defined sub-types. On that basis, we shall then argue that this scheme of application can be extended to any other case of comparable neurological specificities, since neuropsychology aims to keep contact with behaviour, even in case of a complex neuropsychopathology such as Schneiderian schizophrenia, which has prima facie very ill-defined behavioural effects. Finally, we shall provide empirical support to the claim that physical differences imply functional differences as well by considering the molecular mechanism of memory consolidation, illustrating thereby the applicability of this strategy of reduction down to neurobiology. At this point, we hope that convincing arguments will have been provided in order to show that reductionism is the only position that can to make sense of mental causation. In particular, the multiple realization argument does neither refute physicalism conceived as an ontological monism, nor does it conclusively block the way to the epistemological reduction of psychology to neuroscience. Rather, multiple realization vindicates the scientific indispensability of psychology, since it is only by using psychological descriptions that we are able to bring out what relatively large sets of individuals differing from the physical point of view have objectively in common.

Chapter 2 THE PROBLEM OF MENTAL CAUSATION This chapter proposes to establish mental causation as a genuine philosophical problem. The contemporary version of the mind-body problem is framed by four premises that each, individually, are intuitively reasonable, but that, brought together, form an inconsistent whole. The list of the four premises goes like this: 1. Mental property tokens cause physical property tokens. 2. Mental property tokens are distinct from physical property tokens. 3. The causal completeness of physics. 4. The absence of causal overdetermination. These four premises form a self-contradictory whole: to make our premises consistent, one of them has to be dropped. We will argue in the second part of this chapter in favour of abandoning the premise of distinctness, the consequences of which will then be pursued in the next chapters in order to establish a complete reductionist position.

2.1 Objects, events and properties: preliminary remarks There are several formulations of the problem of mental causation in the contemporary debate. They differ in their characterization of the relata of the problematic causal relation between the mind and the body. Let us therefore begin with some preliminary considerations in order to clarify this issue. Ordinary language is full of direct or lightly disguised causal statements such as, for instance, “the car slid off the road because of

16

From psychology to neuroscience

the snow”, or “photosynthesis is caused by sunlight”, etc. This suggests taking objects as the relata of causal relations. However, a closer look at making causality depend upon an ontology of objects presents us with two descriptive insufficiencies. First of all, it is not the case that every property of an object is causally relevant to the occurrence of the effect on the target object described in the causal statement. It is the slipperiness of the snow that causes the car to slide out of the road, rather than its whiteness, and the photosynthetic process that the plant uses to convert sunlight into useable energy does not depend on the mass of the sun. Secondly, objects are subject to change over time, even as they retain their identity. For instance, a signal can be red at time t1 and green at time t2., or even non coloured at time t3, all being possible states of the signal, which we still identify as “the signal”. In the case of a car crash involving a signal that wasn’t properly working, when saying that “the signal” was responsible for the two cars at the intersection crashing together, we refer to to a particular property the signal exemplified at a particular time, namely that it was not working as expected. These factors encourage us to make our causal talk more precise by addressing these two items; the precision will have a bearing on the problem of mental causation. One way of addressing the problem of physical causes is to recast events, as Kim has done, as properties exemplified at a given time (Kim 1976). Kim’s theory of events meets the problem of bringing to bear the relevant properties in our causal story, and retaining the identity of the causal factor, by conceptualising any individual event as a structured particular that “has three unique constituents: a substance (the “constitutive object” of the event), a property it exemplifies (the “constitutive property” or “generic event”, and a time” (Kim 1976: 160). Accordingly, Kim’s picture of an individual event is captured by the formula [x, P, t]. In the case of more complex events, the account has a more general form, taking events to be structures consisting of an n-tuple of objects, an n-adic relational property, and a time. For instance, “Cesar was stabbed by Brutus” can be rendered by [(Cesar, Brutus), is stabbed by, t]. Kim’s account possesses the obvious advantage of fitting with our intuition that an object interacts with another object in virtue of

The problem of mental causation

17

those of the former object’s properties that are causally relevant at the time it exercised its causal power on the latter. Kim’s conception of events, here, responds to the insufficiencies of the theory of events Donald Davidson developed in the 1960s. Davidson also took the relata of causal relations to be particular events, but unstructured particular events. The motivation underlying such an account is that we should draw a strong distinction between causes, i.e. events as particulars extended in space and time, and the specific way we refer to them. Accordingly, if we refer to an event under a specific description, for instance “the match was struck”, it is simply wrong to think that “we have not specified the whole cause of an event when we have not fully specified it” (Davidson 1967: 698). To pick out the cause of a particular effect means to refer to it through a description, but there is no guarantee that the description in question will pick out every causally relevant aspect of the target event, nor does it need to. Davidson’s idea is that our descriptions of the causes of a given event are always incomplete. Instead of specifying sufficient conditions, we rather specify sufficient conditions against the background of tacit assumptions. So, for instance, striking the match is sufficient to light the match only under assumptions that the direct physical environment contains enough oxygen, that the match is not wet, etc. However, when using the description “the match was struck” we refer to the whole cause of the match’s catching fire and have thus given a satisfactory explanation of it. It is however doubtful that such an account of events is sufficiently fine-grained to constitute a framework in which the problem of mental causation can be solved. When Davidson says that we wholly grasp the cause of lighting a match by using the expression “the match was struck”, it is perfectly reasonable to ask if the match lighted qua the head of it being brown or qua its being made of sulphur. This is especially important relatively to the problem of mental causation, as pointed out by Terence Horgan (1989). According to Davidson, an event is mental insofar as it is graspable using a description containing essentially a mental verb, or it is physical insofar it is graspable using a description that contains essentially just physical vocabulary (Davidson 2004: 688). Now, given that Davidson

18

From psychology to neuroscience

assumes that all events are physical events and that some of them support also a mental description, he claims that the problem of mental causation does not appear within the framework of anomalous monism. Mental descriptions pick out causes that also always support a physical description, and in this way we can speak about them as though they were causes of physical effects. Horgan’s objection to the Davidsonian approach to mental causation is simply to point out that a solution to the problem of mental causation is reached just in case we possess a vindication of the causal efficacy of the mental events qua being mental. In other words, what we need, within this framework, is to show not only that the truthmaker of a mental description is the cause of a given physical effect, but also that the mental description refers to mental properties that are causally relevant to the occurrence of the physical effect. If we fail to meet this last condition, then we are led directly to epiphenomenalism with respect to the mental, because Davidson’s account tells us that the physical properties of an event obey strict physical laws that suffice to cause the designated effect. There is consequently no place, in this causal account, for mental properties; they do nothing to cause the event. Horgan concludes that in the Davidsonian framework, mental properties must be “quausally” epiphenomenal (Horgan 1989: 51). Kim’s account of the event structure does not raise similar problems, since events are conceptualized through the properties that objects exemplify at a given time, which allows us a finer grained event than the kind Davidson describes, because it takes as an individual event any triplet made of an object, a property exemplified by the object, and a time. Therefore, in such a framework, for an individual exemplifying n physical properties and m mental properties at time t, there are n+m events at t. Kim’s schema meets Horgan’s objection, since there is no implication that an event causes its effect qua being mental or qua being physical. An event is physical if it consists in the exemplification by an object at time t of a physical property, and it is mental if it consists in the exemplification of a

The problem of mental causation

19

mental property at time t. If an object exemplifies different kinds of properties, then there are several events of different kinds.2 Since the problem of mental causation is about the causal relation between individual mental events and physical causal events, and since mental and physical individual events are defined as such by the kind of property that characterizes them, let us use in what follows the designation ‘property token’ to refer to a property exemplified by a particular object at a given time. The expression “mental property token” designates therefore the exemplification of a mental property by a particular object at a certain time. The expression “physical property token” designates analogously the exemplification of a physical property by a particular object at a certain time. The central question of whether or not mental events cause physical events to occur can therefore be reformulated to ask whether or not mental property tokens cause physical property tokens to occur (Esfeld 2005: 6). The substitution of the latter question for the former retrospectively gives another cast to the terms which articulate the premises of the problem of mental causation.

2.2 Premises of the problem of mental causation 2.2.1 The Premise of mental causation Initially, given our overview of event descriptions above, the problem of mental causation concerns asserting that mental property tokens cause physical property tokens to occur. All accounts of mental causation endorse some variant of this formulation of the problem, since rejecting it outright would be a non-starter, leading to the difficulties that will be sketched out in the section devoted to noninteractionist forms of dualism (2.4.2 Dualism without interaction: parallelism and epiphenomenalism). Thus, our problematic claim can be formulated as follow: (MC) Mental property tokens cause physical property tokens. 2

Technically, Kim (1976, 161) gives the following condition of identity for any two events: [x,P,t] = [y,Q,t’] just in case x = y, P = Q, and t = t’.

20

From psychology to neuroscience

The intuitive justification of the causal efficacy of mental property arises from the following elements. First of all, our linguistic practices are full of references to a presupposed causal link between mental property tokens and physical property tokens. Examples of such statements are legion: “John took the train because he desired to see his brother, who is living in the next city”, “Frida’s belief that surgeons saved her life led her to study in a medical school”, “Peter’s desire of hanging up a photo of his wife caused him to make a hole in the wall”, etc. They are inscribed in the practices of everyday life, and in our mutual expectations of one another. Secondly, these linguistic practices reflect our pre-theoretical conception of human agency, in which the premise of mental causation is firmly grounded. Note that this is true whatever point of view we take on agency. From the first person point of view, we strongly experience our behaviour as the results of our inner mental life. When I want to eat an ice cream and walk to the refrigerator to get one, I feel like my desire caused me to go to the refrigerator. This first person perspective is continuous with the third person perspective, in which we also assume a causal link between the mental properties of the other agents and the physical output that consists of their behaviour. We are always ascribing mental properties to other agent in order to explain and rationalize their behaviour. Indeed, without the assumption that other agents behaviour is caused by their mental properties, the largest part of our social life would be simply incomprehensible. Rationalizing the behaviour of others is not a matter of charity of interpretation, or coddling an illusion. As Davidson put is, crediting people with a large degree of consistency “is unavoidable if we are to be in a position to accuse them meaningfully of error and some degree of irrationality. Global confusion, like universal mistake, is unthinkable, not because imagination boggles, but because too much confusion leaves nothing to be confused about” (Davidson 2004: 696). Speech acts are certainly the most straightforward example of something that is evidently physical – the bodily act of making sounds – that seems to be caused by something mental. When someone says, for instance, “Liberal democracy is the best political system ever devised”, the speech act supposes a very rich background of beliefs about the involved objects,

The problem of mental causation

21

moral judgments, and, finally, the desire of expressing this particular thought. Without ascribing him these other mental properties and without assuming these properties to be causally efficient for the production of the production of behaviour, this behaviour would remain completely obscure. Agency is at the centre of our anthropological conception of ourselves as rational agents in general, along with the idea that we have the capacity to act freely and can be held to be morally responsible. Without brushing against the debate about free will here, let us simply highlight the logical fact that any conception of free will, libertarian or compatibilist,3 presupposes the causal efficiency of mental properties. If, according to libertarian accounts of free will, to act freely means at least in some broad sense to act according to the agent’s self produced desires,4 and if it is in virtue of that freedom that we can be held morally responsible for our actions, not simply in relation to some institution but in all of our everyday relationships with others, then we must have a notion that desires are causally efficacious in bringing about action. If this were not the case, there would be simply no way to impute any responsibility to the guilty. Similarly, compatibilist accounts trying to reconcile free will and causal determination also requires that mental property tokens are causally efficacious in order to get a link with moral responsibility. Harry Frankfurt’s classic defense of compatibilism for instance takes the freedom of action of persons to be a sufficient condition for moral responsibility (Frankfurt 1971: 19). But again, this supposes that the mental property tokens are causally efficacious for the production of behaviour. There is no point in reward or punishment, morally speaking, if the mental properties he exemplifies are not in someway causally related to his behaviour. To sum up, rejecting the efficacy of mental causation would conflict on the deepest level with our general anthropological conception of ourselves as rational, free and morally responsible agents

3 4

We exclude here the eliminativist conceptions of free will.

For instance David Hume takes freedom as « a power of acting or of not acting, according to the determination of the will” (Hume 1977: sect.viii, part 1).

22

From psychology to neuroscience

– in fact, we would lose agency itself (Kim 1998: 32). Any account of the insertion of mental properties in the physical world that neglects their causal efficiency would be incredible, on both the everyday and theoretical level. Jerry Fodor dramatically describes the result of this loss: “if it isn’t literally true that my wanting is causally responsible for my reaching, and my itching is causally responsible for my scratching, and my believing is causally responsible for my saying (...) if none of that is literally true, then practically everything I believe about anything is false and it’s the end of the world” (Fodor 1989: 77). Perhaps end of the world is an exaggeration, but definitively the end of a trans-cultural conception of the mind, and hence of the human being, that would utterly displace him from a distinct causal place in the realm of nature. 2.2.2 The Premise of distinctness Our second premise states that mental property tokens are ontologically distinct from physical property tokens. This premise of distinctness is formulated as follows: (DM) Mental property tokens are ontologically distinct from physical property tokens. Just as we saw with the premise of mental causation, this second premise is supported by several pre-theoretical intuitions with respect to the ontological constitution of our world. Among them, many are tied to certain characteristics of mental property tokens, characteristics that seem to be lacking among physical property tokens. The most notable are their qualitative character, our privileged first person access to them, their intentional character and finally, their content.5 Although each of these items is disputed, we aren’t entering into the lists to defend any position here. We will buttress our argument, in what follows, with reasons to belief that the premise of distinctness is a genuine factor in articulating the problem of mental causation.

5

Note that we do no claim that any mental property token exhibit all these characteristics. Pain is for instance certainly not an intentional property. Pain is caused by something, but not intentionally directed toward that this cause.

The problem of mental causation

23

At least some mental property tokens are characterized by their qualitative character. This qualitative character is the equivalent to the “what it is like” for an individual to exemplify a certain mental property (Chalmers 1996; Nagel 1974), that is to say, experiences associated with the mental life of the individual agent. The point is that being in pain, or seeing a red tomato, or baling out of an airplane, etc, has internal qualitative features that we undergo when exemplifying these mental properties. The term “qualia” is currently used to refer to these qualitative characteristics. Qualia have given rise to a lot of contention in contemporary philosophy of mind, with questions ranging over the following topics: Can the qualia spectrum be inversed without any change in the brain of people instantiating them (Shoemaker 1982)? Are qualia epiphenomenal (Jackson 1982)? Does the phenomenal character of certain mental property tokens establish that physicalism cannot completely describe all what there is in the world (Jackson 1986)? Noting these questions, we are going to narrow our interest to that defined by the causal explanatory scheme we are pursuing here. Qualia are most often conceptualized as purely qualitative, with the result that it is unclear how to integrate them to our system of scientific knowledge (Levine 1993). However, the plausibility of the case for the qualitative character of mental properties clearly hinges upon the fact that we have a privileged access to at least some of our own mental properties through a kind of “inner” perspective. If I feel in pain, it seems that I have a direct and privileged access to my pain, which is different from my access to property tokens laying out in the world, as for instance a falcon flying in the sky. The difference is obvious when we think in term of possible errors. Whereas I could be in error that the bird in the sky I’m seeing is a falcon, I can’t be error about whether or not I am in pain. The point here is that I have an immediate access to at least some of my mental properties, which is evidently not the case for properties belonging to the physical world. The perception of the “outer” world is, at least in part, due to the functioning of our sensory apparatus. However, with feelings like pain, our access is immediate and error free. This suggests that mental property tokens have a distinctive nature. Note that “inner” experience, here, is the only available

24

From psychology to neuroscience

evidence for our privileged epistemic access to mental properties. There is no possible leap that would take us into the world of those properties as they are experienced by someone else. When we perceive the anger of someone else, we also perceive a mental property, which might remind us of our feeling of anger at this or that time; yet the other person’s anger is always given to us from an “outer” perspective not so dissimilar to the perception of physical objects. We cannot feel it inside. A third element suggesting the distinctness of mental properties in comparison with physical properties is the fact that at least some of them are intentionally directed towards, or about other property tokens, of whatever kind. When someone perceives a falcon in the sky, the target mental property token relates somehow to a property token that has wings, flies at a high altitude, etc. Obviously, physical property tokens are not structured by an intentional character. In some sense of course most of the newspaper contributions on the 12th of September 2001 were about the attack of the World Trade Centre in New York, but only in so far as someone exemplifying genuine mental properties reads about it. The fundamental physical property tokens composing newspapers, like electrons, masses or spins, etc, are not about something. Finally, mental property tokens seem to be distinct from physical property tokens because they are content bearers. In addition to be about some object, they represent these under a certain mode of presentation, highlighting certain aspects of the entity rather than others, just as the language makes reference to objects under a certain meaning. As in the case of the intentional character of mental property tokens, the contrast with physical properties is straightforward: single electrons, molecules or charges are not content bearer by themselves. In addition to these intuitive elements, there is a further and more theoretical reason to take mental property tokens to be distinct from physical property tokens, which we will see played out in the coming chapters, namely that mental properties are multiply realized: they are not bound in a necessary relation with some specific physical type of property tokens such that they must be exhibited only when the physical property token occurs, or even must be exhibited when the

The problem of mental causation

25

physical property token occurs. (Fodor 1974; Kim 1992: 1; Putnam 1975). The basic idea is that individuals exemplifying the same mental properties can exemplify very different physical properties. This does not exclude per se any tie between mental and physical property tokens. However, as we shall argue later, this argument is usually taken to exclude any necessary and symmetrical link between mental and physical property types, which suggests that there is something more to mental properties than being only configurations of physical properties. In a nutshell, given that two individuals can be identical from the psychological point of view while being different from the physical point of view, the multiple realization argument suggests prima facie that the referents of psychological descriptions are not something physical and thereby that mental property tokens are ontologically distinct from physical property tokens. We will go into this in detail when ontological (Chapter 3) and epistemological (Chapter 4) reductionisms will be discussed. We mention it here only to show that the premise of distinctness is still a vital premise of the problem of mental causation. 2.2.3 The Premise of the causal completeness of physics Our third premise is formulated in this way: (CCP) For any physical property token p, insofar as p has a cause, it has a complete physical cause. There are several other different formulations of this claim.6 We have chosen a formulation adapted from Esfeld (Esfeld 2005: 5) because it captures precisely the idea that any physical property token occurs in virtue of the occurrence of a previous property token or complex configuration of physical property tokens, but without disallowing directly and explicitly any kind of mental-to-physical causation.7 Let us add two further comments with respect to that 6

See for instance Kim (2005: 43 and 50), Papineau (2002: 17), Harbecke (2008: 18).

7

For a discussion of the interest of ruling out mental-to-physical causation and overdetermination of physical property tokens by additional mental causes directly by the closure premise, see Kim (2005: 50-68).

26

From psychology to neuroscience

formulation of the completeness. Firstly, by the “insofar” clause, we mean that completeness is restricted to physical property tokens that have a cause at all, in contrast to, for instance, the “big bang” in our current cosmological theories. If the big bang is understood as a kind of starting point of the causal history of our universe, then this initial physical event does not admit any prior causes. Secondly, when we talk about the “complete” clause of a physical property p2, insofar as it has a cause, we mean that physicist will never have to look outside of the domain of physical property tokens to isolate a cause that is sufficient for the occurrence of p2. That is, for any physical property token p2, there is always a complete physical cause p1 that is sufficient to bring about p2. This formulation of the premise of causal completeness of physics is neutral with respect to the debate between determinist and probabilistic conceptions of physical causation. Obviously, if physical causation is strictly determinist, (CCP) is simply to be interpreted as explained above. If physical causation turns out to be probabilistic, as quantum physics suggests, the physical causal antecedents still completely determine the probabilities that are involved in the occurrence of any physical effect. In others words, the probability that a given physical property token occurs is fully determined by previous physical conditions, and there is still no need for physicist to go outside of physics to look for the cause of the given occurrence. Therefore, even if causation is probabilistic, the physical world is still causally closed. Probabilistic or not, why do we believe that physics is causally complete? After all, even in the XIXth century, even famous scientists rejected this premise, taking it as self-evident that sui generis vital or mental forces existed (Papineau 2002: 45). Indeed, there is no clear cut a priori argument favouring the completeness of physics. Therefore, since the premise of causal completeness is not grounded in conceptual necessity, it is a contingent claim about the structure of our world, and should consequently be defended on empirical grounds. David Papineau has pursued this argument, pointing out that the evidence for materialism “has not always been available to philosophers” because “its crucial premises, the completeness of physics, rests on empirical

The problem of mental causation

27

evidence, which has emerged only relatively recently” (Papineau 2004: 46). We will borrow the main following argument from that articulated by Papineau (2002: Appendix). The evidence for the causal closure of the physical world is closely related to the discovery of the laws of conservation in the mechanistic conception of the natural world that characterised the Modern period. Already Descartes subscribed to the principle of conservation of motion, which is defined as mass times speed, thinking that the quantity of motion of any collection of physical bodies in an isolated physical system is preserved over time. However, this does not imply that the physical world is causally complete, since the conserved quantity is non-vectorial, or, in other words, motion is a function of speed, a non-directional scalar (Woolhouse 1993: 105). That is precisely what allowed him to reconcile this conservation principle with the causal influence of a distinct mental substance. Since uniquely the total quantity of motion is preserved, this allows the thinking substance to causally interact with the extended bodies by changing the direction of these bodies, as long as the total quantity of motion is preserved. Indeed, from the point of view of the laws of conservation, nothing in the Cartesian framework precludes such a causal interaction between a non-physical substance and a physical substance. According to Leibniz (1646-1716), it is this feature of his mechanics that allowed Descartes to think that the mind exercises a causal influence via the Pineal gland on the body. The situation changed significantly with the improvement by Leibniz himself of the principles of conservation. This improvement consisted in the discovery that both the linear momentum and the kinetic energy of an isolated physical system are conserved. This led him to reject the Cartesian account of mind-body interaction, since the conservation of momentum requires the preservation of not only the total quantity of motion, but also of the direction. In the Leibnizian account the momentum is not a scalar value but a vector. This precludes the changes in direction of physical particles alleged by Descartes. Moreover, the conservation of energy, jointly with the conservation of momentum, fully fixes speeds and directions of the particles after any causal interaction with other particles. There is

28

From psychology to neuroscience

therefore no room for any causal influence of the mental on the physical. Leibniz thus establishes the causal completeness of physics, since the principle of conservation he formulated implies that the physical world is fully determined by physical laws, the determinism being a sufficient condition for the completeness. However, the whig version of history of science that imagines it as a linear progress has long been shown to misread the evidence. At the core of the reconceptualization of mechanics by Isaac Newton (1643-1727) resides the introduction of the notion of impressed force that replaced the narrower notion of interaction by mechanical contact that characterized both Cartesian and Leibnizian accounts. Descartes thought that only physical bodies, with the exception of mental substance, could, by impact, cause changes in another material body. Given his conception of the conservation principle, Leibniz argued such an interaction between the mental substance and physical bodies is not possible. But Leibniz conserved the idea of causal interaction by mechanical contact. However, the concept of impressed force does not require a contact between two items for the first to exercise a causal influence on the second. The paradigmatic case is gravity, which is according to Newton a disembodied force exercising its influence distally. But if such forces acting without any mechanistic contact are allowed, then the door is also open for the action of chemical or magnetic forces, and nothing anymore precludes the non-mechanical action of vital or mental disembodied forces. Thus, the Newtonian reconceptualization of mechanics undermined the claim of completeness of the physical domain. It has to be noticed that laws of conservation did not play a central role in the early Newtonian theory. Newton’s third law indeed implies the conservation of the momentum. But this does not itself impose restrictions on the way forces interacts and on what could count as a force, and especially as a conservative force. And even when a universal principle of conservation of energy was formulated by Hermann Helmholtz (1821-1894) in 1847, it was still possible to imagine that non physical suis generis forces could exist that act jointly with physical forces in living systems, with the restriction that these

The problem of mental causation

29

forces have to be governed by deterministic laws in order to be compatible with the principle of conservation. To put it in a nutshell, since the principle of conservation does not tell us anything about which basic forces act in our world, as long a force acts in a deterministic and conservative way, it cannot be used to rule out nonphysical forces. Given this picture of the world, the fondness of early XIXth century scientists for the notion of vital force, especially popular in anatomy and physiology, was not a throwback to mysticism. Rather, it seemed to describe a distinct force that operated in nature, even if it did not arise from physical causes. Since the middle of the XIXth century, the scientific community was divided with on the question whether or not there are fundamental deterministic forces that exercise a power that is only seen in living beings or in their brain. On the one hand, there were scientists like Helmholz who adhered to a reductionist program and argued, notably in his 1861 talk, “On the Application of the Law of the Conservation of Force to Organic Nature” that there is no vital force at work in organisms, and that the explanation of all organic function could be reduced to the effects of a small set of physical forces that are not specific to living beings. But at this point, Helmholtz was neither arguing on the basis of empirical knowledge of organic function nor drawing his argument out the principle of conservation of energy. Given this state of affairs, most physiological investigations during the second half of the XIXth century were conducted under the assumption that some suis generic animated forces are acting in living systems in addition to the physical forces. Although these researches led to incredible scientific progress, they failed to provide any positive evidence with respect to the existence of such irreducible non-physical forces. However, the balance in this argument tilted decisively to the physicalist side when molecular biology, starting in the 1940s, made discoveries that explained the basic chemical process operating in the cells without referring to any vital force. Once physiologists had explained most of the properties of the cell by referring to the DNA, and once they had explained the neuronal activity in terms of electrical action potential, the case against vitalism was almost sealed.

30

From psychology to neuroscience

Physiology failed to find any evidence for non-physical forces; there was no trace of them even at the molecular level. As Papineau argues, the truth of the premise of causal completeness of physics is grounded in the fact that non-physical vital or mental forces have failed to explain any feature of living systems. The point is that “for once we know which other forces exist, then we will know which anomalous accelerations would indicate the presence of special mental or vital forces. Against this background, the argument from physiology is then simply that detailed modern research has failed to uncover any such anomalous physical processes” (Papineau 2002: 253). Let us sum up this development. First of all, the premise of causal completeness of physics is an empirical claim, rather than an a priori one, and must hence be defended in the light of empirical results. Historically, the discovery of the law of conservation was the first step in the argument for the causal completeness of physics, since, given any isolated physical system, both the momentum and the overall quantity of energy present in the system are preserved. The second step was the result of molecular biology, which has shown that organic processes can be explained through the same kind of chemical and physical activity that we see in the non-organic domain. Or, to put this in other terms, science has so far failed to uncover any positive evidence suggesting the existence of any non-physical force. Against that background, conservation of energy and momentum became a central feature of our scientific understanding of the world. Empirically, it is perfectly reasonable to assume that physical property tokens, either deterministically or probabilistically, are caused by previous physical property tokens, insofar that they have a cause at all. As final word, let us emphasize that we are not thereby endorsing a stronger version of this premise, such as “any cause of a physical property token is itself a physical property token”.8 This would give to our empirical premise the appearance of some conceptual truth: to be a physical property tokens is to have a physical cause. However, this conceptual argument seems weaker than the formulation we have

8

Adapted from Kim (2005: 50).

The problem of mental causation

31

borrowed, for the most part, from the historical analysis of Papineau. We should note that our formulation, consequently, does not stipulate that physical property tokens have no mental causes acting additionally to their physical complete cause. Thus, we don’t construe the completeness of physics to mean that causal overdetermination is directly ruled out. It is for this reason that ruling out causal overdetermination assumes its own separate place on our list. 2.2.4 The Premise of the absence of causal overdetermination Overdetermination is certainly a concept about which we have only few pre-theoretical intuitions. The very idea of causal overdetermination is present in our discourse not because of rooted and ancient social habits, nor because of scientific progress, but is given, rather, as an abstract theoretical possibility arising from the argument that a property token could have more that one lone complete cause. The paradigmatic case of causal overdetermination is the situation in which an event is causally determined by more than one complete cause. The classical illustration is that of two hunters shooting a deer in the heart at the same time. In such a case, the hypothesised causes do not meet the standard counterfactual criteria to count as a cause of the roe’s death. Indeed, if we hypothesize that one of the hunter’s had missed the deer, it would nonetheless have been killed. More precisely, both the counterfactual “if the hunter A would not have shot, then the roe would not have been killed” and the equivalent counterfactual applied to the hunter B are false. Hence, the absence of one of the possible causes would not have prevented the death of the deer. If counterfactuals are the only route to construing causes, then here we seem to have a case of no causation at all. But it is obviously not in that sense that causal overdetermination comes a part of the problem of mental causation. In the case of the two hunters shooting at the same animal at the same time, both causes are ontologically independent, whereas causal overdetermination enters the debate with respect to mental causation in the context of non-reductive physicalism, in which, as we shall see in Chapter 3 (3.5 Non-reductive physicalism), the mental cause of a physical effect is conceptualized as ontologically dependent on the physical cause without violating the

32

From psychology to neuroscience

causal completeness of physics. In other words, systematic overdetermination is a candidate for solving the problem of mental causation within a framework that includes a strong ontological dependence between mental and physical property tokens. Since we will be considering the question of overdetermination later in this chapter (2.4.3 Causal overdetermination), we are going to ground the premise of absence of systematic causal overdetermination here by observing that the presence of overdetermining additional mental causes does not make any distinctive contribution to the effects they supposedly cause, since, by the causal completeness of physics, such physical effects already have a complete physical cause. It seems that the use of Occam’s razor would dispose of overdetermining mental causes. To put it simply, if a physical property token p2 is caused by a physical property token p1, and the p1 is a complete cause for p2, what role could the additional mental cause for p2 play? Let consequently the premise of absence of systematic causal overdetermination of physical property tokens by additional mental causes be formulated as follows: 9 (NO) If mental property tokens cause physical property tokens, there is no causal overdetermination of the physical property tokens in question by complete physical causes and additional mental causes. With this last premise so worded, we can now turn to the formulation of the problem of mental causation. The next section will show that the complex of these four premises constitutes a genuine philosophical problematic by reason of their mutual inconsistency. The last part of this chapter will be then devoted to the presentation of the different possible solutions to this problem.

9

Adapted from Esfeld (2005: 5). This formulation have been chosen to make clear that in the context of the problem of mental causation, the physical property tokens that are possibly causally overdetermined are those that constitute the behaviour. A simple way to designate this particular set of physical property tokens is to specify them as being caused by mental property tokens, as long as we maintain that mental property tokens exists and that they are not epiphenomenon.

The problem of mental causation

33

2.3 Inconsistency of the premises The problem of mental causation arises once it becomes apparent that all four premises cannot be held as jointly true. The conjunction of three of them systematically rules out the fourth. Consider again the four premises: (MC) Mental property tokens cause physical property tokens. (DM) Mental property tokens are ontologically distinct from physical property tokens. (CCP) For any physical property token p, insofar as p has a cause, it has a complete physical cause. (NO) If mental property tokens cause physical property tokens, there is no causal overdetermination of the physical property tokens in question by complete physical causes and additional mental causes. Firstly, consider we take for granted the first three premises. The problem comes down to the logical fact that if (MC) mental property tokens causes the occurrence both of further mental property tokens and of physical property tokens, then we have a picture in which these physical property tokens must have a physical cause by (CCP). Since we have said that they have a mental cause as well, we are evidently in violation of our no overdetermination premise (NO). Formally, we have: (MC) ∧ (DM) ∧ (CCP) → ¬(NO) Secondly, if mental property tokens, which are distinct from physical property tokens, by (DM), cause physical property tokens to occur (MC), then by (NO), we have to eliminate the physical cause of the physical property tokens, in violation of the claim that the physical world is causally closed: (MC) ∧ (DM) ∧ (NO) → ¬(CCP) Thirdly, we can revolve the premises another way. If (CCP) holds, and if we also exclude causal overdetermination of physical property tokens by (NO), then mental property tokens that are according to (DM) ontologically distinct from physical property tokens

34

From psychology to neuroscience

cannot be causally efficacious in causing physical property tokens, per (MC). Thus we get: (CCP) ∧ (NO) ∧ (DM) → ¬(MC) Finally, the last logical possibility is the following. If we hold to the premise that mental property tokens cause physical property tokens (MC), and to the premise that the physical world is causally closed (CCP), and to the premise that causal overdetermination is barred (NO), then one has to endorse the view that, contra our distinctness premise, mental property tokens are identical to physical property tokens (¬DM). In the absence of causal overdetermination, it is the only way to ascribe a causal efficiency to mental property tokens in a world governed by the completeness of physics. (MC) ∧ (CCP) ∧ (NO) → ¬(DM) We started out by jotting down four premises that were supported either by common sense, empirical evidence, or metaphysical inference. The four seem fundamental to the problem of mental causation, meaning that any theory of mental causation must address all four premises. Yet it is evident that they cannot all be true in a consistent theory. The formal problem of the relation between the mind and the natural world is to somehow make the core premises consistent. And that will mean giving one of them up. In the next section, we shall introduce a typology of possible solutions to the problem of mental causation in regards of the different choices that we can make to solve this quadrilemma, giving us a classification of the most important modern and contemporary accounts of the insertion of the mind into the natural world.

2.4 Typology of possible positions in philosophy of mind An important remark must be made at this point. This section does not aim to review in detail all possible positions in philosophy of mind with respect to the problem of mental causation. In the modern and contemporary era, there has been at least four centuries of debate, which has produced an enormous library of tractates, treatises, essays

The problem of mental causation

35

and articles and entering into deep historical discussion would completely exceed our objective, which is to set up a coherent reductionist position in philosophy of mind. Thus, I confine myself in this section to showing, given the four premises, what theoretical alternatives have been generated, and why most of them should be firmly rejected. 2.4.1 Interactionist dualism The “interactionist dualists” are impressed, above all, with the importance of saving the distinctness of the mental (DM). This is the strong motivation for dualism. Secondly, they allow for the causal interaction of two ontologically distinct kinds of entities, what justify the label “interactionist”. Taking for granted the absence of systematic causal overdetermination, the dualist’s program commits her to reject the premise of the causal closure of physics (CCP). Historically, the most important example of interactionist dualism is Descartes’ doctrine of the two substances, in which minds are mental substances characterized by the attribute of thought, and bodies are physical substances, characterized by the attribute of extension. Since each substance possesses exactly one attribute, both kinds of substance are strictly distinct. If the attribute of substance is thought, then the substance is mental whereas if the attribute is extension, then the substance is physical. Therefore, thought and extension are mutually exclusive and it follows that no physical substance thinks and that no mind has a spatial extension. The extended substance, the “res extensae” is strictly ontologically distinct from the non extended substance, the “res cogitans”. The argument motivating the strict distinction between the mental and the physical proceed from the famous Cartesian cogito, which is developed through the Descartes’ Méditation métaphyisques.10 Roughly, the argument states the distinction of bodies and minds starting from the contrast between the fact that I can doubt of the

10

For a systematic reconstruction of the argument, see Rozemond (1998: , chap. 1).

36

From psychology to neuroscience

existence of any physical thing, including the existence of my own body, and the indubitable evidence that as far I am experiencing myself as doubting, I cannot be mistaken with respect to the fact that I am doubting. It follows that it is logically conceivable that my mind exists in the absence of my body. Having gained that certainty, Descartes infers the metaphysical possibility that the minds and bodies exist distinctively, leading him to the idea that minds are one kind of substance and bodies are another, individuated by distinct attributes. In other words, the ontological distinctness of mental and physical substances makes it possible, even likely, that mental substance exists in the absence of any physical substance. To put it in a nutshell, Descartes’ argument tries to move from an epistemic privilege, my indubitable certainty of my actual doubting, to a metaphysical conclusion, the ontological distinctness of my mind and my body. This argument, which is at the root of the Cartesian doctrine, faces a central difficulty. It is generally admitted that the conceivability of mental substances existing independently of physical substance entails the logical possibility of such a situation. However, if the move from conceivability to logical possibility is correct, the move from logical possibility to metaphysical possibility is highly debatable, especially if the goal is to reach an actual metaphysical distinction between mental and physical substances. The argument relies entirely on the Cartesian model of epistemic access to reality, which takes for granted that if you can have clear and distinct ideas of two entities e1 and e2, then the conceivability of the existence of e1 independently of e2 implies the metaphysical possibility that e1 can exist independently of e2 and vice versa. One can argue, however, that as distinct our ideas of minds and bodies are, we cannot be sure on that unique basis firstly, that our ideas correspond to what is metaphysically possible in our world and secondly, even less that it is actually the case that minds do exist independently of bodies in our world. Descartes has at best proven that minds and bodies are logically independent, but further arguments are required to make the move from that logical possibility

The problem of mental causation

37

to metaphysical possibility, and then again we need a transition to whether this possibility is actualized in our world.11 Independently of the internal difficulties proper to this argument, the most central and famous objection to the dualism of substances is that the causal interaction of both kinds of substance seems inexplicable. Physical changes cause changes in the mind, which cause physical changes in return. For instance, “when the nerves in the foot are violently and unusually stimulated, their movement, transmitted through the spinal cord to the inner parts of the brain, there gives a signal to the mind to experience a certain sensation, namely a pain experienced as being in the foot. By this the mind is stimulated to do its best to remove the cause of the pain, as being damaging to the foot” (Descartes 2008: 62). The Cartesian account endorses the existence of causal chain containing both mental and physical elements. However, given the strict distinctness of both substances, how does this causal connection work? The first problem is internal to the interactionist accounts of mental causation and has already been pointed out by Princess Elizabeth of Bohemia, who objects in letter of May 1642 as follows: “How the mind of a human being can determine the bodily spirits [i.e., the fluids in the nerves, muscles, etc.] in producing voluntary actions, being only a thinking substance. For it appears that all determination of movement is produced by the pushing of the thing being moved, by the manner in which it is pushed by that which moves it, or else by the qualification and figure of the surface of the latter. Contact is required for the first two conditions, and extension for the third. [But] you entirely exclude the latter from the notion you have of the body, and the former seems incompatible with an immaterial things”. From Garber (2001: 171).

11

For a contemporary example of a similar debate, see the argument of Kripke (1980) against type-identity and the answer of Levine (Levine 1993: 123). See also Chapter 3, section 3.1 Classical type-identity

38

From psychology to neuroscience

The challenge addressed by Princess Elisabeth to Descartes focuses on the requirement of a physical contact between two substances in order for the first to transmit movement to the second. This requirement is acknowledged by Descartes himself, since he wrote in its second mediation that bodies “can be moved in various ways, not by itself but by whatever else comes into contact” (Descartes 2008: 19). But for entities to be in contact supposes that those entities are extended and localized in space. Now, according to Descartes, the mind is pre-eminently not an extended substance and is by definition not localized in space. Princess Elisabeth’s question then is perfectly well grounded. It seems that the Cartesian doctrine actually excludes mental causation, given a metaphysical framework that takes the mind to be a non-extended substance, while taking mechanical contact of some kind to be a necessary requirement for causation.12 The second difficulty is more general and is relative to the premise of causal closure of the physical world. As said above, Descartes already had a notion of conservation of quantity of movement but this principle was limited to a scalar value and did not take into account the conservation of the overall momentum. This explains his attempt to account for the interaction between the mind and the body in terms of changes of direction of particles in the Pineal gland. However, more recent formulations of the principle of causal

12

It seems in fact that Descartes himself was aware of the gravity of the problem. And it seems that his response, grounded in the theory of the union of the mind and the body already convinced very few of its contemporaries. It seems even that Descartes himself became close of admitting defeat: “The soul is conceived only by the pure intellect; body (i.e. extension, shapes and motions) can likewise be known by the intellect alone, but much better by the intellect aided by the imagination; and finally what belongs to the union of the soul and the body is known only obscurely by the intellect alone or even by the intellect aided by the imagination, but it is known very clearly by the senses. […] It does not seem to me that the human mind is capable of forming a very distinct conception of both the distinction between the soul and the body and their union; for to do this it is necessary to conceive them as a single thing and at the same time to conceive them as two things; and this is absurd” (28 June 1643, AT III:693, CSMK 227). From (Lokhorst 2009).

The problem of mental causation

39

completeness have dramatic consequences on this account. If minds are ontologically independent of bodies, and if one holds that the physical world is causal complete and judges that overdetermination is not a receivable option, then one cannot account for the causal interaction that is supposed to happened between the minds and the bodies. Note that the latter point does not simply count against the Cartesian account and the distinction of substances. The Cartesian ontological framework makes the distinction between substances striking, since what we take today to be properties, such as on the one hand mass, charge, shape, etc, and on the other hand, beliefs, desires, sensations, etc, are taken by Descartes to be modes of the attributes that characterize any substance as being mental or physical. For instance, squareness is a possible mode of the attribute of extension of a substance, which is consequently physical. And a belief – say the belief, to use an example of Descartes, that the wax one holds is solid, is a mode of the attribute of thought, which characterizes the mental substance. In contrast, we formulated the problem of mental causation in terms of mental property tokens, and in the premise of distinction (DN) we distinguish them from physical property tokens. Under this formulation the world does not instantiate two kinds of substances, but distinct kinds of properties. A dualism of properties would hold that mental property tokens are ontologically distinct from physical property tokens. Obviously, if mental property tokens are distinct from physical property tokens and can have a distinct causal impact on them, then the premise of causal closure of physics has to be rejected. To conclude, let us emphasize that interactionist dualisms, either formulated in terms of substance or in terms of properties, assume a strong ontological distinction between the mind and the body, which leads to an uncomfortable dilemma with respect to the causal relation obtaining between both of them. The root of that dilemma result from the rejection of the causal closure of physics (Esfeld 2005: 6). Given that at least some physical substances or properties are causally determined by non-physical substances or properties, interactionists have either to accept that laws of physics do not correctly describe the former, or that the laws of physics simply do not apply to them. But both terms of the dilemma seem philosophically untenable, in the face

40

From psychology to neuroscience

of the strong empirical evidence for the causal closure of physics. Consequently, interactionist dualisms should be rejected. 2.4.2 Dualism without interaction: parallelism and epiphenomenalism If interactionist forms of dualism cannot solve the problem of mental causation without forcing us to reject the causal closure of physics, we are still left with our intuitive notion that mental and physical property tokens are distinct. One line of argument simply abandons the whole interactionist structure of these positions, and postulates either a psycho-physic parallelism or epiphenomenalism. Both preserve our intuition of distinctness, with the difference that psycho-parallelism, repudiating mental to physical causal relations, allows nonetheless for mental to mental causal relations, whereas epiphenomenalism denies any causal efficiency to mental property tokens. According to parallelism, the mental and the physical are distinct and causally develop in a strictly interdependent way. After all, if mental property tokens do not cause any physical changes, they can nonetheless cause further mental properties to occur without violating the completeness of physics. For instance, my feeling of pain is well and truly the cause of my desire to take aspirin, which then causes the perception of the aspirin tablet in my mouth, etc. But none of these mental property tokens is causally involved in the physical process of bringing my arm to my mouth to swallow the tablet. This effect has as cause previous physical property tokens, pace the completeness of physics. A classical example of parallelist position is the one that Leibniz developed upon seeing the problems produced by the Cartesian account of the mind-body relation. It seems plausible to assume that Leibniz does, in the end, take minds and bodies to be distinct substances (Kulstad and Carlin 2008). Leibniz’s justification of this distinctness lies in the fact that qualitative phenomena such as perception are inexplicable on the basis of mechanistic principles. Given his view on physical causation, characterized by the paradigm of the mechanical contact and the conservation of momentum of movement, Leibniz took minds and bodies to be acting according to distinct laws, staying

The problem of mental causation

41

however perfectly synchronized in virtue of a pre-established harmony grounded in the perfection of God’s creation. The best metaphor provided by Leibniz is certainly the one presenting mental and physical substances working perfectly synchronously just as two clocks may stay synchronised even without causally interacting. This line of reasoning faces however important difficulties. The first one is simply that parallelism radically conflicts with our ordinary experience of the insertion of our mental properties in the physical world, since parallelism simply isolates mental property tokens from physical property token, making our experience of agency illusory while having to admit that physical property tokens have no causal effect on minds. Here, too, according to parallelist views, mind and bodies just appear to interact because sequences of mental property tokens and sequences of physical property tokens are simply synchronized. In others words, they co-vary without any causal interaction. However, this co-variance raises a second difficulty. If minds and bodies operate in synch, but without any causal interaction, how can parallelists account for the fact that patterns of mental property tokens co-vary systematically with pattern of physical property tokens? An interactionist dualist has a straightforward account of the close relation obtaining between mental and physical property tokens. They co-vary because they stand in causal relations. But if we renounce this account because the explanation of this causal interaction doesn’t seem plausible, and we still want mental properties to be causes, the synchronicity between the mental and the physical becomes puzzling. One can hold that the coincidence between the mental and the physical is a matter of contingent brute fact about our world, but this seems excessively ad hoc, especially as the parallelist position was formed in the wake of the rejection of the Cartesian account the mind-body problem precisely because it left us with these unresolvable problems. In place of the causal interaction of the mental and the physical, we get the grand coincidence, presided over by God who, for reasons of his own, has ordained pre-established harmony as his principle in the universe. Leibniz’s appeal to God, in the absence of an argument establishing his existence and justifying the doctrine of the pre-

42

From psychology to neuroscience

established harmony, is no more convincing than the “brute fact thesis”, since this move appears also to be excessively ad hoc. An even greater difficulty with psycho-physical parallelism is that it implies panpsychism (Esfeld 2005: 34-35). Leibniz’s monads are constructed so that mental properties are present in every bit of matter there is in the universe, since a mental property token corresponds to each physical property token occurring in the world, and vice-versa. Again, there is no empirical indication that each electron is percipient. This commitment is completely at odds with our contemporary views on mind and on fundamental physical particles. Finally, if, as Leibniz claimed, the mental property tokens operate causally without interacting with physical property tokens according to purely psychological laws, then it suggests that ideally psychology should somehow be complete in the sense that it could someday, theoretically, explain completely the occurrences of mental property tokens only by reference to other mental property tokens, in the same way that physics is causally complete. This ideal seems the very opposite of procedures in current psychology, scientific or ordinary, which refers to physical property tokens, be it as causes of the occurrence of mental property tokens or as their neurobiological substratum. It is difficult to conceive of a scientific psychology that concerned itself only with mental property tokens. To sum up, the parallelist accounts suffer from an excessively ad hoc character, are committed to a panpsychic cosmology, and have to propose a scientific psychology that would refer to mental property tokens only. These disadvantages explain why parallelism will here not be pursued, and why it has nearly disappeared of the contemporary debate in philosophy of mind.13 Granted the parallelists plunged into absurdity by trying to maintain a substance-based distinction between the mental and the physical while trying to find a way around interactional dualism. It nonetheless remains logically possible that we deny any causal efficiency to mental property tokens, to defend the premise of 13

For a contemporary parallelist account, see Ferber (2003: 138-55).

The problem of mental causation

43

distinction of mental and physical property tokens, preserve the claim of completeness of physics and reject systematic causal overdetermination. According to epiphenomenalism, mental property tokens do have an actual existence in distinction from physical property tokens, but without having any causative power at all. Mental property tokens are taken to be by-products of certain complex arrangements of physical property tokens (Heil 2004: 37). A mental property token of being thirsty neither causes the desire to drink a bottle of beer nor the behaviour of going to get the bottle. The picture drawn by epiphenomenalism is the following. When my body lacks liquid, it causes the occurrence of a particular brain state, and as a by-product of that, the occurrence of a mental property token of being thirsty. But this mental property token itself does not cause any further effect. All the causal work occurs on the physical level: my desire of getting a glass of fresh water is the result of an impulse generated within my body that brings about changes in the brain, which in turn brings about the behavioural pattern consisting in going to get the bottle of beer. The mental property tokens are, so to speak, just along for the ride: they have no kind of effect. The classical metaphor is the case of the temporal parts of the shadow of a moving car. In motion the car generates a temporal sequence of successive complex configurations of physical property tokens, and no matter how we divide this continuum, each temporal slice of the car causes the next one to occur. Incidentally, this also causes the shadow running along side of the car, but the shadow’s temporal slices do not operate one on the other sequentially, nor influence the car’s movement. There are several versions of epiphenomenalism. Some stick to the agenda we have outlined in the last paragraph; more recently, the notion of supervenience (3.3 Supervenience) has given philosophers another way of describing the mental property tokens that are metaphysically necessitated by the physical property tokens. The latter, though, still do all the work. But every version of epiphenomenalism denies causal efficacy to mental property tokens, in order to conciliate the distinct existence of mental property tokens with the causal closure of the physical world.

44

From psychology to neuroscience

Epiphenomenalism can be traced back to Thomas H. Huxley (1825-1895, see Robinson 2009), who called animals, including human beings, “conscious automata” and compared the consciousness to the steam-whistle of a train, which “accompanies the work of a locomotive engine [but] is without influence on its machinery” (Huxley 1898: 240). Huxley believed that the antecedents of animal and human behaviour are exhaustively determined by their physiological composition. Accordingly, mental property tokens exist as accidents that come with the organic machinery. They do not contribute causally to the production of the behaviour. A typical example of contemporary reasoning leading to epiphenomenalism is the so-called Knowledge argument of Frank Jackson (Jackson 1982: 128-30), which has the following form: 1) Mary knows all physical information there is to be known about colours perceptions. 2) Mary never experienced redness. 3) Mary acquires new information about colours perception when she has her first experience of redness. 4) Therefore, all information is not physical. 5) Therefore, physicalism is false. Note that several responses have been addressed to this antiphysicalist conclusion of Jackson.14 But the important point for this section is that Jackson follows the most common line of reasoning that leads to epiphenomenalism. This line of reasoning has two argumentative steps. The first consists in arguing for the endorsement of an ontological distinction between some particular kinds of 14

Especially, Terence Horgan (1984: 150-51) pointed out that this argument uses too broad a notion of physical information, which confuses concepts of physical information as information worded using physical predicates and physical information as information about physical property tokens. If we stick to this distinction, as Horgan argues, this argument does not constitute a refutation of physicalism, since there is no reason to believe that the new information Mary acquires is not about physical property tokens, even if not worded using physical predicates.

The problem of mental causation

45

properties. The second step is to bring to bear the causal closure of the physical. Given that the physical world is causally closed, if we have to recognize non-physical property tokens, we have to treat them as causally inert. Consequently, Jackson argues that epiphenomenalism is a reasonable position with respect to mental property tokens. This solution has not proved popular. The reason is that epiphenomenalism completely denies mental causation simply by removing it from the picture. This total opposition to our intuitive anthropological conception of ourselves leaves us with the question of why, at least, we universally have this illusory experience of agency. To deprive the human subject of agency seems way too great a price to pay for solving the mind-body problem, especially as the starting motivation was to solve the problem of mental causation. But even if we were willing to take the epiphenomenalist’s deal, we would still run into a tangle of theoretical issues. Firstly, the introduction of non-causal entities would allow for an inflation of them without limit. Once we detach ourselves from the fundamental ontological criteria that entities have to have some effect, then we could run wild with forces and spirits. For instance, we could endorse a non-causal, non-observable vital force. Secondly, if mental property tokens do exist without being causally efficient, it is a perfectly reasonable to ask why we should take them seriously. In the ordinary scientific practice, scientists are looking for the causes of a particular phenomenon when trying to explain it. Consequently, it is difficult to see how we could explain anything by making reference to non-causal entities. If to give an explanation of a phenomenon means to make reference to the cause of the phenomenon to be explained within the framework of a scientific theory, then property tokens that fail to cause anything also fail to explain anything or offer themselves for explanation. Thirdly, if, we assume that our non-physical properties are ontologically distinct from physical properties and that the physical world is causally complete, then we face the risk of expanding the epiphenomenalist position to deal with other kinds of property tokens as well. Chemical property tokens and biological property tokens would also turn out to be formally epiphenomenal. This is an argument

46

From psychology to neuroscience

we will pursue in more detail in our next chapter (3.7.5 Token-identity and the layered model of the world). The result of this is that epiphenomenalism should be rejected as a variation on the puzzle that firstly relies on an ad hoc solution to the problem of mental causation motivated by purely conceptual requirements. Secondly, although it exists as a solution in the logical space left by our four premises, it asks too high a price when it asks us to give up our intuitive conception of agency. Finally, the endorsement of non-causal mental properties opens the way either to an ontological inflation in which a lot of non-causal entities have to be accepted or to the absolute rejection of mental property tokens. 2.4.3 Causal overdetermination Given the logical space we have been moving in, the next solutions would seem to come from voiding our premise of absence of overdetermination of behaviour by physical causes and additional mental causes, or, in other words, accepting causal overdetermination. If we rely on our intuitions and judge that mental property tokens are distinct from physical property tokens, and still claim that physics is causally closed and mental properties are causally efficient, seemingly we need to retreat to some kind of causal overdetermination. If we must sacrifice one of our premises, that of the absence of overdetermination seems the least painful, as it is the one about which we have the fewest pre-theoretical intuitions or empirical data. Indeed, the very idea of causal overdetermination exists as an abstract logical possibility of solving the problem of mental causation. Yet the endorsement of causal overdetermination turns out to be not as painless as it seems. Although the current overdeterministic accounts of mental causation (see for instance Harbecke 2008; Loewer 2007) are firmly grounded in the framework of non-reductive physicalism, about which we will have a lot to say in the next chapter, we will here foreground our discussion by showing why overdetermination should be rejected. As we mentioned above, it seems that causal overdetermination does not present us with a superior explanation of mental causation. As our completeness premise already tells us that there is already a complete physical cause for any physical

The problem of mental causation

47

behaviour, the additional mental cause is simply extraneous, a cause with no distinctive force. In that sense, Occam’s razor applies: mental properties have no scientific status as causes, since acknowledging the existence of causes that do not have any distinctive effect fails to explain anything that could not be explained without reference to these causes. Considering mental causes seems thus superfluous. That is why prima facie causal overdetermination leads to the elimination of any kind of psychology, which is obviously a result we would try to avoid.15 Secondly, the endorsement of causal overdetermination is clearly incompatible with the majority of causation theories. If causation is supposed to be the literal production of an effect by a prior cause, then it is not possible for a property token to have more than one complete cause. Since by completeness every physical property token has a sufficient physical cause, if a physical property token p1 is sufficient to bring about the effect p2, then there is simply no room for a mental property token to bring about the occurrence of p2. This argument applies also to the case of the conception of causation as a transfer of a conserved quantity of energy. If any physical property token is the result of a process of transmission of a conserved quantity of energy, and if physics is complete, then mental property tokens cannot be causes – for if they were, we would plunge back into the dualist’s dilemma of trying to fit mental causes into a world in which the conservation of energy holds. Both claim of incompatibility are acknowledge by Lower himself (2007), who favours an approach to causation in term of counterfactual dependency between property tokens, arguing that mental property tokens stand in a distinct counterfactual relation to their physical effect, rather than is the case with the physical cause, which stands as it is in 15

This could be taken as a general argument stating that posited mental property tokens fail to explain anything that is not already explained by physics, since this one is complete. However, this should be specifically read in relation to systematic overdetermination as a solution to the problem of mental causation. In Chapter 6, we will argue that psychology is scientifically indispensable, describing patterns of similarities that physics is unable to grasp.

48

From psychology to neuroscience

virtue of claim of the completeness of physics. Although the viability of overdetermination in a counterfactualist framework is currently disputed (see for instance Sparber 2005), the commitment to a counterfactual theory of causation leads overdetermination to face at least two kinds of major difficulties, both linked to the fact that such accounts of causation are conceived in a Humean metaphysical framework, which is characterized by taking properties to be categorical and to have an intrinsic essence. Given that we will come back to these issues later (4.2.4.1 Troubles with Humeanism), let us be brief here. Firstly, such a framework fundamentally disconnects the nature of properties from the causal relation in which they enter, problematizing how we come to know anything with respect to these intrinsic essences, since our cognitive access to what properties are strongly depend on their causal manifestation (for instance, see Black 2000: 94). The second kind of difficulty is that counterfactual dependency seems to be too weak to do justice to our experience of mental causation. According to the Humean accounts of causation, the fact that a causal relation obtains between two property tokens depends on laws of nature, which depend themselves on the entire distribution of fundamental intrinsic and categorical property tokens in the universe. Now, this distribution of fundamental property tokens is contingent and has to be considered as primitive (cf. Beebee 2006). Consequently, the fact that a causal relation obtains between my desire to drink a glass of beer and my drinking the beer depends on the whole fabric of the universe, which is at odds with our experience of mental causation. Our experience tells us that our mental properties are by themselves sufficient to bring about their behavioural effect. The counterfactual conception of causation used to buttress overdetermination seems to indicate that this experience of agency is not true (Esfeld 2007: 4-5). Our belief that the ‘user illusion’ does not extend that far is a point which we have already emphasized several times in this chapter. In summary, causal overdetermination can be seen at first glance as a solution that fits with our strongest pre-theoretical intuitions with respect to the problem of mental causation. However, once we have a closer look at it, we see that overdetermination is mired in theoretical

The problem of mental causation

49

difficulties that make it unsatisfactory as a solution to the problem of mental causation. Among them figures firstly the fact that it leads to an ontological inflation that is undesirable in regards of Occam’s razor, since it admits several sufficient causes in order to explain a certain behavioural effect whereas one of them would be explicatively sufficient. Secondly, qua being committed to a counterfactual account of causation, overdetermination is subject to the charges faced by any Humean metaphysical framework, which cannot vindicate agency and is subject to important epistemic difficulties (this will be discussed in detail in section 4.2.4.1 Troubles with Humeanism). 2.4.4 Physicalism Given the premises of mental causation, the last possibility to fix the problem of mental causation is to give up the premise of distinction (DM) stating that mental property tokens are distinct from physical property tokens and to endorse the claim that mental property tokens are ontologically identical to physical property tokens. At this stage, it is obvious that this claim is motivated by the primacy accorded to the three other premises of the problem of mental causation. Taken jointly, these constitute the core of what is known as the “causal argument” in the contemporary debate (see for instance Papineau 2002: section 1.2), of which Kim’s supervenience argument (Kim 2005: chapter 2) is an elaborated version. These arguments infer from the premise of causal completeness of physics, the premise of mental causation and the premise of causal exclusion, which is nothing but a reformulation of the premise of absence of causal overdetermination, that mental property tokens are identical with physical property tokens. Such a position is often interpreted as conferring a decisive primacy on the physical constitution of the world and is thus labelled “physicalism”. This label designates a relatively large spectrum of philosophical positions and there is no general agreement about its definition, although it is generally agreed that any physicalist position should entail that “any world which is a minimal physical duplicate of our world is duplicate simpliciter of our world”. This claim grasps what will be described in the next chapter as “global supervenience”, namely that the distribution of fundamental physical property tokens in

50

From psychology to neuroscience

our world fix the distribution of everything that there is in the world in the sense that if one could make a duplicate of our physical world, he would make a duplicate simpliciter of the entire world. Global supervenience does not entail that everything that there is in our world is physical, but does entail that everything there is in our world depends ontologically on the distribution of physical property tokens. In other words, such a minimal physicalist position can countenance non-physical entities as long as their existence is entirely determined by the existence of certain configurations of fundamental physical property tokens and their relations. The different versions of physicalism will be discussed in the next chapter. Let us here simply focus on the striking difficulty that any physicalist position faces and repeat why we should nonetheless tackle the mind-body problem from the causal point of view. Physicalist accounts of the mind-body relation are in trouble precisely because they repudiate the premise of the distinction between mental and physical property tokens. They consequently have to account for the distinctive character of mental properties on the basis of the physical constitution of our world, which is far from obvious. However, as pointed out in the first chapter (1.2 The metaphysical issue of mental causation), any account explaining the distinctive features of mental property tokens has to account for the causal relevance of these features to the production of the behaviour. For instance, there is no point in giving an account of the intentional character of certain mental property tokens if this account does not explain how this intentional character enters into the causal determination of behaviour. In that sense, the goal of developing a powerful account of the way mental property tokens are causally inserted into the physical world is of primary interest. As we will argue in the next chapter, a good account should avoid any ontological distinction between mental and physical property tokens. We will therefore be highly concerned by the different version of physicalism, which will therefore be reviewed all along our development of systematic reductionist position.

The problem of mental causation

51

2.5 Summary and transition This chapter was devoted to sorting through the problem space formed by the four premises that together have made up the way mental causation in the world is construed. In order to formulate this precisely, we briefly went over the ontological background. There, we saw that neither an ontology of the object nor one of Davidsonian events proves to be sufficient, since we want to know how the mental properties of a particular individual are causally relevant to the production of behaviour. Therein lays the motivation for endorsing an ontology of property tokens, defined as Kim-ian events. Against that background, the problem of mental causation has four constitutive premises, namely, the premise of mental causation (MC), the premise of distinctness (DM), the premise of causal completeness of physics (CCP) and the premise of absence of systematic causal overdetermination (NO), which constitute together a genuine philosophical problem because of their mutual logical inconsistency. Each of them is supported by quite reasonable intuitions or by empirical data. The problem space of mental causation is thus, in a way, structured by these four cardinal points, a quadrilemma that has been solved by taking one or another premise out. This gave us four main lines of thought. Accordingly, we developed a brief typology of possible philosophical positions with respect to this formulation of the problem of mental causation. Interactionist dualisms repudiate the premise of completeness of physics; parallelisms and epiphenomenalism reject the premise of mental causation. The proponents of causal overdetermination argue that multiple sufficient causes are acceptable as an explanation of the role of the mental in the world. And finally, physicalism argues for dropping the distinctness premise. We went through the first three positions, showing where each ran into untenable entailments. This left us with physicalism, which does give us an ontological framework that does not run into conceptual difficulties or do violence to our scientific worldview. The next chapter will show how we can do justice to the causal efficacy of mental properties within that framework. We shall therefore progressively examine the different physicalist account of mental causation, in particular the type-identity

52

From psychology to neuroscience

thesis (3.1 Classical type-identity), non-reductive physicalism (3.5 Non-reductive physicalism), token-identity thesis (3.7 The tokenidentity thesis as ontological reductionism), saving eliminative materialism for examination in a later chapter (5.4.1.3 NWR and eliminative materialism).

Chapter 3 ONTOLOGICAL REDUCTIONISM This chapter will examine the contemporary debate about the tenability of certain forms of physicalism in the philosophy of mind, with the aim of proposing ontological reductionism as the general framework within which the problem of mental causation can be explained. Ontological reductionism is the position that ensues from our rejection of the premise of distinctness as outlined in the last chapter – which, positively, is equivalent to the assertion that there is no ontological difference between a causally efficient mental property token and an appropriate configuration of physical property tokens. Since the label “physicalism” encompasses a quite broad range of philosophical positions, we shall restrict our discussion to the principal alternatives. This chapter starts therefore by examining the type-identity thesis as the first approach to the issue of mental causation in the direction outlined in the last section of the previous chapter. As is well known, the multiple realization argument has been used to dispute the pertinence of type-identity. According to the multiple realization thesis, mental and physical property types are not co-extensional which would invalidate any physicalist argument based on the naïve type-identity position. However, even if the historical result of the multiple realizability thesis has been to show that mental property tokens are not type-identical with physical property tokens, we can still reasonably assume that mental property tokens stand nonetheless in an asymmetrical relation of determination with physical property tokens. This is all we need to support the claim that the former supervene on the latter somehow. Along with the claim of causal completeness of physics, the development of supervenience since the seventies has made it a central notion in the preferred mainstream strategy of defending physicalism in the philosophy of mind, namely non-

54

From psychology to neuroscience

reductive physicalism. However, Kim’s supervenience argument actually suggests problems with non-reductive physicalism such that the token-identity thesis should be considered the superior solution to the puzzle of mental causation. We shall end this chapter by characterizing in detail the ontological consequences of the endorsement of token-identity.

3.1 Classical type-identity The initial advocates of the physicalist program in the philosophy of mind, figures like Place (1956), Feigl (1958) and Smart (1959), assumed that the problem of mental causation could best be analyzed in terms of the claim that mental property tokens are type-identical with physical property tokens. Physicalism was identified with the decisive rejection of the premise that mental and physical property tokens are distinct. Behind the physicalist doctrine was the idea that the relation between the mind and the body should be conceived on the model of other identifications in the natural sciences. According to the standard example, the mental property tokens of the type ‘being in pain’ are identical to the physical property tokens of the type ‘c-fibres firing’, just as property tokens of the type ‘water’ are identical to property tokens of the type ‘H2O’. Although the distinction between types and tokens of properties was not yet clearly established for these thinkers, but one can generally say that that mental properties types, for these philosophers, are identical in the strong sense with physical property types, in particular with types of brain states. Type-identity is a metaphysical thesis that aims to establish that it is no contradiction in conceiving of mental property tokens as being nothing more than brain processes. If one takes other descriptions - say the relation between a flash of light and an electrical discharge - that have different linguistic contexts, this difference is not an ontological claim that there are really two different particular entities in the world: “there is one thing, a flash of lightening, which is described scientifically as an electrical discharge to the earth from a cloud of ionized water molecules” (Smart 1959: 146-47). In other words, in spite of the difference of meaning between these two descriptions, both of them refer to the very same set of entities in reality, along the lines

Ontological reductionism

55

of Frege’s classical example, the Evening star and the Morning star. Even if we have two different descriptions that are semantically nonequivalent, both of them refer to the very same thing in the world, namely the planet Venus. Analogously, if I report, say, having an afterimage, what I refer to in the world could also be describable as a certain type of complex brain process. This is the core idea of the typeidentity thesis. From the physicalist point of view, the type-identity thesis does a lot of work: namely, it inserts causally mental property tokens in the physical world as being ontologically identical to complex arrangements of physical property tokens. That way, the puzzle of mental causality neatly reduces to the kind of physical causality we are used to, thereby escaping epiphenomenalism since we have a description that vitiates the need for additional causes that would enter into conflict with the causal completeness of physics. Such an account would have the advantage of being ontologically parsimonious, which did not escape the notice of this generation of physicalists: indeed, Smart argues for his position by appealing to Occam’s razor. Given that certain mental property tokens appear to be highly correlated with certain brain states, from the scientific point of view it seems reasonable to identify the referents of the verbal report of the mental experience with referents of the corresponding brain descriptions. This reduction is, as well, in accordance with the view that nothing irreducibly non-physical subsists in the world, so that the ontology of philosophy of mind is not burdened with defending the existence of mental, non-physical entities, which would in the best case have to be explained in terms of a “dangling” relation to the physical world (see Feigl 1958: 428) or, in the worst case, would simply contradict the scientific image of the world. Finally, since identity is a symmetrical relation, type-identity would allow us to preserve mental property language, or psychological terms, while also allowing us to make true descriptions couched in physical terms. In other words, the framework of type-identity is not meant to deny that mental descriptions refer, but to show that they are ontologically identical with – and therefore reducible to – complex arrangements of physical properties, which are the referents and the truth conditions for mental descriptions.

56

From psychology to neuroscience

In spite of its apparent attractiveness in integrating the mental into the physical world of science, the identity theory as an identity holding between types of properties enjoyed an incredibly short philosophical popularity, even if there are some authors today who still mount arguments in its favour (for instance Bickle 2003: 153). Indeed, already during the 1960s, several major objections have been raised against type-identity. There is no need to catalogue them all; we will confine ourselves to two attacks that will be critically important in the development of the argument in this chapter. Famously, Saul Kripke (1980) made a general criticism of the attempt to describe the relation between mental property types and physical property types according to the model of scientific identification. On the one hand, Kripke argued that scientific identifications are metaphysically necessary, that is, necessary about the essence of the things identified, in spite of the apparent contingency of the scientific discovery process. That water is H2O states a necessary fact about water that is true in all possible worlds, even if the discovery of the elements of hydrogen, oxygen and electrons, etc., is a fact of human history, and thus contingent. Yet, on the other hand, Kripke pointed out that there is a significant dissimilarity between classical scientific identification – for instance of heat to mean molecular motion – and psychophysical identifications – for instance pain with c-fibers firing. This dissimilarity emerges when one attempts to justify the apparent contingency of each kind of identification. In the case of scientific identifications, one can distinguish between, firstly, the metaphysical necessity obtaining between what we name “heat” and what physics designs as “mean molecular motion” and, secondly, the contingency of the relation holding between the sensation of heat and what causes this mental property to occur. The fact is that it is always conceivable that this feeling can be produced in human beings by something else which is neither heat nor mean molecular motion. The point of Kripke is to show that this distinction is not available in the case of psychophysical identifications. If one takes the mental property of being in pain to be type-identical with the brain process of c-fibers firing, then this identification has to be necessary. But we definitely cannot conceive a mental property token that is

Ontological reductionism

57

qualitatively indistinguishable from a sensation of pain and which is not a pain. In the case of psychophysical identifications, the distinction between the essence of the property and its appearance collapses. According to Kripke, mental property tokens are appearance, and appearances are only contingently related to what causes them. Therefore, psychophysical identifications fail to be necessary, and, consequently, it is erroneous to conceive the relation between minds and bodies on the model of scientific identifications. Therefore, the type-identity thesis is false. While Kripke’s argument against the contingency of scientific identifications has been largely accepted, the application he makes to mental properties, namely, that they are ontologically distinct from physical properties, has not put an end to the project of physicalist reductionism. First of all, even if we grant that it is logically possible for pain to be felt without c-fibers firing, this does not establish that it is not in fact the case. Levine argued explicitly against Kripke that “what is imaginable is an epistemological matter, and therefore what imagining pain without c-fibers does establish the epistemological possibility that pain is not identical with the firing of C-fibres. It takes another argument to get from the epistemological possibility that pain is not the firing of c-fibers to the metaphysical possibility, which is what you need to show that isn’t in fact identical to the firing of cfibers” (Levine 1993: 123). The charge is straightforwardly that Kripke’s dualist conclusion tries to decide a matter of material fact on the basis of a purely epistemological argument: in this context, nothing in the argument strictly prevents us from concluding that the mental description “being in pain” and the physical description “c-fibers firing” can possibly refer to the same entity. A second response to Kripke’s argument premises that the essence of mental properties is not their qualitative character – their phenomenological appearance - but rather their typical causal role. Under such an assumption, the Kripkean distinction is perfectly available to the type-identity theorist. If the mental property type of being in pain is defined by its typical causal role, then, on the one hand, it seems that the mental property in this case is necessarily identical with the physical property of c-fibers firing, in as much as both of them

58

From psychology to neuroscience

have the same causal role; while, on the other hand, we can maintain that in some possible worlds we could conceive of a mental property token which shares the phenomenal character of pain, but which, differing completely with respect to its causal role, must be a different type of property. Under this conception of the essence of mental properties, Kripke’s point about the dissimilarity between scientific identifications and psycho-physical identification has disappeared. But this result is nothing more than a half victory for the identity theorist. Even if we refute the Kripkean argument by showing that the critical distinction proper to scientific statements of identity is available to the type-identity theorist, the latter is still committed to the claim that the identities between mental and physical types of properties are necessary. However, as we shall see in the next section, this claim is presumably false, even if we accept that mental properties are individuated by their typical causal role.

3.2 Multiple realization in an ontological context The major objection addressed to type-identity is known as the multiple realization argument, developed principally by Putnam (1967) and Fodor (1974). This argument is still, as we will see later (5.3 Multiple realization in an epistemological context), the master argument with respect to the epistemological reducibility of psychology, and in a more general manner of the special sciences to physics. As such, this argument crucially shapes the approach developed in this book. Let us focus here on multiple realization argument as addressed to the typeidentity thesis, and thereby to its effect in the ontological context in which we have been representing physicalism in the philosophy of mind so far. As we have seen, if mental property tokens are type-identical to complex arrangements of physical property tokens, this identity has to be necessary. This commits the type-identity theorist to the following claim, where P is a type of complex arrangement of physical property tokens, such as a brain state characterized by the firing of the c-fibers, and M is a correlated type of mental property token, such as being in pain:

Ontological reductionism

59

(1)∀x (Px ↔ Mx) The proposition (1), which comprehends what we mean by the “typeidentity thesis”, entails propositions (2) and (3): (2)∀x (Px → Mx) (3)∀x (Mx → Px) Proposition (2) means that for any entity x, if this entity is a token of the complex physical property type P, then, necessarily, x is also a token of the mental property type M. This expression articulates the relation of local supervenience, which will be discussed below in this chapter (3.3.1 Varieties of supervenience). Without entering in more details here, let us simply say that this claim is supported by the fact that two identical brains cannot diverge with respect to their mental properties if we accept that mental property types are individuated through their typical causal role,16 since the two identical brains will have the same physical causal dispositions. If mental properties are individuated relative to the behaviour they produce, both brains’ mental properties would be in close connection with these dispositions, since, otherwise, the mental properties exemplified by one of them would confer extra causal powers to this brain. This would contradict the premise of causal completeness of physics. Therefore, in this context, two identical brains exemplify the same mental properties. Proposition (3) is the target of the multiple realization argument. According to (3), any entity x exemplifying a mental property of type M necessarily exemplifies the complex physical property P as well. This is clearly disputable, at least until we have a clearer grasp of the individuation of mental and physical types of properties. First of all, stated as such, (3) means, according to the standard example, that if an entity is in pain, this entity is necessarily in a type of brain state in which c-fibers are firing, and that it is not possible to be in pain without 16

Of course the situation is quite different if mental property types are individuated in another way. For instance, if mental property types are individuated relatively to their content, then it is at least possible that two individuals that are physically perfectly identical could exemplify different mental properties, as Putnam has famously argued with his twin-earth example (Putnam 1975).

60

From psychology to neuroscience

being in that type of physical state (see for instance Putnam 1967). Consequently, any entity that does not exemplify this type of physical property cannot be in pain. Now, it seems that creatures quite different from the human can be in pain. Mammals that have comparable but different brains, as well as reptiles with very different brain configurations, as well as animals lacking what we would call a brain, can all be in pain. The claim that being in a brain state characterized by the firing of c-fibers is a sufficient but also necessary condition to be in pain seems unreasonably too strong. Secondly, medical history has shown that human individuals with severe neurological abnormalities can still exemplify most of our common mental properties. There are many cases in which neurological functions are relocalized as a consequence of congenital alterations to the brain or traumas or illness or even surgery. Yet they are otherwise normal psychological subjects exemplifying comparable mental properties – thoughts, sensations - in spite of their neurological differences. Moreover, the study of severe brain trauma has shown that the organic structure of the human brain is extremely flexible in terms of the distribution of mental functions. In particular, it has been discovered that intact parts of injured brain can assume the execution of cognitive functions that were previously located in damaged areas. The re-functionalizing of neural activity in these people may well constitute a very important difference with respect to the localization of functions in the normal brain; but these people are certainly able to exemplify mental properties in a comparable manner to normal subjects. Thirdly, if we stick to the functional individuation of mental properties, it is perfectly conceivable that entities having no biological properties at all could nonetheless be in pain. Famously, Lewis (1980) argued that hypothetical Martian aliens, in spite of having some kind of hydraulic minds containing varying amounts of fluids in many inflatable cavities, can exemplify pain as well, in as much as they could exhibit behavioural patterns that are typical of pain. It is also conceivable that in the future some supercomputer will be able to display intelligent behaviour such that we would judge them able to exemplify mental properties, such as remembering, choosing, judging, believing and so on, that are close to our own. Even discounting the

Ontological reductionism

61

more speculative examples, we have enough neurological and ethological evidence to show that (3), as a condition of type-identity, is clearly too strong to lead to an acceptable solution to the problem of mental causation, since insurmountable problems in making reasonable ascriptions of mental properties to entities that are physically different of human beings. The multiple realization argument shows that it is not the case, and especially not necessarily the case, that any entity that exemplifies a mental property of type M also exemplifies physical property type P. This point can be expressed in terms of the extensions of types. The type-identity thesis requires mental property types to be co-extensional with complex physical property types. Consider proposition (1) that asserts that any particular entity exemplifying a mental property of type M also exemplifies a complex physical property of type P and viceversa. Therefore, the set of all entities that exemplify M is the same as the set of entities that exemplify P. The multiple realization argument points to the faulty logic here, since in fact mental property types have a larger extension than the complex physical property types with which they are supposedly co-extensive. Instead, mental property types are co-extensive with an open disjunction of possible complex physical property types (Fodor 1974: 108), as stated by proposition (4): (4)∀x ( Mx ↔ P1x ˅ P2x ˅ … ˅ Pnx) Proposition (4) entails propositions (5) and (6), which are at the root of the doctrine that will be examined in more details in the next sections, namely non-reductive physicalism: (5)∀x (Mx → P1x ˅ P2x ˅ … ˅ Pnx) (6)∀x (Mx ← P1x ˅ P2x ˅ … ˅ Pnx) Proposition (5) grasps the fact that any entity, which exemplifies a mental property, exemplifies some physical properties as well, even if these physical properties belong to an open disjunction of possible physical property types. This claim is grounded in the conjunction of the premises of mental causation and of causal completeness of physics. If an entity exemplifying a mental property of type M causes a physical behaviour of type B to occur, then, by the premise of

62

From psychology to neuroscience

completeness of physics, this behaviour has a physical cause as well. Therefore, this entity exemplifies one of the different physical property types P1, P2, …, Pn of which the exemplification causes the behaviour B. Stated as such, proposition (6) is desperately weak, since proposition (2) was perfectly acceptable. This statement does however, grasp the idea that exemplifying a given complex type of physical property is a sufficient condition for exemplifying a given mental property type, but it is surely not a necessary condition, which is the lesson of the multiple realization argument. The reasoning behind this proposition is that we can grant the principle of multiple realization, that is, that there may be a disjunction ranging over several possible complex types of arrangement of physical property tokens, so that one of them will bring about a physical behaviour B. Any entity that exemplifies one of these complex physical types will cause a behaviour of type B to occur. This entity thus will exemplify, as well, the mental property type M, which is defined by causing a behaviour such of type B. At this point, it is important to notice that that the original typeidentity thesis is mistaken in taking the relation between types of mental properties and types of physical properties to be symmetrical. This relation is in fact asymmetrical in the following sense. On the one hand, the physical properties exemplified by a particular entity imply that this entity exhibits a certain type of mental properties. Here, physical property tokens appear to determine mental property tokens. On the other hand, if an entity exhibits a given type of mental property, then, by the joint premises of mental causation and the completeness of physics, this entity necessarily exemplifies physical properties as well. However, there is no particular commitment to there being a specific type of physical property this entity exhibits. Let us summarize briefly the first two sections of this chapter. The type-identity thesis can be seen as an initial attempt to discard the dualism between mental and physical property tokens, and instead aims to conceive the relation between both kinds of properties on the model of scientific identification. Type-identity takes mental property types to be identical with physical property types in the same way that we say

Ontological reductionism

63

water is identical with H20. As Kripke has argued, scientific identifications are necessary, even if such statements are matters of discovery a posteriori. Now, the multiple realization argument has convincingly shown that different physical entities can nonetheless exemplify mental properties of the same type. Consequently, for any entity, to exemplify a given type of physical property is not a necessary condition for the exemplification of a particular type of mental property. This result can be read as meaning that the extensions of the respective types of psychology and physics are potentially different, contradicting the requirement of co-extensionality characterizing the classical type-identity thesis. This difference of extension will show up again in the chapter devoted to the reducibility of psychology to neurosciences, where we will discuss the multiple realization argument in an epistemological context (5.3 Multiple realization in an epistemological context). We shall now turn to non-reductive physicalism, which, as we pointed out, took heed of the criticisms directed at the classical physicalist school, mainly by appealing to the concept of supervenience to explain the asymmetrical relation between mental and physical property types. Against this background we shall then present the supervenience argument formulated by Kim, who makes the token-identity thesis a general metaphysical framework for both philosophy of mind and philosophy of science.

3.3 Supervenience The upshot of our presentation of the vagaries of the discussion of the type identity-thesis was to underline the fact that mental and physical properties are asymmetrically related. One the one hand, mental property tokens are somehow related to physical property tokens, since otherwise we would again have to confront the paradox of dualism, which can never coherently explain the causal interaction between mind and body. This motivates the assumption that if an entity exemplifies a certain appropriate arrangement of physical properties, then this entity exemplifies a certain type of mental properties as well (see proposition (2) in the last section). On the other hand, the multiple realization argument shows clearly that mental property tokens are not type-identical to physical property tokens. The claim that if an entity

64

From psychology to neuroscience

exemplifies a given type of mental property it also necessarily exemplifies a certain type of physical property has therefore to be rejected (cf. our rejection of proposition (3) and acceptance of proposition (5) in the previous section). Mental property types are in this sense extensionally distinct from physical property types. The mainstream position in philosophy of mind since the seventies, namely non-reductive physicalism, has been designed to meet these constraints in a way that confer causal efficiency to mental property without contradicting the claim of causal completeness of physics by using the concept of supervenience to grasp the relation between physical and mental property types. Supervenience is a notion that has been imported from moral philosophy to the field of philosophy of mind consequent to the difficulties that emerged in the critique of the classical physicalist program.17 This concept enables us to grasp the asymmetrical relation of determination between two sets of properties A and B in the sense that, for a given entity, the determination of its properties with respect to B entails the determination of its properties with respect to A, without this relation being symmetrical, i.e. without the identity of an entity with respect to A determining any particular identity with respect to B. We borrow the following simplified example to Kim (1993) in order to clarify the general idea of supervenience. Let our two sets of properties to be such that A contains the property F and B contains properties G, H, V, both sets being closed under Boolean operations. If A-properties supervene on B-properties, then the identity of an entity with respect to its B-properties determines its identity with respect to their A-properties. This enables to define the three following relations (Kim 1993: 58-59): (Det1) ∀(x)[Bm(x) → F*(x)]

17

According to Kim(1993: 54-55), the initial formulation of the idea of supervenience is due to G. E. Moore in Philosophical Studies, London, 1922, even if did not use himself the term “supervenience”.

Ontological reductionism

65

where Bm designates any property maximally defined18 in B and F* designates either F or ¬F but not both; (Dep1) ∀(x)[B*(x) ↔ F(x)] where B* designates a disjunction of B-maximal properties; (Dep2) ∀(x)[B#(x) ↔ ¬F(x)] where B# designates another disjunction of B-properties. (Det1) gives us the desired determination relation from the set B to the set A. For any entity, its identity with respect to set B-properties determines its identity with respect to A-properties. This relation is not symmetrical, which is to say that the identity with respect to Aproperties does not imply any particular identity with respect to Bproperties. (Dep1) expresses the idea that for a property in A, there is a sub-set B* of the set Btot of all B-maximal properties that can be constructed on the basis of B, where every element is such that any entity that exemplifies one of them would exemplify the same property in A. In other words, for any entity, if this entity exemplifies a Bmaximal property belonging to B*, then this entity also exhibits the property F. Conversely, if an entity exemplifies the property F, then this entity necessary exemplifies a property belonging to B*. This characterizes the dependency aspect of the supervenience relation. If Aproperties supervene on B-properties, then A-properties depend on Bproperties, i.e. it is not possible to exemplify the property F without instantiating a B-maximal property belonging to B*. (Dep2) expresses the very same idea with respect to relation between ~F and a sub-set B# of Btot. Note that sub-sets B* and B# are non-intersecting sets and that the union of B* and B# is nothing but the set Btot. This enables us to grasp another important aspect of the supervenience relation. If two entities are different with respect to their A-properties, then these entities must be different with respect to their 18

The notion of maximal property with respect to set B designates any consistent property that is entirely determined with respect to every element of B. In the present case, there are eight of these properties, which are mutually exclusive: C ∧ V ∧ H, C ∧ V ∧ ¬H, C ∧ ¬V ∧ H, …, ¬C ∧ ¬V ∧ ¬H.

66

From psychology to neuroscience

B-properties. In our simplified example, if any entity x is F then, according to (Dep1), this entity exemplifies a B-maximal property belonging to B*. If an entity y is ¬F, then, according to (Dep2), this entity exemplifies a B-maximal property belonging to B#. Therefore, any difference with respect to A-properties implies a difference with respect to B-properties even if it is not necessarily true that the identity of an entity with respect to A-properties determines its identity with respect to B-properties. 3.3.1 Varieties of supervenience Keeping this outline of Kim’s thesis in mind, let us tackle a first general formulation of supervenience with respect to mental property tokens and physical property tokens (adapted from Kim 1993: 58): (WSup) Mental property tokens supervene on physical property tokens if and only if, necessarily, for any x and any y in the world w, if x and y are identical with respect to their physical properties, then x and y are identical with respect to their mental properties. This formulation of the relation between mental and physical properties appears to satisfy the constraints that emerged in our discussion of type-identity. On the one hand, mental property tokens are ontologically dependent on physical property tokens in as much as the identity of an individual with respect to its physical properties determines its identity with respect to its mental properties. On the other hand, this formulation is compatible with the result of the discussion of type-identity, since it allows for the fact that two identical individuals sharing all their mental properties can nevertheless be different with respect to their physical properties. Consequently, using supervenience we can grasp the asymmetrical relation of determination that obtains between mental and physical properties, giving us a thesis that is compatible, as type-identity is not, with the multiple realization argument. It is what Davidson seems to have in mind with respect to the relation between mental and physical properties when he says that “although, as I am urging, psychological characteristics cannot be

Ontological reductionism

67

reduced to the others [i.e. the physical characteristics], nevertheless they may be (and I think are) strongly dependent on them” (Davidson 1980: 253). Accordingly, the physical properties an entity exemplifies determines its mental properties, although mental property tokens still remain ontologically distinct from physical property tokens, that is to say, they cannot be identified to physical properties, discarding thereby the classical type-identity thesis. At this point, several questions arise. First of all, one could simply ask why it should be the case that the relation described by the definition (WSup) obtains in the actual world. Let us argue in favour of this claim in the basis of the two following arguments. The first one is closely related to the causal scheme of argumentation we already developed several times. Imagine an individual playing golf, say Jack, who desires to put the ball in the hole on his next try. This desire, in virtue of the premise of mental causation, causes the ball he strikes to follow a certain trajectory. By the premise of completeness of physics, the trajectory of the ball has a complete physical cause as well, which includes John’s body, the composition of the ball, the immediate environment in which it was hit, the physical laws governing motion in the universe, etc. As we pointed out before, if one accepts that the trajectory of the ball has two complete but distinct causes, then there has to be a story showing how both causes are related. According to the multiple realization thesis, mental properties and physical properties exemplified by Jack are not type-identical and therefore the two candidate causes cannot be straightforwardly identified following the model of scientific identification. There is however a close relation between the desire of Jack and the configuration of physical property tokens that caused the ball to be initially impelled to follow a particular trajectory. If we take for granted the individuation of mental properties by their typical causal role in the production of behaviour (cf. 4.2.1 Mental properties as functional properties), then the causal roles of Jack’s mental properties are in close relation with their physical causal dispositions. Otherwise, it would be impossible to ascribe mental properties to agents in order interpret their behaviour and we would be lost in what Davidson has described as global confusion. Thus, in this case, Jack’s mental properties simply have to be properties of which

68

From psychology to neuroscience

the respective causal roles tend to cause this precise effect. In other words, due to causal role individuation of mental properties, the mental properties an individual exemplifies are determined by the physical properties that cause its physical behaviour. Accordingly, two individuals that are alike with respect to their physical properties are alike with respect to the dependency condition of their mental properties. We reach here the claim stated by the definition (WSup). Secondly, as we will see in the chapters devoted to the relation between psychology and neurosciences, an enormous amount of empirical research suggests that mental properties are dependent on physical properties. For instance, the ability of perceiving colours and to mentally represent them is highly correlated with the normal activity of areas V4 of the occipital lobes (Kolb and Whishaw 2003: 332-33). Patients suffering of cerebral achromatopsia, which is characterized at the neurological level by bilateral damages to or the congenital malformation of these areas report that they are unable to perceive colours and that they even dream in shades of gray. They also report that they are unable to remember what colours look like. These two last features distinguish them clearly from patients suffering from lesions of the peripheral nervous system, since these report that they are able to perceive colours. In other words, the mental properties exemplified by these patients appear to be highly correlated with their brain specificities, preventing them to exemplify usual mental properties. In the light of such results, it is reasonable to assume that the mental properties an individual exemplifies depend on the physical properties that characterize him. This argument is obviously grounded in an empirical correlation between a particular type of brain lesion and verbal reports by afflicted individuals of an inability to perceive and mentally represent colours. As such it is a contingent relation that cannot by itself ground the necessity embedded in the definition of the psychophysical supervenience. However, we completely lack contrary evidence according to which two individuals that are perfectly alike could

Ontological reductionism

69

nonetheless exemplify different mental properties.19 From that point of view, this argument in favour of the psychophysical supervenience shares a central feature with Papineau’s argument in favour of the completeness of physics, namely that there is no evidence to the contrary. Note that according to our first argument in favour of (WSup), the case of physically identical individual exemplifying different mental properties is simply metaphysically impossible, as long as we conceive mental properties as individuated by their causal role in a world in which laws of nature are held to be constant. The question then arises as to whether or not a position characterized by the claim that supervenience obtains between mental and physical properties is a physicalist position, given that psychophysical supervenience gives up the argument that identifies mental and physical properties, accepting that both kinds of properties remain distinct. After all, according to Davidson, the world does not contain only physical property tokens, since mental property tokens constitute a genuine and distinct ontological class of fundamental items. Such a position can nonetheless be conceived as a physicalist one in light of the fact that it is based on a claim of, so to speak, “vertical” determination of properties of a certain kind by properties belonging to another kind. Once the physical properties of a given entity are determined, mental properties of this entity are fully determined, i.e. there is no possibility for two physically identical entities to differ from the psychological point of view. Under this conception, physical properties benefit from an ontological privilege in regard to mental properties. What kind of privilege is this? This brings us to the question of the modal strength of supervenience, whether it is weak or strong. As stated, definition (WSup) describes the relation between mental and physical properties in a particular world w, without any commitment relative to other possible worlds. It is perfectly compatible with (WSup) for an individual x in a world w1 to be physically identical with

19

We assume here a functional individuation of mental properties. We shall briefly tackle the externalist issue below.

70

From psychology to neuroscience

an individual y in world w2 while differing in regard of their respective mental properties. It that sense, there is no trans-world metaphysical necessity embedded in this formulation of supervenience, which allows for distinctions between worlds: duplicates of complex configurations of physical property tokens existing in our world might therefore fail to exemplify the same mental properties in others worlds. (WSup) thus indicates that our intra-word relation lacks metaphysical necessity. Let us consequently call (WSup) “weak supervenience”. The lack of metaphysical necessity embedded in weak supervenience makes it the case that a position based on this formulation of supervenience does not satisfy all the requirements of the physicalist thesis. As said above, if we take mental property tokens to be distinct from physical property tokens, but hold as well that the physical properties an entity exemplifies completely determine the mental properties it exemplifies, then we remain within the broad parameters of physicalism. However, weak supervenience surrenders the argument for such a complete determination. It is not possible to speak of complete determination of one set of properties by another set of properties if the determinacy of the subvenient properties does not necessarily determine the supervening properties. This failure implies a major objection to the endorsement of weak supervenience as a satisfactory description of the relation between mental and physical property tokens relative to the causal efficiency of mental property tokens (Sachse 2007: 56-57). Imagine that in our world w1, in which weak supervenience holds, an individual whom we will call Jack exemplifies the mental property m1, believing that it is raining outside, and that this mental property token causes a mental property token m2, for instance the Jack’s desire to take his umbrella when going outside. The mental property token m1 is weakly determined by a complex configuration of property tokens p1, namely Jack’s body when exemplifying m1. The same goes for the mental property token m2, which is determined by another complex configuration of physical property tokens p2. Assume also that p1 is the physical cause of p2. Imagine now a world w2, in which p1 also causes p2 and in which the mental property token m1 also weakly supervenes on p1. Weak supervenience is however compatible with the case in

Ontological reductionism

71

which no mental property token supervenes on p2 in w2. Now if we compare world w1 with world w2, both worlds are alike with respect to their physical properties. However, the mental property token m1 fails to cause m2 to occur in w2, with the result that there is dissimilarity with respect to Jack’s mental properties between w1 and w2. The mental property m1 is clearly an epiphenomenon in w2, since m1 fails to cause any further mental property. This suggests that we should also consider m1 in w1 as an epiphenomenon, because m1 in w1 and m1 in w2 are supposed to be identical and p2 is already a sufficient condition for the occurrence of m2 in w1. Weak supervenience leads consequently to the epiphenomenality of mental property tokens, which is an unacceptable result (cf. 2.4.2 Dualism without interaction: parallelism and epiphenomenalism). These considerations motivate the strong supervenience thesis, which holds across comparisons of possible worlds in order to account for the fact that physical property tokens completely determine mental property tokens. Strong supervenience can be phrased as follows: (SSup) Mental property tokens strongly supervene on property tokens if and only if, necessarily, for any possible worlds w1 and w2 and for any x in w1 and any y in w2, if x and y are identical with respect to their physical properties, then x and y are identical with respect to their mental properties. This formulation rules out the problematical cases proper to weak supervenience according to which two physically identical individuals located in two different possible worlds can nevertheless differ with respect to their mental properties. Therefore, the charge of epiphenomenality of mental property tokens does not occur under the assumption of strong supervenience. Strong supervenience appears to be sufficiently strong to capture a genuine determination of mental property tokens by physical property tokens. Strong supervenience is controversial, however, even in the circle of physicalist philosophers, since it appears to be too strong. The objections having created the discord among the physicalists are directed against the local character of both weak and strong

72

From psychology to neuroscience

supervenience. Indeed, as formulated both (WSup) and (SSup) take mental property tokens to supervene locally on the complex arrangement of property tokens constituting an individual. In both cases, the claim is that for any individual x, its intrinsic physical properties determine its mental properties, independently of the environmental context in which this individual is inserted. Under such a conception, mental properties are therefore thought to be intrinsically individuated. However, in the literature, several arguments directed against such an internal individuation of mental properties have been offered. The most famous of these externalist arguments with respect to mental properties is certainly the Putnam’s Twin-Earth thought experiment20 (Putnam 1975), in which we are to imagine a XVIIIth century individual, Oscar, who lives on our Earth in the actual world w1 and, given the limited extent of chemical knowledge of his time, knows nothing about the molecular constitution of water. Oscar has a physically identical counterpart, Twin-Oscar, who lives on Twin-Earth in a possible world w2 and speaks the same language as Oscar. Earth and Twin-Earth are physically perfectly identical except in as much as Twin-Earth water is composed of a molecule of XYZ rather than H2O. Imagine now that Oscar is teleported across possible world to TwinEarth in w2 and drinks a glass of water with Twin-Oscar. The point of Putnam is to argue that when Oscar believes on Twin-Earth that he is drinking water, the content of his belief is false, because “the body of liquid [Oscar] is pointing to bears a certain sameness relation (say, x is the same liquid as y, or x is the same as y) to most of the stuff [he] and other speakers in [its] linguistic community have on other occasion called ‘water’ ” (Putnam 1975: 225). In other words, Oscar believes

20

The original formulation of the argument considers two identical planets, except for the molecular composition of water, in the actual world. There are good reasons to think that there is only one chemical kind that plays the role of water in our world. That is why we formulate here the argument that way. The original argument has also been designed as concerning first of all the content of sentences. However, it can be directly applied to the content of intentional mental properties. See McGinn (1977).

Ontological reductionism

73

implicitly that the water he is thinking about has the same chemical essence as the one he usually drinks on Earth, but this is obviously false, whereas this relation of identity holds in the case of Twin-Oscar, his belief being consequently true. Provided the truth value of the respective beliefs of Oscar and Twin-Oscar differ when they refer to the water of Twin-Earth, the content of these beliefs cannot be the same and, therefore, Oscar and Twin-Oscar do not exemplify the same mental properties. It seems, consequently, that the physical constitution of the referents of intentional mental properties enters in the condition of individuation of their content. As Putnam put it, “‘meaning’ just ain’t in the head” (1975: 226) This result is in obvious contradiction with the local character of both weak and strong supervenience, and explains thereby why strong supervenience does not represent a consensual formulation of physicalism in contemporary philosophy of mind, since it seems to presuppose an internalist semantic of mental properties. As a result of this discussion, an important reconceptualization of the very notion of supervenience led to the idea of global supervenience, which is designed expressly to allow supervenient properties to depend on the environment of the individual, who exemplifies them. Global supervenience can be formulated as follows (adapted from McLaughlin and Bennett 2010): (GSup) Mental property tokens globally supervene on physical properties if and only if, for any worlds w1 and w2, if w1 and w2 are identical with respect to the distribution of their physical property tokens, then w1 and w2 are identical with respect to the distribution of their mental property tokens. This formulation can be rephrased in terms of duplicates of entire worlds in order to clarify the kind of determination adduced by global supervenience. If one duplicates the entire distribution of physical property tokens existing in a world w1 to a world w2, then global supervenience implies that the mental property tokens existing in w1 are present in w2 as well. The proposed (GSup) formulation preserves the usual asymmetry proper to supervenience, since if one duplicates every mental property tokens from w1 to w2, global supervenience remains

74

From psychology to neuroscience

perfectly silent with respect to the physical constitution of w2. Global supervenience is therefore compatible with the multiple realization of mental properties. However, we have reasons to doubt that global supervenience finally gives us the correct conception of the determination of mental property tokens by physical property tokens. First of all, global supervenience is logically weaker that strong supervenience, since strong supervenience entails global supervenience whereas the converse relation does not obtain.21 This weakness leads to our second point. Although supervenience grasps the general idea of a systematic and asymmetric relation of determination of mental properties by physical properties at, so to speak, a world-scale, it is compatible with the case in which two worlds vary in an extremely minimal way from the physical point of view, say one of these two worlds has a single additional molecule of oxygen, whilst having a completely different distribution of mental properties tokens. Moreover, global supervenience allows intra-world variations for individuals, which even weak supervenience blocks. Two physically identical individuals can exemplify completely different mental properties. Consequently, global supervenience appears to allow only a very weak dependency between mental and physical property tokens, putting it in the contrary direction to the causal approach of the individuation of mental property types and to the case from empirical evidence, which all suggest that we need a stronger dependence of mental properties on physical properties to account for mental properties.

21

A simple way to be convinced of this implication is the following. If strong supervenience is true and if every physical systems of world w1 are duplicated into a world w2, then all physical systems in w2 are alike with respect to their mental properties (including system having no mental properties in w1). Now, if the spatiotemporal relations between the physical systems in w1, are respected in this duplication process, then w1 and w2 are genuine physical duplicates, and, by strong supervenience, the distribution of mental properties in both world is also identical (See Kim 1984 p. 168; note that Kim tries to prove, mistakenly, the converse relation in this paper).

Ontological reductionism

75

It is important to note that the argument giving the impulse leading to global supervenience, the Twin-Earth’s thought experiment, is itself debatable, since it assumes, without discussion, that the content of mental properties is their essential feature. First of all, it is not the case that every mental property has a meaningful content. For instance, pain is not a meaningful mental property in that it is not intentionally directed toward the world. Of course, when I am in pain, something caused me to be in pain. But this doesn’t mean that my pain represents something under a certain mode of presentation, which characterizes beliefs and other intentional property tokens. In fact, Putnam’s argument applies only to a sub-set of mental properties, namely propositional attitudes. Secondly, even if the content of mental properties is extrinsically individuated per Putnam, one could interpret the content of propositional attitudes as a relational feature of mental properties grasped by reports of propositional attitudes rather than an argument leading straightforwardly to the extrinsicality of these properties (Kim 1998: 36). Mental property tokens such as beliefs and desires represent actual or possible states of affairs. The content of such properties is usually described using reports of propositional attitudes, which embed a sentence s, whose function is to specify the representational content of the mental property. This sentence s is nothing but the one we would use to describe directly the relevant aspect of a given state of affairs to which a certain mental property is supposed to refer. Now, one should not confuse the characteristics of the mental property itself and those of the proposition that is embedded in a propositional attitude reporting its content, since only the latter are necessarily conceptual and for instance true or false of certain entities in the world. By contrast to the classical Computational and Representational Theory of Mind (for a paradigmatic example, see Fodor 1987), the Map Theory of Mind (Braddon-Mitchell and Jackson 1996; Johnson-Laird 1983; McGinn 1989), but also certain connectionist accounts (Bechtel 2001), do not presuppose that mental representations have a conceptual or linguisticlike structure. According to such accounts, we use linguistic expressions to depict the representational content of our mental properties. Now, given that propositional attitudes embed a proposition

76

From psychology to neuroscience

that we would use to describe what our mental properties represent, we can conceive reports of propositional attitudes as describing the relations obtaining between the world and the content of mental properties. Under such an interpretation, it is the conceptual content of propositional attitude that is extrinsically individuated rather than the representational content of our mental properties (McGinn 1989: 192). Therefore, the fact that the conceptual content of propositional attitudes depends on the environment does not straightforwardly imply the extrinsicality of mental properties. Finally, let us insist on the fact that the extrinsic individuation of mental properties brings up the problem of their causal efficiency with respect to the production of behaviour. If the mental properties depend for their individuation on distal objects, then how can these properties be causally relevant for the production of behaviour, given the fact that causation is a local process (Esfeld 2005: 136). If mental properties are individuated distally, we have no story explaining how they can be causally relevant for the production of behaviour in as much as part of the behaviour has been integrated into just what the mental properties are, with that result that our experience of agency and of naïve psychology, that is, our everyday mode of explaining the behaviour of other agents by means of the ascription of mental properties, would be seriously undermined. After all, this is nothing but what grounds the premise of mental causation, as explained in Chapter 2. Since causation is a local process, then mental properties have to be local properties in order to be able to cause the behaviour (Fodor 1991; Horgan 1993: 573; Kim 1998: 37). As we explained in the introduction (1.2 The metaphysical issue of mental causation), any account of mental content has to be compatible on some level with mental causation, because we expect the ascription of mental content to the agent to explain their behaviour. This is the great pragmatic motivation for a causal approach of mental properties, which understands their local causal role to be their essential feature rather than any phenomenology or content they may possess. Accordingly to this hypothesis, externalism should be rejected because it leads to a theoretical situation in which the essential feature of mental property tokens is entirely obscured. We shall therefore stick

Ontological reductionism

77

to our causal approach to mental properties, and its derived conditions, such as the fact that causal roles are purely local. From this perspective, we can endorse the strong supervenience of mental property tokens on physical property tokens. 3.3.2 Status of supervenience Before moving on, let us emphasize that supervenience is not in itself a solution to the problem of mental causation, since supervenience is not a “deep” explanatory notion (Kim 1993: 167). First of all, if supervenience is a useful conceptual tool to express an asymmetrical dependency between two kinds of property tokens, supervenience fails in itself to explain why this relation holds, simply because supervenience does not provide an account of what relates mental and physical property tokens and how they are related. To give a complete account of the relations of dependence between physical and mental property tokens, we would need answers to those questions. In the absence of such an account, supervenience simply expresses a dependency between different kinds of property tokens that we observe in the world. It is however perfectly reasonable to expect a theoretical explanation of such relations of dependency. As Horgan put it, “supervenience relations, in order to figure in a broadly materialistic worldview, must be explainable rather than sui generis” (Horgan 1993: 556). Secondly, supervenience is perfectly neutral with respect to the question of the ontological identity or distinctness of mental and physical property tokens and is consequently compatible with different positions in philosophy of mind. On the one hand, we can interpret supervenience as the thesis that mental property tokens are nothing “over and above” physical property tokens. In that sense, supervenience is compatible with token-identity thesis (3.7.4 Tokenidentity and supervenience). Mental property tokens are identical to configurations of physical property tokens, in spite of not being typeidentical to them. On the other hand, we can interpret supervenience as taking mental and physical property tokens to be ontologically distinct, in spite of the dependency of the former on the latter. In that case, supervenience is compatible with epiphenomenalism: for any

78

From psychology to neuroscience

individual, its physical properties determine its mental properties, which are ontologically distinct and causally impotent. Supervenience excludes however interactionist dualism, since this assumes that mental property tokens exist independently of physical property tokens. In that case, mental property tokens can cause other mental properties to occur without any corresponding physical change, contradicting the rule of supervenience, according to which it is a sine qua non condition that any change with respect to mental properties of an individual requires a physical change. Supervenience rules also out panpsychism and consequently parallelism because both of them take the relation of determination between mental property tokens and physical property tokens to be symmetrical, contradicting the asymmetrical premise of supervenience. Let us briefly sum up here our considerations with respect to supervenience. We saw that the type-identity thesis as originally propounded turns out to fail in its purpose, in that mental and physical property tokens, by way of the multiple realization argument, are not type-identical. The examination of this result showcased the fact that the mental property tokens stand in an asymmetrical relation of dependency with respect to physical property tokens. This dependency is conceptualized as supervenience. Because the dependency relation can be variously interpreted, there are several formulations of supervenience in the current literature. As we showed, weak supervenience doesn’t give us a strong enough determination of mental properties by physical properties for us to remain within the parameters of physicalism. This motivated us to turn to strong supervenience, which adequately represents this relation as long as we take the essence of mental property tokens to be causal. Due to its “ontological innocence”, strong supervenience is not in itself a solution to the problem of mental causation, nor is it a complete position in philosophy of mind. However, supervenience plays a key role in the mainstream position in philosophy of mind of the second half of the twentieth century. Before we turn to the examination of this position, let us add some further considerations with respect to the other important feature of any physicalist position, namely the claim of completeness of

Ontological reductionism

79

physics. Against that background, we shall then formulate an interpretation of non-reductive physicalism, the latter being the target of the Kim’s supervenience argument.

3.4 The causal completeness: psychology and physics 3.4.1 Completeness of physics The main reasons to endorse the causal completeness of physics have already been presented in chapter two. However, we did not elaborate in that chapter on the further implications of the completeness principle, which will now prove to be important in the next stage of our development. For, as we pointed out, the premise of causal completeness of physics asserts that for any physical property token p, if p has a cause, then it has a complete physical cause. This claim is grounded in the physical principles of conservation and ultimately in the absence of empirical research suggesting that physical property tokens are not universally caused by previous physical property tokens. In other words, this absence of counter-evidence suggests strongly that physicists never have to go outside of the realm of physics when investigating the causes of any physical phenomena, independently of the determinist or probabilistic character of physical causation. In so far as causal relations are concerned, the causal completeness of physics implies the nomological and the explicative completeness of physics. The nomological completeness of physics is the claim that for any physical property token p, insofar as there are laws that apply to p, then there is a physical law that applies to p. This means that as long as physicists are looking for laws that apply to p, they do not need to search outside the domain of physical laws. The causal completeness of physics implies the nomological completeness of physics, because any relation between property tokens that allows us to speak of causal relation obtaining between them is sufficient to ground a causal law-like relation. For instance, if any physical property token p1 causes a physical property p2 to occur, then this causal relation

80

From psychology to neuroscience

is sufficient to assert that there is a law-like connection between property tokens p1 and p2. Suppose now that physics is causally incomplete. In that case, there are physical property tokens that have non-physical causes. Or, if a physical token p2 is caused only by some mental property tokens m1, then, in this case, there would be no physical law according to which p2 occurs, since any law under which the occurrence of a property token is subsumed refers to the cause of this property token. The only law under which p2 can be brought about is a psychophysical law. Physicists searching for a law under which to subsume p2 would have to look outside of the domain of physical laws. Consequently, if physics is not causally complete physics is also not nomologicaly complete. Finally, note that if it turns out that physics is not nomologically complete, then the causal completeness of physics would have to be rejected as well. The reason is that if there is a property token p2 that falls under some non-physical law and not at all under a physical law, this would mean that there is a causal relation between p2 and a causal non-physical antecedent without there being a physical causal antecedent. In this case, the claim of causal completeness of physics would obviously be false. In the same way, causal explanations dependent on the nomological completeness of physical would be vitiated. By explanation, we mean a description which act as a nomological predicate in the context of the laws of a theory. The explicative completeness of physics relies on the nomological completeness of physics, which means that any explanation of a physical property token p2 would assume that physicists would never have to appeal to nonphysical concepts to explain the occurrence of p2. The implication holds because of the fact that any possible causal relation between property tokens in the world must fall under a law-like relation, even if the discovery of exactly how that is the case, i.e., our current scientific understanding, remains incomplete. Thus, to formulate a physical explanation of p2 is synonymous with asserting that p1 caused p2 to occur in virtue of the laws of physics. If physics is causally complete, then we can always assume that there is a physical cause for p2, which

Ontological reductionism

81

is equivalent to saying that there is a physical law under which p2 falls. There is consequently a physical explanation for the p2. But what if we suppose that physics is causally incomplete? As we have said, this would be like saying that some physical property tokens have no physical cause at all, but instead have a purely mental cause. In this case, there is no corresponding physical law under which these physical property tokens could appear as effects; rather, we must adduce some autonomous psychological law. Thus, physics would be explicatively incomplete. Now, if it turns out that physics is not, indeed, explicatively complete, then the implications are clear: physics must be nomologically and causally incomplete as well. This would throw us into a world in which there are physical property tokens that fail to come under a physical law, coming instead under a psychological law. This would finally unravel the causal completeness of physics, as some physical property tokens would be wholly governed by a psychological law. 3.4.2 Incompleteness of psychology Contrary to physics, psychology is causally, nomologically and explicatively incomplete relatively to physics. Let us consider the claim that psychology is not causally complete, which means that for any mental property token m1, insofar m1 has a cause, it will not always have a complete psychological cause. Note that here, “insofar” leaves open the possibility that there is a physical cause for m1. Our two arguments in favour of the incompleteness of the psychological are as follows. Firstly, it can be the case that mental property tokens are caused by purely physical property tokens, having consequently a cause but no mental causes. For instance, suppose that Jack is in pain after having sat on a nail. In such a case, as naive psychologist, we have, in order to describe the cause of the pain of Jack, to go outside the realm of purely psychological causation, since the cause of this pain is not in itself a mental property token. There is therefore no pure psychological law under which the causal relation between the sitting on the nail and the pain comes, since such a law

82

From psychology to neuroscience

would be expressed using both mental and physical terms. Furthermore, there is also no purely psychological explanation of Jack’s pain, which cannot be understood in exclusively psychological terms. Secondly, as we will see in details later (7.1.1 Inter-hemispheric disconnection syndrome), it is not the case that psychology is always able to provide the best explanation of a mental property token. For instance, to return to the example we explored, above, of colour perception, the best explanation of the cause of the inability of a patient suffering of cerebral achromatopsia to perceive colours is not psychological – that is, not something that occurred within the sensation of the colours themselves, or their cognition, and thus referencing only previous mental property tokens – but is instead physiological and neurological, referencing damage to the areas IV of the occipital cortex. In order to explain such cases, psychology must resort to the physical cause, as for instance in the present case, where we must look at the neurobiological property tokens in order to explain psychological effects. The causal incompleteness of psychology is relative to the completeness of physics. Psychology is causally incomplete because some mental property tokens have purely physical causes, to which the psychologist must refer. But if psychology is causally incomplete with respect to physics, that doesn’t mean that it is also incomplete in the same way relative to other sciences. For instance, it is never the case that psychologists have to refer to social property tokens, if any, in order to give an account of the occurrence of some mental property tokens. Of course, we often explain mental properties by referring to the social properties of believing agents, as when we say that such and such a person votes for the left because he belongs to the lower middle class. But in such cases, psychology can always provide an equivalent or even more precise explanation. For instance, the psychologist can say that the left wing voter believes that his preferred party will defend the economical interest of the social class he belongs to. That way, psychology is complete with respect to sociology while being at the same time incomplete with respect to physics.

Ontological reductionism

83

Let us assume on that basis that psychology is not nomologically and explicatively complete. This means that there are mental property tokens that do not have a mental cause and that they do thereby not always comes under a purely psychological law. They are therefore not always explained by making reference to previous mental property tokens against the background of a theory worded using only psychological predicates.

3.5 Non-reductive physicalism Against the background of our tuning of the premises of supervenience in conjunction with our interpretation of the causal completeness of physics, we are now in a better position to understand the intermediate position with respect to the relation between mental property tokens and physical property tokens, namely non-reductive physicalism. Of course, non-reductive physicalism, as the mainstream contemporary position in philosophy of mind as well as in philosophy of science, is not a rigidly homogeneous doctrine; there are numerous variants and tweaks of the position in the literature. However, the label nonreductive physicalism still points to common assumptions shared by various philosophical accounts (adapted from Kim, 2003, 151-158), although this label is sometimes used to designate other positions.22 A minimal list of these assumptions would include: 1) (SSup) Mental property tokens strongly supervene on physical property tokens if and only if, necessarily, for any possible worlds w1 and w2 and for any x in w1 and any y in w2, if x and y are identical with respect to their physical properties, then x and y are necessarily identical with respect to their mental properties.

22

We take here non-reductive physicalism to be a purely ontological thesis rather than a position according to which one the one hand mental property tokens are ontologically identical to configurations of physical property tokens while taking on the other hand psychology to be epistemologically irreducible to physics. We shall consider such hybrid positions later (5.1 Why epistemological reductionism?).

84

From psychology to neuroscience 2) (Irr) Mental property tokens are not reducible to and are therefore not identical to physical property tokens. 3) (MC) Mental property tokens cause physical property tokens.

Non-reductive physicalism is an intermediate position between property dualism and ontological monism, meaning that, on the one hand, strong supervenience affirms a close relation between physical and mental property tokens; and on the other hand, non-reductive physicalism takes mental property tokens to be irreducible to physical property tokens. The first condition establishes that an agent’s mental properties are determined by, and thus ontologically dependent on, his physical properties. This is what confers on non-reductive physicalism its “physicalist” character, since in such a view fundamental physical property tokens are ontologically prior to mental property tokens. The second condition highlights the “irreducibility” in an ontological sense – namely, mental and physical property tokens are “ontologically non-identical”. This claim of ontological distinctness is grounded in the lesson of the multiple realization argument. Since physically different individuals can nonetheless be identical with respect to the mental property they exemplify, these individuals have something ontological in common that is not itself physical. This is taken to ground the ontological distinctness of mental and physical property tokens. This is what confers on non-reductive physicalism its “non-reductive” character. Finally, non-reductive physicalism endorses the claim that mental property tokens are causally relevant, i.e. they have a causative role in making physical property tokens occur, for reasons we have already gone through in the previous chapter. It endorses therefore the premise of mental causation. The picture proposed by non-reductive physicalism is the following. Suppose a mental property token m1 cause a mental property token m2 to occur. These mental property tokens supervene on configurations of physical property tokens p1 and p2 respectively, which stand in a causal relation with each other as well, so that we can say that two causal relations obtain: one between mental property tokens and one between physical property tokens. A way of making

Ontological reductionism

85

sense of this picture is closely related to the endorsement of a counterfactual account of causation. If mental property tokens supervene on physical property tokens, then m1 and m2 stand in a counterfactual dependency that is closely related to the counterfactual dependency that obtains between the respective physical correlates that are the basis of the supervenience relation. We take the regularity between tokens p1 and p2 to be causal if the following counterfactual is true: if p1 had not had occurred, p2 would not have either.23 Thus, if p1 and any alternative supervenience basis p1* for m1 had failed to occur, then p2 and any other supervenience basis p2* of m2 would have failed to occur as well, with that result that m2 would also have fail to occurs. Consequently, it seems that if we place this counterfactual analysis of the relation between m1 and m2 against the background of supervenience, m1 and m2 stand in a counterfactual dependency that allows them to stand in a causal relation. It is according to this schema that we can safely speak of mental to mental causation without violating the principles of physicalism. The same holds if we consider the situation in which a mental property token causes a physical property token to occur directly. In the context of the counterfactual account of causation, it means that if this property token had not occurred, then its physical effect would have failed to occur as well. By supervenience, any mental property token supervenes on a configuration of physical property tokens so that (1) if this configuration of physical property tokens had not occurred, and if, in addition, (2) no other alternative basis of supervenience for the mental supervenient property token had occurred, then we can predict that its physical effect would have failed to occur. The tricky point here consists in the fact that, given that a mental property token necessarily

23

It is a simplified way of presenting counterfactuals to stay as close as possible to the discussion of non-reductive physicalism. The worry of such a Lewisian style of counterfactuals is that it merely expresses a counterfactual dependency or relevance, but does not show that a token is a sufficient condition for causation. Yablo added for this reason an adequacy condition, using the following formulation: if a had not occurred, then if it had, b would have occurred as well (1992: 274).

86

From psychology to neuroscience

has a physical supervenience basis, the counterfactual relation that obtain between the actual supervenience basis of a mental property and its physical effect e will also obtain between the supervening mental property token and the effect e, because mental property tokens and physical property tokens are ontologically dependent.

3.6 The causal argument for the token-identity thesis Against the background of the premise of strong supervenience of mental property tokens on physical property tokens in conjunction with the premise of the causal completeness of physics, we can now formulate a general argument against non-reductive physicalism (cf. Kim 2005: chapter 2). This argument aims to show that non-reductive physicalism collapses into epiphenomenalism with respect to the mental property tokens and that the lone tenable position allowing mental property tokens to be causally efficient is the one assuming that mental and physical property tokens are ontologically identical, except if we are ready to endorse causal overdetermination, against which we have already argued in the last chapter (2.4.3 Causal overdetermination). Let us start by assuming that mental property tokens cause other mental property tokens to occur. For instance, the Jack’s belief that it is raining outside is a mental token m1 and it causes another mental token m2, namely his desire to carry an umbrella when he goes out. Understanding mental properties as being causally efficient in that sense is unproblematic, since the disputed case is the one of the nature of the causal relation obtaining between mental property tokens and physical property tokens. (1) m1 causes m2. Secondly, by premise of supervenience, both m1 and m2 have, respectively, a supervenience basis. More precisely, supervenience entails that for any mental property m1, there is a property p1 such that m1 supervenes on p1. The same goes for m2, with that result we can infer that mental property token m1 supervenes on the physical configuration of property token p1 and that mental property token m2

Ontological reductionism

87

supervene on physical property token p2. Note that given that m1 is different from m2, supervenience implies also that p1 and p2 are different configurations of physical property tokens (p1 ≠ p2). (2) m1 supervenes on p1 and causes m2, which supervenes on p2. Thirdly, it is not possible for m1 to cause m2 without causing the supervenience basis of m2 to occur, as this would mean that mental property tokens could exist that simply fail to supervene on physical property tokens. At this stage, invoking the multiple realization thesis, one could argue that m1 could have caused another supervenience basis to occur than p2. However, independently of all these possibilities, the important point is that there is obviously only one actual basis of supervenience for m2, which we have already labelled “p2”. Consequently, m1 causes both m2 and p2 to occur. (3) m1 causes both m2 and p2. Fourth, by the premise of the causal, nomological and explicative completeness of physics, p2, has a complete physical cause, say p*. There are at least three reasons to admit that p* is identical to p1 and is therefore the cause of p2. First, p1 is nomologically sufficient for p2 because the occurrence of p1 is, by supervenience, sufficient for m1, which is itself causally sufficient for p2, since it is not possible for m1 to cause m2 without causing p2. Secondly, assuming that we track causal relations obtaining in the world using counterfactual analysis, it seems that, according to supervenience, if p1 does not occur (nor any alternative basis of supervenience of m1), then m1 will not occur, with that result that p2, caused by m1, would not occur. Finally, it is reasonable to assume that p1 is the physical cause of p2 by reason of the argument we gave in favour of the supervenience premise itself, namely that p1 is the supervenience basis of m1 because p1 and m1 have the same physical effect p2 . We have therefore two causes for p2. (4) m1 causes p2 and p1 cause p2. As said above, non-reductive physicalism takes mental and physical property tokens to be ontologically distinct. (5) m1 ≠ p1.

88

From psychology to neuroscience

This puts us in a dilemma. If, as we have tried to show, claims for causal overdetermination are not convincing, then we can apply the premise of absence of systematic overdetermination of physical property tokens by additional mental properties tokens (2.2.4 The Premise of the absence of causal overdetermination): (6) p2 is not causally overdetermined by p1 and m1. Consequently, we must pare off one of our causes. In this situation, it obvious that either one gives up the causal efficacy of the mental property token m1, or one allows the causal efficacy of m1 at the price of contradicting the premise of causal completeness of physics. Since we took the premise of causal completeness of physics for nonnegotiable, the causal relation between p1 and p2 has to be favoured with respect to the relation between m1 and p2. The mental property token is consequently relegated onto the status of epiphenomenon. (7) p1 is the cause of p2. (8) m1 is an epiphenomenon. This last stage contradicts the non-disputable premise of mental causation. Taking m1 to be an epiphenomenon would lead us to surrender our anthropological conception of the self as a cognitive agent. Our argument leads therefore at this point to an unacceptable result, especially given the fact that non-reductive physicalism premises mental causation. But our argument seems to lead us to a cul de sac in which our options have dwindled to those we previously rejected, that is – either accepting causal overdetermination or rejecting the causal completeness of physics. At this point we would like to go back to focus on the crucial assumption that mental and physical property tokens are ontologically distinct, which is what has driven the argument into a dilemma composed of a choice between, on the one hand, causal overdetermination, and on the other, epiphenomenalism. Once this is clearly seen, there are no other option than a) to assume that p2 has two distinct sufficient causes and thus accepting causal overdetermination, or b) to exclude one of those causes, namely the mental one, and thus accepting epiphenomenalism. Consequently, the argument gives us a compelling reason to drop the assumption of ontological distinctness

Ontological reductionism

89

between m1 and p1. If m1 is ontologically identical to p1, then there is no reason to admit either that p2 is causally overdetermined or that m1 is an epiphenomenon. There is one single causal relation between p1 = m1 and p2 = m2. (9) (p1 = m1) causes (p2 = m2). Against the background of the ontological identity of physical and mental property tokens, all of our three central premises hold. The causal efficacy of mental property tokens is safe, because qua being identical to physical property tokens, mental property tokens are causally efficient. The premise of causal completeness of physics is also respected, since p2 has a complete physical cause, p1. Finally, there is no reason to assume that p2 is causally overdetermined, because p2 has only one single cause, namely p1 being identical to m1. Moreover, note that the supervenience premise is also respected, since any pair of entities, which are identical with respect to their physical properties, are also identical with respect to their mental properties. Let us briefly sum up these developments. In any world in which there is both strong supervenience of mental property tokens on physical property tokens conjoined with the causal completeness of physics, any causally efficient mental property token is identical to a certain configuration of physical property tokens. To recapitulate, the crucial argumentative steps leading to that conclusion are the following. First, we admit that at least some mental property tokens cause further mental property tokens to occur. Second, by supervenience, mental properties tokens have a physical supervenience basis. Third, a mental property token cannot cause a further mental property tokens without causing its supervenience basis to occur. Therefore, mental property tokens cause physical property tokens to occur as well. Fourth, by premise of completeness of physics, these physical effects have physical property tokens as causal antecedent as well. Fifth, if one takes mental property tokens and physical property tokens to be ontologically distinct, then, sixth: (1) Either we accept that physical property tokens can have more than one complete physical cause, or, in other words, are causally overdetermined; or, (2) one of the causes must be demoted. Seventh, if we agree with the arguments against causal overdetermination, the second option leads us to favour

90

From psychology to neuroscience

the physical cause, under the premise of the completeness of physics. Consequently, eighth, mental property tokens are demoted to epiphenomena. But finally, ninth, we can avoid the recourse to epiphenomena if mental and physical property tokens are ontologically identical, since then there would be no question of having two different kinds of causes. This development constitutes the causal argument for the token-identity thesis.

3.7 The token-identity reductionism

thesis

as

ontological

Within the broad parameters in which some physicalist explanation of mental phenomena is desired, the causal argument for token-identity is compelling. Assuming strong supervenience conjoined with the causal completeness of physics, the case for ontologically identifying mental property tokens with physical property tokens seems, at least given our criticism of non-reductive physicalism, to be the strongest physicalist argument. Moving away from the type-identity thesis because of the multiple realization argument, we move towards the token-identity thesis because it captures the core idea of ontological reductionism, namely that mental property tokens are nothing ontologically special above physical property tokens, which allows us to treat them as causally efficient in the physical world without contradicting the claim of causal completeness of physics. Let us add the followings clarifications with respect to token-identity thesis.

Ontological reductionism M1

91 M2

Psychological descriptions

m2

Mental property tokens

p2

Configurations of physical property tokens

P2

Physical constructed types

Applies to m1 Ontologically identical to p1 Applies to P1

Token-identity thesis. The property token m1 is ontologically identical to the configuration of physical property tokens p1. The same holds for m2 and p2. The causal relations r and r’ are therefore ontologically one and the same causal relation. 3.7.1 Remarks on the concept of ‘identity’ First of all, it is important to clarify how mental property tokens are identical to physical property tokens. Identity statements come in two flavours, like supervenience: the strong and the weak kind. The later is often expressed as perfect or exact similarity, whereas the former is strict identity. Strict identity is the relation of selfsameness expressed by the Leibnizian principle of identity of the indiscernibles. If a is strictly identical to b, then a is b, with the result that a and b necessarily share all of their properties, whatever those properties are. For instance, in the possible world imagined by Jerry Siegel and Joe Shuster, Superman is strictly identical to Clark Kent. They are one and the same individual in this world with that result that the individual we refer to using the name “Superman” shares all of its properties with the individual we refer to using the name “Clark Kent”. In other words, in that world, both proper names refer to one and the same entity. Perfect similarity is a weaker relation. For instance, two red billiard balls b1 and b2 standing at two different locations on a table can be said to be identical. They are perfectly similar in the sense that their

92

From psychology to neuroscience

share at least all of their intrinsic properties. Both of them are red, are spherical, have an equal mass, and so one. Nevertheless, they are not one and the same ball. They are numerically distinct and some of their respective properties are different, for instance they do not share their location on the table. Therefore expression “b1” and “b2” do not refer to one and the same entity. The token-identity thesis is the claim that any mental property token is strictly identical to a certain configuration of physical property tokens. Each mental token and its subvenient configuration of physical property tokens are one and the same thing, which makes true both a mental and a physical description. At this stage emerges an element that will be of primary importance for the subsequent discussion (6.2 Starting point). Both strict identity and perfect similarity are transitive relations. Tokenidentity thesis asserts that a particular mental property token m1 is strictly identical to a certain configuration of physical property tokens p1. However, another mental property token m2, which is perfectly similar to m1, can be strictly identical to another physical property token p2, which, per multiple realization, fails to be perfectly similar to p1. The transitivity requirement holding for both strict identity and perfect similarity is obviously transgressed here. Keeping that point in mind, we will come back to that issue in chapter 6 (section 6.2.2 Implications for the multiple realization argument), arguing that if two mental property tokens are strictly identical to their respective two configurations of physical property tokens, and these latter are not perfectly similar, then these two mental property tokens also fail to be perfectly similar, but are nonetheless similar enough to count as tokens of the same psychological type, pace the multiple realization argument. 3.7.2 The conservative character of strict identity Secondly, it has to be emphasized that token-identity is not eliminative, but rather preserves the symmetrical relation in opting for strict identity. Let me illustrate that by a simple analogy. At some point, it was discovered that the Morning Star was in reality nothing other than the Evening Star – that is, both entities were so to speak strictly identified. But this identification neither eliminated nor “removed”

Ontological reductionism

93

anything from the world. Both the Morning star and the Evening Star exist and they are simply one and the very same entity in the universe. The Morning Star is the Evening Star and vice-versa. Analogously, within the framework of token-identity thesis, any mental property token just is a complex configuration of physical property tokens and, reversely, some configurations of physical property tokens are mental property tokens. The token-identity thesis is consequently a conservative position with respect to mental property tokens and is therefore significantly different from an eliminative materialism that infers, from the conjunction of multiple realization and causal completeness, that no mental property tokens exist in the world (5.4.1.3 NWR and eliminative materialism). Rather, the token-identity thesis is from that point of view the claim that, among entities existing in our world, certain of them verify descriptions belonging to different descriptive modalities. 3.7.3 Token-identity vs. and type-identity Third, the token-identity thesis, being a weaker claim than typeidentity, is compatible with the multiple realization argument. The type-identity thesis is the claim that any entity exemplifying a mental property of type M exemplifies also a certain complex physical property of type P and vice-versa (3.1 Classical type-identity). Our discussion of the decisive criticism of this thesis, which gave rise to the multiple realization argument, showed that any future physicalist theory would have to weaken the claim of identity in the following way: although it is true that any entity having certain physical properties of type P necessarily expresses a mental property of type M, it is not the case that M necessarily implies P (3.2 Multiple realization in an ontological context). The token-identity thesis is the claim that any mental property token is ontologically identical to a configuration of physical property tokens, but without any commitment with respect to the type of those physical property tokens. The token-identity thesis does not therefore posit any bi-conditional relations between types of properties. That way, it allows for the possibility that a mental property token of type M is ontologically identical to a configuration of physical property tokens

94

From psychology to neuroscience

of type P1, whereas some other mental property token of type M is ontologically identical to physical property tokens of type P2, etc. The important difference to emphasize here is therefore that type-identity entails that every mental property token of type M is necessarily identical to a configuration of physical property tokens of the same type P, whereas token-identity just entails that every mental property tokens of type M are necessary identical to configurations of physical property tokens, each of them belonging to possibly different physical types P1 to Pn (see statement (4), 3.2 Multiple realization in an ontological context). 3.7.4 Token-identity and supervenience Fourth, our token-identity thesis is compatible with supervenience, although it suggests interpreting it as a relation between different sets of descriptions rather than a relation between sets of properties. On the compatibility side, it is clear that every element of the definition of strong supervenience we gave above (3.3.1 Varieties of supervenience) is consistent with token-identity. If mental property token m1 is identical to a complex configuration of physical property tokens p1, it is obvious that if another complex configuration of physical property tokens p2 is identical with p1 either in the same world or in another possible world, then p2 must be identical with m1 from the psychological point of view. In other words, duplicating p1 means duplicating a mental duplicate of p1 given that mental property tokens have no ontological existence above the physical configurations of physical property tokens. Token-identity is therefore consistent with supervenience. Similarly, if one were to duplicate the entire distribution of physical property tokens of our world, he would also be duplicating the entirety of mental property tokens. However, according to the token-identity thesis, it is not the case that mental property tokens are asymmetrically dependent on and determined by some other kinds of property tokens. As explained above, mental property tokens are individually identical to configurations of physical property tokens. In other words, there are no ontologically different kinds of property tokens or levels of being. Moreover, strict identity is a symmetrical relation and consequently, if

Ontological reductionism

95

a mental property token m1 is identical to a configuration of property tokens p1, then this mental property token cannot be perfectly similar with another mental property token m2, which is not physically perfectly similar to p1. This contradicts the lesson of multiple realization as interpreted in the ontological context, namely that two mental property tokens can be perfectly similar in spite of being respectively identical to two different configurations property tokens. Both elements suggest interpreting supervenience epistemologically, that is as a relation holding between different sets of descriptions in the following sense. For a configuration of property tokens c, the fact that c falls under a given mental description M implies that c falls under a certain physical description P, and any other configuration of physical property tokens falling under P falls as well under the mental description M. Here, the asymmetry of supervenience is explained in terms of a dependence between the applicability of different descriptions. On the one hand, a given entity that falls under a certain physical description will therefore fall under a certain mental description. But to fall under a certain mental description does not necessarily imply falling under a particular physical description, for this depends as well on the way mental descriptions are conceived. Let us consequently insist, here, premising a certain background epistemological interpretation, that supervenience and multiple realization are entirely compatible with token-identity. Accordingly, both issues have to be discussed as the question of the relations obtaining between two different descriptive modalities. 3.7.5 Token-identity and the layered model of the world These reflections with respect to the status of supervenience and the multiple realization argument against the background of the tokenidentity thesis have some general implications for the so-called “layered” view of the world. It is a commonplace of philosophy of science that our sciences are divided into distinct bodies of theories, each of them using a particular set of descriptions in order to formulate laws that account for the phenomena we observe at distinct levels of organization in nature. There are in nature simple systems such as electrons and atoms, and these are described by physics; then there are

96

From psychology to neuroscience

more complex systems such molecules, cells and individuals, and these are described by molecular chemistry, biology and psychology, etc. Although the notion of levels of organization and the development of an account of their relations is highly controversial,24 let us draw here the following naïve picture. The intuition behind the notion of levels of organization is that our world contains complex systems that are characterized by various degrees of causal and structural organization of their parts. For instance, cells are entities that are composed by highly structured and interacting parts such as molecules, which are themselves systems in turn composed of interacting structures of atoms. Therefore, cells are more complex systems than molecules, since their description in chemical terms would be more complex than the description of one of the molecules contained in the cell, this description having to take into account the complete structure of molecules plus their causal interactions. Similarly, molecules are more complex than atoms, quarks or hypothetical Higgs bosons. This hierarchy of levels of organization is itself mirrored by a hierarchy of levels of descriptions, each of them being devoted to the explanation of a corresponding level of complexity through the formulation of an adequate explanation of the domain of the particular set of entities belonging to that level. Physics enjoys a privilege in this hierarchy as being the lone all-encompassing descriptive modality. Anything existing in any space-time region of the universe is therefore theoretically describable by physics and is governed by laws of fundamental physics. Among the entities describable by physics, some of them are describable by chemistry and some by a smaller subset of

24

There are many different accounts of what a level is and how the different levels are laid out. See for instance Morgan (1923) for a layered metaphysical framework devoted to emergentism; Putnam & Oppenheim (1958) who develop a mereological account of levels that seeks to be compatible with the unification of sciences; Craver (2007: chapter 5) for an account of level restricted to neuroscience; and Kim (2002) for a criticism of the criterion of composition commonly supposed to individuate levels and for a criticism of any ontic conception of the notion of level.

Ontological reductionism

97

chemistry, biology. Entities supporting mental descriptions enter this picture as an even smaller set. As such this picture sketching a linear scale of degree of organization paralleled by a hierarchy of descriptive modalities does not convey any particular ontological commitments. However, it is often associated with the notion of ontological levels of being, especially in the framework of non-reductive physicalism, which takes the different sciences to pick up ontologically distinct kinds of properties that stand in relations of supervenience to the adjacent levels. Given that physics is universal, physical property tokens are the most fundamental and they are therefore the ultimate supervenience basis for any other distinct kind of properties. Roughly, chemical property tokens supervene on physical property tokens, then biological property tokens on chemical property token and, finally, mental property tokens on biological property token. Since supervenience is transitive, this is fully compatible with our previous formulation, according to which mental property tokens supervene on physical property tokens. Let me refer to that ontologically loaded interpretation of the layered view of the world as the “ontic layered model of the world”. However, there are at least three difficulties with this ontic layered model of the world that suggest that the framework of tokenidentity thesis may be the most preferable form of non-reductive physicalism, as the framework does not reify levels by taking tokens of different levels to be ontologically distinct. Firstly and most importantly, after our preceding development, it is clear that the causal argument for token-identity applies to any property token that supervenes on some physical property token whilst being ontologically distinct from it, as long as the premise of causal completeness of physics holds. This is the “generalization argument” (Block 2003: 13840) of Kim’s argument (Kim 2005: chapter 2). After all, according to the ontic layered model of the world, mental causation is just a particular case of downward inter-level causation. We needn’t rehearse the steps of the argument for property tokens of each level. Roughly, for any token of any level that supervenes on some physical property tokens, if this token is ontologically distinct of physical property

98

From psychology to neuroscience

tokens, then the causal argument for token-identity shows that this token is an epiphenomenon. As a consequence, there would be no mental causation, no biological causation, no physiological causation or chemical causation, but rather just physical causation. The generalization argument aims therefore to show that supervenience not only creates trouble for mental causation but also for any causal relation that obtains outside of those between physical property tokens. However, we argued previously that if we give up the premise of irreducibility, nothing prevents mental property tokens from being causally efficient qua being identical with physical property tokens. Since the generalization argument has the same form for any other kind of property tokens, the same lesson applies as well: if we subtract the premise of irreducibility, property tokens of any kinds that supervene on physical property tokens are causally efficient qua identical with physical property tokens. But this is obviously rejecting the “ontic” interpretation of the layered model of the world. Still under assumption of irreducibility, Block’s generalization argument has a second aspect: this kicks in if we suppose that there is possibly no fundamental level in the sense that there is an infinite regress of subvenient property tokens of a level L-1 for each property token of level L. Let us clarify that the causal argument for tokenidentity holds in such a context. First, the premise of supervenience does not require the existence of a fundamental level. Empirical data suggest that chemical property tokens are an entirely adequate base of supervenience for mental property tokens, even though the domain of chemistry is not ontologically fundamental. Therefore, supervenience holds in the absence of a fundamental level. Secondly, it is sufficient for the causal argument for token-identity that the supervenience basis of mental property tokens is causally complete with respect to psychology. Again, chemistry is causally complete with respect to psychology, just in the same sense as we argued that psychology is complete with respect to sociology (3.4.2 Incompleteness of psychology). It is never the case that chemists have to refer to psychological causes in order to explain a chemical phenomenon. Hence, chemistry is complete with respect to psychology without being absolutely complete, since chemistry is not complete with respect to

Ontological reductionism

99

physics. Against that background, the causal argument for tokenidentity holds independently of the existence of a fundamental physical theory. Consequently, if we subscribe to the ontic layered model of the world, then it follows that any causal relation that obtains between two property tokens at a given level L is screened-off by the causal relation obtaining between their subvenient configurations of property tokens. If now there is no fundamental ontological level at all, then there is obviously a regress ad infinitum, with the result that there is no causal relation at all. But, again, if the premise of irreducibility is rejected in constructing the layered model of the world, then the list of all subvenient property tokens are identical to their supervenient property, in which case, to rehearse it one more time, there is one single causal relation between two entities that are hypothetically describable according to an infinity of descriptive modalities. But still, it is not the case that there is no causation at all, as Block’s argument suggests. Thirdly, it is important to note that the place of mental property tokens in the linear hierarchy proposed by the ontic layered model of the world is litigious (Kim 2002: 16). According to the early picture proposed by Morgan (1923), but also by Oppenheim & Putnam (1958), mental property tokens succeed to biological property tokens in the hierarchy of levels that all of them propose. However, if we consider seriously the multiple realizability of mental property tokens, then it can be the case that the set of entities supporting a mental description is only partially included in the set of biological entities, since it is possible for mental entities to completely lack biological properties, although being describable using the tools of chemistry or physics, as might be the case with some hypothetical future supercomputers. From that point of view, it is not the case these levels are ordered as linearly as one might expect at first glance. To put it in terms of supervenience, although any mental property token supervenes on physical property tokens, it is possible that only some of them supervene on biological property tokens whereas other mental property tokens supervene so to speak directly on chemical property tokens. Against the background of such a possibility, if levels of organization are interpreted as levels of being, it is not clear how the psychological level is located within this ontic hierarchy, or if it doesn’t traverse different levels. In other words,

100

From psychology to neuroscience

the mental level would be adjacent to the chemical level, while at the same time in some cases, if we consider the cases of mental property tokens in the organic world, the biological level is supposed to be located between the chemical and the mental level. This observation seem to be more easily integrated is if we consider that there is no ordered level of being but simply sets of entities that are describable by certain descriptive modalities. These different considerations strongly suggest avoiding drawing any ontological conclusions from the layered model of the world and instead interpreting levels purely as epistemological interpretations. As pointed out by Ned Block, Kim’s supervenience argument can be generalized; but contrary to Block’s interpretation, we take it to imply that ontological reductionism has to be understood as a general framework for the philosophy of science enabling us to consider any property token of the so-called special sciences as identical in theory to some configuration of physical property tokens. According to tokenidentity thesis, there is no ontological difference between causally efficient property tokens of different levels. From ontological reductionism we can derive an ontological monism, according to which there is in the end a single level of being that makes true different descriptive modalities, and it is the latter which construct different levels of organization out of what is in reality a single level of being. These different descriptive modalities consequently share their truthmakers. For instance, when a chemist explains some reaction of oxidation, he refers to phenomena that are describable using physical conceptual resources. Similarly, to come back to the particular case that interests us in this dissertation, psychology explains phenomena that are also describable from the point of view of neuroscience, which is, as we shall argue in the next chapter, devoted to the study of the physical mechanisms that are causally relevant for the production of behaviour.

3.8 Summary and transition We argued in the previous chapter that the problem of mental causation involves the premise of rejecting the distinction between physical and

Ontological reductionism

101

mental property tokens. The type-identity thesis can be interpreted as the first attempt going in this theoretical direction in order to insert the mind in the natural world. Subsequent criticism made the type-identity thesis untenable, however; the multiple realization argument, which was directed against it in the 1960s, showed that the bi-conditional link between physical and mental property types, modelled on scientific identification, was metaphysically and empirically unpalatable. This leads to the central idea of non-reductive physicalism, according to which mental property tokens, supervening on physical property tokens, retain their ontologically distinction from the latter. We criticized this common thesis against the two background premises of strong supervenience of mental property tokens on physical property tokens conjoined with the causal completeness of physics. The causal argument for token-identity establishes that ontological reductionism, which is immune to the multiple realization argument, holds as the only acceptable solution to the problem of the causal efficiency of mental property tokens in a world described by the natural sciences, given that neither epiphenomenalism nor causal overdetermination constitute satisfactory accounts. Consequently, any mental property token is identical to a complex configuration of physical property tokens, with the more general result that it is not the case that there is a plurality of ontological levels of being. Ontological reductionism is therefore an ontological monism, according to which one and the same level of being makes true different descriptive modalities. Thus, if psychology and neuroscience share their truthmakers, it suggests strongly that both descriptive modalities should be somehow related, since both of them aim to explain within a specific theoretical apparatus the same phenomena, namely the property tokens that are involved in the production of the human behaviour. We shall therefore have a closer look in the next chapter at these theories, before addressing in detail the question of the epistemological relations obtaining between psychology and neuroscience.

Chapter 4 PSYCHOLOGY AND NEUROSCIENCE This chapter will examine in a more detailed way the theories that we use in order to explain the causal origins of behaviour. We shall therefore start by introducing some general remarks about what a theory is and what are the relations between theories, concepts, descriptive categories and property types. Against that background, we shall argue that psychology is a functional theory of mind. Since there are several functionalist positions, we present the main alternatives in order to clarify why common-sense functionalism is the only stance that is able to account for the causal relevance of our ordinary mental properties for the production of behaviour. With respect to the ontological status of mental properties, we shall argue that causal role functionalism is misguided if it takes functional properties to be second-order properties realized by first-order configurations of physical properties. However, if one endorses the claim that functional properties are identical to configurations of physical property tokens, one must reject classical Humeanism, which does not admit genuine functional property tokens. We shall therefore argue in favour of a causal theory of properties, in order to vindicate the identity of ordinary mental property tokens to fundamental configurations of physical properties. In contrast to psychology, neuroscience explains the production of behaviour in terms of the operation of the nervous system. Such is the complexity of the latter that neuroscience builds models of the mechanisms by means of which the nervous system actually produces the behaviour. Without endowing the models with an entourage of elaborate laws directly linking specific brain states to specific behaviours, neuroscientific models nevertheless enable us to understand how a hierarchy of neurobiological mechanisms produces

104

From psychology to neuroscience

human behaviour. Neuroscience, by explaining the capacity of the nervous system to produce behaviour in terms of mechanisms which contain no component or relation that are not, in turn, amenable to mechanistic explanation, reduces the nervous system to the fundamental physical components of mechanisms, which are nothing but physical property tokens. Against the background of the supervenience argument for token-identity, we shall argue that the mechanistic hierarchy outlined by neuroscience is descriptive rather than ontological. Higher level cognitive phenomena, mechanisms accounting for them and their fundamental constitutive components are ontologically identical.

4.1 General background 4.1.1 Theories, concepts and descriptions There are at least two distinct fields that are devoted to the study of mental phenomena and to their implication in the production of the very rich behaviour displayed not only by humans, but also by other living species: psychology, including both scientific and folk psychology, and neuroscience, under which title are bundled together several scientific disciplines devoted to the study of the brain’s functioning. Given this disparity in the parameters defining these fields, scientific and otherwise, some general considerations of the notion of “theory” are called for, in order to provide a broad idea of what we mean by it and how it is articulated with the notion of “property type”. Theories are our epistemological accounts of what obtains in the actual world. Using theories, we are able to explain the phenomena we observe around us, as well in our ordinary life as in the most advanced scientific practices. Theories also enable us to predict the occurrence and trend of their target phenomena, given relevant information concerning the initial conditions under which these phenomena are caused. For instance, each of us daily – and even hourly - uses the resource of folk psychology in order to explain the behaviour of ourselves and other humans acting around us. If John went to the train station to travel to Paris, we presuppose that John wanted to go to the

Psychology and neuroscience

105

train station to travel to Paris, that John believed that the train he took was going to Paris and that the train would satisfy his desire, etc. If Jack indicates that he has these desires and beliefs beforehand (for instance, by saying, “I’m going to Paris tomorrow on the 8:00 train”), one can predict that it is likely that Jack will go to the train station. Any theory consists of a collection of interconnected concepts standing in inferential relations each to the others such that they grasp the causal relations obtaining between property tokens to which they apply. For instance, any thing coming under the concept “having a negative charge” falls within the scope of the concept “has the disposition to attract any entity having a positive charge”. Similarly, folk psychology grasps the relation obtaining between mental property tokens falling within the scope of the concept “believing that Jack is a boy” and mental property tokens falling within the scope of the concept “believing that Jack is a male” by ascribing to the former description the entailment of the latter. Note that the relations obtaining between the different concepts proper to a theory can be of various sorts and that they are highly sensitive to the particular theory under consideration. Concepts are themselves descriptive categories we use in order to classify entities existing in the world. In ordinary life we make daily use of concepts such as “believing that p” or “desiring that q”, ascribing mental properties to agents that are thereby classified against the background of the taxonomy of psychology. Physicists classify entities with respect to their “mass”, “charge”, “quantity of motion” or “molecular structure” and so on. The (correct) application of concepts to entities in the world aims to sort them according to their similarities and differences against the background of the different concepts embedded in a theory. If two entities are similar from the point of view a given theory, then they both fall under the scope of the theory’s concepts. For instance, if John and Jack both desire to take the train to Paris, both of them fall within the parameters of the psychological concept “desiring to take the train to Paris”. This does not imply that they are entirely similar with respect to the psychological classification, for their beliefs about train travel, Paris, and innumerable other matters connected with their train trip could be vastly different. Analogously,

106

From psychology to neuroscience

the physical concept of “having a mass of 8.3g” could be attributed to two things that were otherwise dissimilar with respect to their respective molecular structure, etc. The set of the different descriptive categories proper to a given theory constitutes the taxonomy of this theory. Since distinct theories are made of different, and differently interconnected, concepts, they develop distinct taxonomies and focus on different sets of property tokens. Obviously, the relation obtaining between descriptive modalities devoted to the explanation of the production of human characteristics can be rephrased in term of relation obtaining between psychological and neuroscientific taxonomies: given that psychology and neurosciences share their truthmakers via ontological reductionism, it is perfectly legitimate to expect some close relation between the respective taxonomies of each theory. Let us briefly sum up our views with respect to what theories are. Any theory is a network of interconnected concepts that aims to account for the relations obtaining between occurrences of property tokens of a certain kind. Property tokens are classified through the application of concepts, which are essentially descriptive categories, plus its taxonomy, or the set of the different descriptive categories used by a theory. Since psychology and neuroscience use different conceptual resources, both theories develop different taxonomies in order to explain the causes of the behaviour. 4.1.2 Property types Along with these precisions with respect to the relations between theories, concepts and descriptions, we must specify the status of property types in order to avoid any possible confusion when, later on, we discuss the epistemological reduction of psychology to neuroscience. First, according to ontological reductionism, everything in the world is or is identical to a certain configuration of physical property tokens. Any mental property token of type M is therefore identical to a complex configuration of physical property tokens. The multiple realization thesis tells us that there is no pre-given certainty about the type of this complex configuration of physical property tokens. Nonetheless, focusing on tokens, any entity exemplifying

Psychology and neuroscience

107

mental properties appears to belong to several types of properties, namely at least a mental type and a physical type. Taking a property token to be a token of a certain mental type M or of a certain physical type P depends on the theoretical point of view from which the token is studied. In other words, an entity may appear to be a mental property token of type M or a physical property token of type P depending on whether this token comes under descriptions belonging to the mental or physical taxonomy. This suggests that property types are descriptive categories. These basic comments have a bearing on the multiple realization thesis, in as much as it presents the possibility that two entities e1 and e2 belonging to two completely different physical types can nonetheless belong to the same mental type. For the sake of the argument, suppose that property types are ontological types. In this case, it means that e1 and e2 share an ontological feature, namely their psychological type, in spite of the fact that they share no relevant physical feature. This would however contradict ontological reductionism. Let us recall that ontological reductionism implies that any mental property token is identical to a configuration of physical property tokens. Therefore, neither e1 nor e2 are something above configurations of physical property tokens. This excludes that the psychological type to which e1 and e2 belong can be something ontological, excepted if one claims that mental and physical types are identical. However, due to the multiple realization argument, we have already rejected the type-identity thesis (3.2 Multiple realization in an ontological context). This is why types cannot be something ontological beyond physical types. They are rather descriptive categories, by means of which we classify entities in the world. Consequently, for an entity to be of a certain type does not mean anything else than for that entity to fall within the scope of a certain concept proper to the taxonomy developed by a given theory. A property type is a concept describing a set of entities in the world and its application is made true by any of the member of the set in question. Obviously, the fact that physically different entities can nonetheless fall under the same description of psychology cries out for some further explanations with respect to the individuation of the descriptive

108

From psychology to neuroscience

categories used by psychology. We shall consequently now turn to the individuation of psychological types.

4.2 Folks psychology as a functional theory of the mind 4.2.1 Mental properties as functional properties Functionalism has become the mainstream methodological trend for the interpretation of the relation between the mind and the natural world. It takes mental property tokens to be essentially individuated by their typical causal role in the production of behaviour and of some further mental property tokens. Mental property tokens are therefore functionalised, making psychology a theory using functionally defined concepts in order to account for the production of the behaviour by mental property tokens. However, functionalism is in itself not, as we shall see below, a well defined philosophical position, since there are several contending versions of it that differ concerning the individuation of psychological types and the ontological status of mental property tokens. The disparities in functionalist interpretations point us back to the conditions giving rise to functionalism: the failures of logical behaviourism and classical type-identity, and the emergence of the vast investigation program of the computational cognitive sciences. Without entering into a vast discussion of the history of functionalism, let us however provide the following explanations in order to get clear about the fundamental insight of functionalism with respect to the nature of mental properties. Logical behaviourism, which emerged during the second quarter of the XXth century from the behaviourist research programs in psychology, took the position that minds could be relegated to the status of black boxes in discussing what psychology was really about – behaviour. The lone relevant mental characteristics were taken to be publicly observable features, namely pairings of sensory stimulations and behavioural responses. As an empirical program, behaviourism led to the systematic banning from any theory of mind of any reference to unobservable and non-physical property tokens, like desires, beliefs,

Psychology and neuroscience

109

thoughts, etc. This allowed psychology to concentrate on purely behavioural characterizations of any type of mental property. The motivation for behaviourism was to raise psychology to the rank of an objective science of the human behaviour, comparable to other highlevel sciences, such as chemistry or biology, which dealt solely with observables. As a philosophical position, the central thesis of logical behaviourism is that any meaningful statement containing mental terms can be safely translated into a statement describing the behaviour that any individual would adopt in certain environmental circumstances. Thus, the meaning of mental concepts is equivalent to the meaning of adequate descriptions of stimuli/behaviour pairings. Evidently, behaviourism can be interpreted as a clear physicalist attempt to solve the problem of mental causation, since mental property tokens are nothing “over and above” behavioural responses to environmental stimulations, which can themselves be described without any reference to internal, subjective non-physical property tokens. Mental property tokens are therewith reduced to complex patterns of physical property tokens. According to behaviourists, no immaterial substance or non-physical property token causes or contributes to the production of human behaviour, and mental properties should be translated into talk about behavioural dispositions, which can be empirically observed by varying environmental circumstances. Behaviourism appears to be a full fledged reductionist program. Empirical and philosophical behaviourism were met with devastating objections in the 1960s and 1970s. On the empirical side, the behaviourist investigations were always conducted against the background of normalized conditions, which were justified by the pragmatic need of obtaining regular patterns of stimulations and behavioural responses, no single stimulus giving rise to the same reaction in every possible circumstances. However, the “all thing being equal” conditions, as was pointed out by Chomsky, contained ineliminable reference to other mental property tokens that were taken tacitly to set the norms (cf. Chomsky 1959). The behaviourist program thus was accused of circularity, failing to analyse psychological phenomena in pure behavioural terms. The problem is even graver if

110

From psychology to neuroscience

we reflect on the philosophical project of behaviourism, which aimed to establish a semantic connection between mental terms and behavioural descriptions. As several philosophers argued (for instance Chisolm 1959: chapter 11), it is not possible for any mental property token to give rise to any precise behaviour if it is not assumed that the subject exemplifying this property also exemplifies other types of mental properties. In other words, a single mental property token such as “believing that is it raining outside” or “desiring a beer” will not determine behaviour as long as other mental property tokens are not presupposed. No possible meaning-preserving translation of mental descriptions in purely behaviouristic descriptions can be generated without tacit mentalist assumptions. In this way, the overall project of analysing the mental purely behaviourally notoriously failed. Functionalism inherited from behaviourism the idea that the exemplification of mental properties is closely related to the manifestation of certain behavioural dispositions in certain background conditions, the central issue being the characterization of such conditions. It was in the 1960s that theoretical computational sciences began to provide a major hint as to how the individuation of mental property types in terms of functional causal role can be understood. This constituted the second important source of inspiration for the functionalist paradigm. Historically, the rise of functionalism coincides with formidable progress in computing during the 1950s and 1960s and related applied technologies, which contributed largely to clarify in which sense mental property tokens can individuated by their causal role and how they are related to their material embodiment (see notably Putnam 1960; also Turing 1950). Accordingly, any computable inputoutput function can be computed by an abstract Turing machine, of which the internal states are specified in terms of some possible inputs, some possible outputs and a transition function to some other states of the machine. Roughly, at each step of the computation, the machine reads an input, gives an output as a function of the input plus its current state and finally registers changes of state as requested by the transition function before moving on to reading the input of the next computational step. In other words, the outputs of an abstract Turing machine are given as a strict function of its inputs plus its internal

Psychology and neuroscience

111

current state. From the philosophical point of view, this approach appeared as promising answer to the difficulties that led to the failure of the behaviourist program, by taking behavioural responses to be a systematic function of the environmental stimulation received by a mental system, inasmuch the internal states of the system are taken into account. The analogy between minds and machines is also supported by the relation obtaining between an abstract machine computing a certain algorithm and the concrete physical device that is causally responsible for the execution of the computing process. It dawned on philosophers that there was something philosophically important in the fact that a large variety of physical devices, or many material makes and designs, could implement the same abstract function, i.e. several physical types of devices d1, ..., dn, be they made of vacuum tubes or silicon chips, are capable of generating outputs that correspond to the specification of an given abstract Turing machine T. The phrase ‘multiply realizable’ gained currency when it was understood that the Turing machines, which were taken as models for the mind, could be constructed out of different matter to generate the same outputs. By the strength of the analogy, mental properties could be expressed by individuals exemplifying physical properties of different types, as shown by the discussion of the classical type-identity thesis. What does it mean for a physical device P to “realize” an abstract function F in this context? A complex configuration of physical property tokens P realizes an abstract function F to the extent that the causal behaviour of P is such that P produces outputs whose pairing with inputs match the input/output specification of F. If a mental property type M is causally defined by a causal role φ, a certain type of configuration of physical property tokens P* is a realizer of M to the extent that P* satisfies the causal specification φ. Jointly, these elements firmly grounded the idea that mental property types are definable as abstract causal roles, individuated by their typical causal antecedent, their behavioural effects and some relations to other types of mental property. This opened the way to the mainstream position in analytical philosophy of mind, associated with

112

From psychology to neuroscience

the work of Hilary Putnam, Jerry Fodor and David Lewis,25 according to which mental property tokens are functional property tokens realized in human being and possibly in other type of physical structures. The idea that mental property tokens are individuated by their causal role and that they are realized by physical property tokens formed an entirely new paradigm in the philosophical landscape of the middle of the XXth century. Against the background of old fights between nonreductionists and reductionists, functionalism seemed to be a big breath of fresh air, providing the possibility of studying the mind in abstraction of the compositional characteristics of its possible realizers, following the model of the computational theoretical sciences. Moreover, in a context in which it was not common to distinguish between the ontological reduction property tokens and the epistemological reduction of theories describing different kinds of property tokens, the functionalist approach suggested that one could hold on a physicalist point of view, according to which mental property tokens are second-order functional property tokens supervening on complex configurations of physical property tokens, without committing oneself to the epistemological reducibility of mental property types, as suggested by the multiple realization thesis. This is what obviously gave rise to the idea of the autonomy of special science in general and of psychology in particular, of which Fodor (in particular in his 1974) has been an influential advocate. From that point of view, the overthrow of the type-identity thesis by the multiple realization argument seemed to drive one to functionalism. Although the general idea of functionalism is pretty clear, the name is applied to a vast research program in philosophy of mind that is now subdivided into different approaches. A first source of divergence among functionalists concerns the characterization of the functional causal role defining mental property types. Accordingly, is has become customary to distinguish between “machine state functionalism”, “empirical functionalism” and “common-sense functionalism”. A second disagreement concerns whether to conceive functional mental property tokens, ontologically, as second-order 25

See notably Lewis(1972), Fodor (1968) and Putnam (1975).

Psychology and neuroscience

113

functional properties, or to take functional property tokens as nothing other than the realizers of abstract mental causal roles. In order to clarify the position pursued here, the first distinction will be discussed in the next section, whereas the question of the ontological status of functional property tokens will follow afterwards. 4.2.1 Varieties of functionalism 4.2.1.1 Machine state functionalism A first strand within the functionalist paradigm can be seen as a reinterpretation of the failure of empirical behaviourism on the basis of the development of computational sciences. According to machine state functionalism, minds are abstract Turing machines. The intuition underlying this claim is that if any computable function is computable by a Turing machine, per the Church-Turing thesis, then this applies to the function by means of which the brain produces the behaviour. We must imagine the mind as something like an abstract Turing machine realized by the brain. Although the early protagonists variously defined the precise type of computing machine that would serve as the best model for the human mind,26 let us comment on this approach by taking as an example the simple case of a deterministic finite automaton. A finite deterministic automaton is an abstract machine characterized by a transition function δ which determines, for any possible pair of current state Si and input Ij, the output Ok of the machine and into which other state Sl the machine has to move to wait for the next input, except if Sl is a final state, ending thereby the machine computation of the input string,27 The transition function δ is defined by a table instructing the machine for each state and for any input, which output the machine has to produce and in which state the

26

Putnam for instance began by taking the Turing machine to be the best mind model (1960), and then refined his analogy to the probabilistic automaton (1967).

27

Formally, a deterministic finite automaton is defined as a 5-tuple (Q, Σ, δ, q0, F), consisting of a finite set of states (Q), a finite set of input symbols called the alphabet (Σ), a transition function (δ : Q × Σ → Q), a start state (q0 ∈Q) and set of final states (F ⊆Q). From Sipser (2005: 35).

114

From psychology to neuroscience

machine has move in order to receive the next input. Any state Si of the machine is therefore characterised by a function from inputs to outputs and other states of the machine. Machine states functionalism identifies mental property types with the states of such machines, which allows us to speak of functionally defined type mental states (S1, …, Sn),28 To illustrate this, let us quote Putnam on the functional definition of pain: “The functional state we have in mind is the state of receiving sensory inputs, which play a certain role in the Functional Organization of the organism. This role is characterized, at least partially, by the fact that the sense organs responsible for the inputs in question are organs whose function is to detect damage to the body, or dangerous extremes of temperature, pressure, etc., and by the fact that the “inputs” themselves, whatever their physical realization, represent a condition that the organism assigns a high disvalue […], that does not mean that the Machine will always avoid being in the condition in question (“pain”); it only means that the condition will be avoided unless not avoiding it is necessary to the attainment of some more highly valued goal. Since the behaviour of the Machine (in this case, an organism) will depend not merely on the sensory inputs, but also on the Total State (i.e. on other values, beliefs, etc.), it seems hopeless to make any general statement about how an organism in such a condition must behave; but this does not mean that we must abandon hope of characterizing the condition. Indeed, we have just characterized it.” (Putnam 1979: 438) 28

We took mental property tokens to be the exemplification of a property by an object at a given time (cf. 2.1 Objects, events and properties: preliminary remarks). If an object is a complex entity having different components c1, …, cn constituting a system s, then a state of s is define as a complex Kimian event having component c1, …, cn standing in a n-adic relational property at a time t, that is [(c1, ...., cn, Pn, t].

Psychology and neuroscience

115

“Pain” is therefore that type of state of the mind, caused by some damage to the body, which normally causes a behaviour aiming to avoid being in that type of state, and which is connected with some other states, depending on which the system can behave in a different way to the input, the behaviour being entirely determined by what Putnam call the “Total state” of the system. Surprisingly, this classical passage of Putnam illustrates clearly the idea of machine state functionalism as well as the difficulties with which such early accounts were encumbered. The problem is obviously that any computational state of any kind of machine is a total state of a system, whereas the pain state characterized by Putnam is not. According to his example, a system can be in different states at the same time and the behaviour is generated by inputs and the totality of states characterizing the system at a given time. An automaton, finite or not, deterministic or not, as a Turing machine, is always in a total state. In that sense, machine state functionalism, guided by a mistaken analogy, seems to have an inadequate grasp of the relation between individual mental states and the production of behaviour. As the sense of this inadequacy became general in the philosophical community, machine state functionalism was progressively abandoned in favour of a more comprehensive account of the relation between the production of the behaviour and a plurality of internal mental properties exemplified simultaneously by an individual. 4.2.1.2 Empirical functionalism Empirical functionalism holds that the investigation of causal roles relevant to the production of behaviour is entirely a matter of empirical fact. To be in pain depends not only on exemplifying the typical behavioural patterns that common-sense associate with pain, but more essentially on exemplifying further internal causal roles that jointly process the local sensory inputs in order to produce the motor outputs whose paring is typical of pain (Braddon-Mitchell and Jackson 1996: 80). From that point of view, empirical functionalism can be seen as characterizing mental property types in terms of their role in some scientific cognitive psychological theory or other (Levin 2009). Hence, the precise way of individuating functionally mental property tokens

116

From psychology to neuroscience

depends critically on which psychological theory is taken into account and, thereby, on how the inputs and outputs of the systems are described by the selected theory. There are at least three different worries about empirical functionalism. First of all, if one takes the best theory of the production of human behaviour to be the most fine grained theory we have at our disposal, for instance neurobiology, then the inputs and outputs of the cognitive system are going to be characterized in neurophysiological terms and the relevant “mental causal roles” will consist in a multiplicity of complex computational operations applied to those inputs in the production of behavioural outputs. In such a case, the problem is that any organism, although behaving as if he was in pain, is denied by empirical functionalism to be able to exemplify the property of “being in pain” if its nervous system does not have the human-like internal functional organisation linking sensory inputs to motor outputs.29 The same point holds for the characterization of sensory inputs and motor outputs once we go beyond carbon based sensory organs as described in neurophysiology and deal with, for instance, silicon based sensoria. Therefore, if empirical functionalism is developed in the direction of a very fine-grained theory of the production of human behaviour, it appears to be as anthropocentric as the classical type-identity thesis, literally missing the point underlying the multiple realization thesis. In contrast, we can try to avoid the charge of anthropocentrism by adopting a more abstract way of describing both the internal collection of causal roles generating the behaviour and the sensory

29

At this point, one could argue that this does not constitute a real objection to empirical functionalism, since it is still possible to admit related concepts such as for instance “pain in species S”. However, we shall argue in detail in Chapter 5 that such a move leads to the elimination of abstract psychological types (see especially 5.4.2.2 Multiple realization and functional reduction). In a nutshell, the problem is that the construction of such concepts, which are individuated both in a functional and in a non-functional way, lets unexplained the ability of psychology to provide abstract and homogeneous explanations of entities that are highly heterogeneous from a lower level point of view.

Psychology and neuroscience

117

inputs and motor outputs themselves. Roughly, the general idea here is that the same abstract causal role, which is for instance typical of pain, can be realized by different internal cognitive processes, themselves conceived as distinct causal roles, and that the different possible types of sensory inputs and motor outputs could be described using abstract variables. In such a case, one can imagine a completely abstract description of a functional system whose specification resemble an abstract machine table. However, as Block points out, it seems perfectly conceivable that an abstract machine describing the inputoutput function characteristic of the human behaviour could also describe the general function driving the economy of a small country such as Bolivia. In that sense, such an attempt seems to be clearly too liberal, attributing mental properties to what seems obviously a lack of such properties. In fact, it seems inevitably that a proximal characterization of inputs and outputs of a functional systems, either in very precise physical terms or in very abstract terms will systematically collapse – it will either be too anthropocentric or too liberal (Block 1980: 305). Another serious worry arises from the fact that empirical functionalism focuses on the internal cognitive functions that are causally involved in the production of behaviour within the human cognitive architecture. Empirical functionalism deals with the internal economy of a system whose functional components are cognitive “encapsulated modules”, to borrow Fodor’s words. But such functional components are memory traces, emotional responses, visual integration, action planning, and so on. Prima facie, this creates a gap between what empirical functionalism is focused on and what ordinary psychology, with its categories such as beliefs and desires, describes and explains. Within empirical functionalism, our common-sense mental descriptions refer to internal cognitive causal roles, but focus on the latter obscures the precise characterization of the former.30 30

From that point of view, the Fodor’s famous hypothesis of a language of thought (Fodor 1975)can be seen as a attempt to integrate ordinary mental property tokens at the heart of the human cognitive architecture as propositional attitudes that can be computationally grasped by their syntactic characteristics within a classical computational framework. It is not necessary here to rehearse

118

From psychology to neuroscience

Although empirical functionalism does interest itself in the question of the human cognitive architecture, which is assuredly a primary domain of interest for the philosophy of mind as well as for the philosophy of psychology, the empirical functionalist program seems incapable of delivering a precise account of the causal contribution of the ordinary mental property token to the production of behaviour. In a nutshell, qua focusing on internal causal roles, empirical functionalism fails to provide an account of the relation between these roles and ordinary mental properties. Such an explanation requires a functional approach of these ordinary properties, which is nothing but the issue addressed by common-sense functionalism. In that case, it would be possible to relate the internal cognitive roles to functionalized ordinary mental properties, in order to explain how the former ones implement the latter ones. However, as it stands, it seems that empirical functionalism has just changed the topic and that it consequently fails to account for the way ordinary beliefs and desires play a key role in the production of behaviour, our primary goal (1.2 The metaphysical issue of mental causation). 4.2.1.3 Common-sense functionalism The failure of behaviourist influenced or empirical functionalism did not discredit functionalism so much as help us understand some constraints that shape descriptions of mental property types. First, mental types are neither co-extensional with physical types nor types of behaviour. According to the multiple realization thesis, mental property types are not straightforwardly identifiable with complex physical property types, which are defined specifically in terms of a particular structure and composition. The discussion of logical behaviourism showed that these types cannot be identified with behavioural types connecting environmental stimulations with behavioural dispositions either, since a sufficient psychological explanation of behaviour must

the philosophical argument that largely undermines the hypothesis of such an internal and private language of thought, nor the fact that, as model of cognition in biological systems, the connectionist alternative is supported by strong empirical evidences.

Psychology and neuroscience

119

make reference to other mental property tokens. This motivated the functionalist research project. Our second constraint, taking account of the complexity of mental property types as producers of behaviour, is that we must functionally define mental property types in terms of the possible interactions between multiple tokens of mental properties that occur in order to produce a given behaviour. In spite of the prima facie attractiveness of the computer metaphor, we can’t rely on a one to one model between the mind and a primitive simulation of the mind. It is a mistake to create a model that is only good for describing mono-valent mental states generating some behaviour and causing the mind to mechanically proceed to another monovalent mental state. Of course, mental property tokens do cause further mental properties to occur. But identifying the mind with a single global state at a time t in analogy with a Turing machine, which computes in a sequential way, one total state at a time, fails to grasp the subtle interaction between individual mental property tokens by means of which the human behaviour is produced. A functionalist account should therefore be able to account for the causal efficacy of our desires and beliefs, i.e. our ordinary psychological categories, or be marked as failing to explain how such property tokens contribute, through their mutual interactions, to the production of rich and subtle behavioural patterns. The third functionalist desideratum can be seen as an equilibrium point between excessive anthropocentrism and excessive liberalism in the ascription of mental properties: a valid functional theory of the mind must ascribe mental properties to individuals in some rough correspondence with our intuitions. On the one hand, a functional theory of mental properties should not lead us to think that infants, or large mammals or even possible aliens completely lack of mentality. On the other hand, it should not commit us to ascribe mental properties to entities to which no one would usually ascribe a single belief or desire. The individuation of mental property types should therefore start by taking a close look at the categories of folk psychology, in order to have a standard by which to make our intuition accord with our philosophical theory of how mental property tokens are implicated in

120

From psychology to neuroscience

the production of behaviour; here, we want to stay as close as possible to our habits concerning the domain of individuals who can exemplify mental properties, and finally and most importantly, grasp the categories of mental properties which are of first interest with respect to the problem of mental causation. This is one of the intuitions underlying Lewis’s common-sense functionalism, which confers the status of theory on the discourse of folk psychology. In spite of not being generated from formalized propositions, such as we see in quantum physics, for instance, folk psychology does imply a theory or theories of mental properties mastered by any ordinary human agent that enables him to explain, rationalize and even predict to some extent the behaviour of himself and other individuals. The basic idea is clearly spelt out as follows: “Collect all the platitudes you can think of regarding the causal relations of mental states, sensory stimuli, and motor responses. Perhaps we can think of them as having the form: When someone is in so-and-so combination of mental states and receives sensory stimuli of so-ands-so kind, he tends with so-and-so probability to be caused thereby to go into so-and-so mental states and produce soand-so motor responses”(Lewis 1972: 256). Accordingly, any mental property type is implicitly functionally defined in common-sense terms by its relations with types of certain motor inputs and outputs and its relations to other types of mental properties. In other words, we do not have to engage in potentially endless empirical investigation in order to discover roughly which mental property type is connected to which other one and how they are related to perception and behaviour. Folk psychology, which has evolved through thousands of years and has proved itself in practical terms in all human communities, provides the relevant data for our more sophisticated theories (Lewis 1994: 56). However, even if common-sense psychology is an informal theory, to be useful it has to be made accurate as a method for characterizing mental property types, sensory inputs and motors outputs.

Psychology and neuroscience

121

4.2.2 Common-sense functionalism pursued According to common-sense functionalism, the definitions of mental property types contain the three critical elements of any functional definition, namely, a disjunction of possible sensory inputs, a disjunction of possible outputs and a collection of possible relations to other types of mental properties. Let us start with the characterization of sensory inputs. First of all, taking to heart Block’s lesson relative to empirical functionalism, the “inputs” of the functional systems should neither be characterized purely physically nor proximally. In any case, constructing such detailed physical or abstract characterizations would completely exceed the domain of common sense knowledge. Inputs are therefore things like daylight, chairs, tigers, etc., which cause some perceptual property token to occur. Secondly, the description of the possible causes of any type of mental property token is necessarily disjunctive because there are different environmental conditions that can cause the same type of mental property token to occur, depending on other of the agent’s mental properties. One can form the belief that it is raining by looking out through the window or by listening to the radio. Block’s lessons hold also for behavioural outputs, which, if described solely in terms of nerves firing or abstract computational variables, will bring about the anthropocentric-liberalism dilemma. But common sense doesn’t really suffer from a problem here, for such analytic categories are not part of folk psychological classification. Lewis offers us a conditional way of describing behavioural outputs in terms of environmental impact, which avoids the problem of describing the behaviour in an intentionally loaded way as an action. Hence, behavioural outputs can be described as follows “when you kicked the ball, your body moved in such a way that if you had been on a flat surface in earth-normal gravity with a suitable placed ball in front of you and a suitably placed teammate some distance away, then the impact of your foot upon the ball would have propelled the ball onto a trajectory bringing it within the teammate’s reach” (Lewis 1994: 300, italics in the original). Such a conditional description does not presuppose anything with respect to the kicker’s mental properties nor the physical constitution of the kicker, save for the latter’s foot.

122

From psychology to neuroscience

The description of behavioural outputs of the cognitive system is also disjunctive, since the same mental property token can be involved in the production of different behaviours, depending upon the other mental properties that are exemplified by the agent under consideration. For instance, a person who believes it is raining outside may for that reason take an umbrella when leaving home or simply not leave home, depending on, for instance, his plans for the day. Finally, the functional definition of mental property types must include some description acknowledging the relation between mental property types. The lack of this factor was the reason logical behaviourism was discredited. Instead of bracketing the mind as a black box, functionalism acknowledges that the behavioural outputs of a functional system depend on a two-fold relation both to the environmental inputs of the situation in which behaviour is expressed and the other exemplified mental properties. Since such references are an essential part of the functional definitions, mental property types are systematically inter-defined as standing in a network of relations between environmental causes, behavioural responses and relations to other mental property types. As such, they are positionally defined. Escaping the stalemate of behaviourism, functionalism confronts, here, two problems: circularity – admitting further reference to other types of mental properties might form a logical circle – and the undecided ontological status of the mental properties. There are at least two ways to dissipate the fear of circularity. First, circularity does not always leads to an infinite regress. A functional definition of a carburator, for instance, keeps well within the logically acceptable by defining it in terms of its function in the engine vis-à-vis other parts of the engine. Although such a method of explanation would make it the case that the explanation of any component would refer to the others, the overall explanation made of a description of each component simply lets us understand the contribution of the carburettor to the functioning of the engine as a system. What then is a carburettor? Nothing but an entity that occupies the relevant role in the network of causal interactions. Secondly, a more formal way to get rid of such doubts is provided by what is known as the “Ramsification” of functional

Psychology and neuroscience

123

definition, which aims to define a mental type without any reference to other mental types (Lewis 1972: 253-55). According to Lewis, folk psychology is a theory about the interaction between various types of mental property tokens, behavioural outputs and sensory inputs. For the sake of simplicity, let’s confine the argument to the relations between mental property types. Let T be the postulate sentence specifying the relations between the different mental types in a conjunctive way. Such a theory has the general form: T(m1, m2, m3, …, mn) Every mental property types is then replaced by an existentially quantified variable to form the ramsey-sentence of T: (∃x1)( ∃x2)( ∃x3)…( ∃xn) T(x1, x2, x3…xn) Now, suppose we are defining the property of “having pain”, replacing the bound variable xi. In such a case, we have: y has pain if and only if (∃x1)( ∃x2)( ∃x3)…( ∃xn) [T(x1, x2, x3…xn) and y has xi] The result is that the mental term “has pain” is defined in non-mental terms, showing that there is no vicious circularity in common-sense functionalism. Let me furthermore borrow a simplified example incorporating the typical environmental stimulation and the behavioural effects of mental property tokens from Levin. Imagine that folk psychology contains the following theory Tp of pain: “pain tends to be caused by bodily injury; pain tends to produce the belief that something is wrong with the body and the desire to be out of that state; pain tends to produce anxiety; pain tends to produce wincing or moaning” (Levin 2009). The corresponding ramsey-sentence is something like: (∃x1)(∃x2)(∃x3)(∃x4) Tp(x1 tends to be caused by bodily injury & x1 tends to produce property tokens x2, x3 and x4 & x1 tends to produce wincing or moaning) As a result, we can define “having pain” as follows: y is having pain if and only if (∃x1)(∃x2)(∃x3)(∃x4) [Tp(x1 tends to be caused by bodily injury & x1 tends to produce

124

From psychology to neuroscience property tokens x2, x3 and x4 & x1 tends to produce wincing or moaning) & y has x1]

A generic definition of any mental property type can therefore be provided as follows: y has mental property Φ if and only if (∃x1)(∃x2)(∃x3)… (∃xn) [T((x1 is caused by environmental circumstances c1 ˅ c2 ˅ … ˅ cn) & (x1 stands in relation to x2, x3…xn such that x1R1x2 ∧ x1R2x3, ∧ … ∧ x1Rn-1xn) & (x1 causes behaviour b1 ˅ b2 ˅… ˅ bn) & y has x1)] Accordingly, the definitions in use in folk psychology are embedded in a network of interrelated mental properties, behavioural outputs and environmental stimulations which makes it the case that exemplifying a particular property such as “having pain” means nothing more than standing in the right place in the network; the key is that the network can be described non-mentally. Lewis’s proposed transformation assures us that the circularity obtaining in classical functional descriptions is not vicious. In a more general way, the inter-definition of mental property types proper to functionalism imposes a holistic vision of the mental on the ontological side. This holistic character follows from the functionalist account quite naturally, simply because the idea of an individual exemplifying one single mental property is nonsensical. To imagine an individual believing for instance that there are cool beers in the fridge makes sense only if this belief is endowed with sense by a vast number of other beliefs, such as what beer is, what a fridge is, what cool is, and so on, each belief in turn bringing into view a cohort of other beliefs. Secondly, it is hard to see how isolated mental property tokens could play an active role in the production of the behaviour. Think of an automaton being always in the same state. Such a device would always answer in the same way to the same input. In such a case, the inputs would be nomologically sufficient for the occurrence of the effect, without any distinct causal influence of the state on the behaviour of the automaton. It seems that the space enabling the state to play a distinct causal contribution is lost. The

Psychology and neuroscience

125

same holds for individual mental property tokens within the framework of common-sense functionalism The second problem for functionalism is how to assess the ontological status of functional property tokens. As we saw, logical behaviourism can be seen as a fully fledged reductionist program, identifying on the one hand mental property tokens with tokens of stimulus/behaviour pairs, while, on the other hand, reducing mental property types to types of stimuli/behaviour pairs. As we shall see in the next section, for functionalism, the situation is not so straightforward. 4.2.3 The ontological status of functional property tokens, or functionalism’s dilemma We have seen that the Ramsification of functional definitions enables us to fully define mental property types without using further mental vocabulary, which also holds out the promise that common-sense functionalism might converge with logical behaviourism to justify the reductionist program. However, functionalism actually faces two options with respect to the ontological status of the referents of mental functional descriptions, seeming at first glance to drag the overall functionalist program into an uncomfortable dilemma. On the one hand, “causal role” functionalism takes mental property tokens to be second-order property tokens supervening on and being realized by first-order complex configurations of physical property tokens, which occupy the relevant causal role.31 More precisely, mental property tokens are second order property tokens dependent on first order configurations of property tokens fulfilling a certain causal role. Such a conception takes second-order functional property tokens to be ontologically distinct from first-order physical property tokens. The motivation to endorse this view lays in an ontological interpretation of the multiple realization thesis. According to causal role functionalism, two physically different entities e1 and e2 can nonetheless be identical with respect to their role in the expression 31

Hilary Putnam has been one of main proponent of causal role functionalism (Putnam 1967; Putnam 1975).

126

From psychology to neuroscience

of mental properties. However, to be identical from the point of view of psychology, e1 and e2 have to have something ontological in common, which constitutes the truth condition for the application of a given mental description. The multiple realization thesis excludes the physical characteristics of e1 an e2 from the list of candidates to this ontological commonality. The causal role functionalist’s answer is therefore to say that what both entities have in common is a second order functional property of fulfilling a certain causal role, which is ontologically distinct from physical properties and to which mental description rigidly refer. However, against the background of the supervenience argument for token-identity (3.6 The causal argument for the token-identity), it is obvious that supervening second-order property tokens face the charge of epiphenomenalism, which has proved to be a dead end in the philosophy of mind. The problem is that if a mental second-order property token m1 supervenes on a physical property token p1, which is nomologically sufficient for the occurrence of an effect p2, then the supervenience argument for token-identity shows that either this second-order property token m1 is epiphenomenal with respect to p2, or identical to the physical property token p1. Since causal role functionalists refuse to endorse token-identity for the reasons explained above, it seems that they are driven to conclude that mental property tokens are epiphenomenal, contradicting the premise of causal efficacy of the mental. This unfortunate consequence of causal role functionalism was pointed out by Block (1990) in the early 1990s. On the other hand, “realizer” functionalism takes mental descriptions to refer directly to the configurations of physical property tokens that satisfy the functional characterizations.32 Accordingly, for a complex configuration of physical property tokens p1, to make true a mental description M is nothing but to be embedded in a causal network such that p1 has a cause c, an effect e and some other causal relations to property tokens p2, … pn such that p1 satisfies the overall

32

The main proponent of realizer functionalism is certainly David Lewis (see notably 1966; 1972; 1994).

Psychology and neuroscience

127

functional specification of M. From that point of view, whether or not some configuration of physical property makes true a mental description depends on its causal relations with other configurations of physical property tokens rather than on its internal physical constitution. This is obviously what allows realizer functionalism to be compatible with the multiple realization thesis. Different configurations of physical property tokens can make true the same mental description because the intrinsic features of these physical entities are not what matters for psychology, but rather the extrinsic causal relations in which there are embedded. Mental descriptions are consequently considered as non-rigid designators. The obvious advantage of such a conception is that the causal exclusion problem is bypassed, since it avoids the consideration of distinct kinds of property tokens. Complex patterns of physical property tokens are all that exist, some of them making true mental descriptions. Hence, nothing precludes the referent of mental descriptions to be causally efficient. Against the background of our development of Chapter 3, this seems to be a more adequate metaphysical version of functionalism, since this position avoids any commitment to a distinction between two kinds of property tokens having to interact within the production of behaviour. There are however two main difficulties proper to realizer functionalism. The first ensues directly from the motivations of proponents of causal role functionalism in having a stricto sensu ontological “commonality” shared by the different configurations of physical property tokens that make true a certain mental description. Indeed, if there are such differences, then these entities enter in slightly different causal and nomological relations and thus, although satisfying the same functional specification, they are not causally identical. As a result, realizer functionalism drops the fundamental insight of functionalism with respect to multiple realization, according to which different physical entities can exemplify identical mental properties playing the same role in the production of the behaviour as well as the further mental property tokens. In other words, within the framework of realizer functionalism, physically different individuals exemplify at best imperfectly similar mental properties.

128

From psychology to neuroscience

Note that this commitment of realizer functionalism is consistent with the result of our discussion of the implication of the token-identity thesis (3.7.1 Remarks on the concept of ‘identity’). On the one hand, we shall argue that the requirement for causal identity as taken up and pursued by causal role functionalism leads to an uncomfortable dilemma with respect to multiple realization (5.5 The general dilemma of multiple realization). On the other hand, we will argue that causal similarities between realizers are sufficient to ground a reductionist position, which is conservative with respect to mental property tokens while providing an explanation of the ability of psychology to abstract from the causal details distinguishing the different realizers (6.3.6 Conservative reduction of psychology). The metaphysical background against which realizer functionalism is usually conceived raises a second and more serious difficulty, stemming from the fact that realizer functionalism is usually built within an orthodox Humean metaphysical framework, which denies the existence of fundamental necessary connections between fundamental property tokens and, thereby, rules that genuine functional property tokens do not exist. Accordingly, everything there is in the world is a vast mosaic of fundamental, intrinsic and categorical physical property tokens, whose distribution makes true functional descriptions and, derivatively, mental descriptions, but which does not admit genuine functional property tokens. Realizer functionalism thus sets itself up to slip down the slope to eliminativism. Let us review these elements one by one. The general idea of the Human metaphysical framework is grasped in the Lewisian thesis of Humean supervenience: “Humean Supervenience is named in honour of the greater denier of necessary connections. It is the doctrine that all there is to the world is a vast mosaic of local matters of particular fact, just one little thing and then another. [...] We have geometry: a system of external relations of spatiotemporal distance between points. [...] And at those points we have local qualities: perfectly natural intrinsic properties, which need nothing bigger than a point at which to be instantiated. For short: we have an arrangement of

Psychology and neuroscience

129

qualities. And that is all. There is no difference without difference in the arrangement of qualities. All else supervenes on that (Lewis 1986: ix-x)”. Following the sceptical inclination of classical empiricism with respect to causation (see notably Hume 1978: 162), David Lewis’s picture of the world offers only a vast collection of fundamental physical property tokens, whose dispositions are accepted as primitive (Beebee 2006: 511). Those fundamental physical properties exemplified at space-time points are purely intrinsic and qualitative. Their respective essence does not depend in any respect on what is exemplified at other space-time points elsewhere in the universe. To take the overall distribution of fundamental property tokens as primitive amounts to taking this overall distribution as contingent, but also to regarding any relation obtaining between any two precise tokens as contingent. There is consequently no necessary connection between the fundamental tokens constituting the vast Humean mosaic, so that the exemplification of a specific quality at one space-time point does not impose any restrictions on what can be exemplified elsewhere in the universe. The second part of the Lewis’ Humean supervenience thesis claims that this fundamental distribution of intrinsic and qualitative fundamental physical property tokens constitutes the ultimate truthmaker for any true empirical description of the world, including statements describing causal relations and laws of nature. However, such a metaphysical framework cannot admit functional properties. As said above, physical properties are essentially intrinsic and categorical. As such, their identity is independent of what obtains anywhere else in the universe and, in particular, it is independent of the causal relation in which fundamental property tokens are involved. Roughly, whether or not two property tokens of types F and type G stand in a causal relation depends on the pattern of regularity between exemplifications of F’s and G’s, but neither exemplification has per se an ontological pertinence for the other. The pattern of regularities in question supervenes on the overall distribution of categorical properties in the world. The identity of a particular exemplification of a property F is therefore ontologically prior to the causal character of the relations in which it stands, and it would remain

130

From psychology to neuroscience

an exemplification of an F whatever else there is in the universe. The same holds in a counterfactual account of causation, since in such an account, the truth of counterfactual dependencies is a matter of the laws of nature obtaining in the world, and those laws also supervene on the overall distribution of fundamental physical property tokens (for instance Lewis 2004). In contrast, the identity of functional properties tokens stems essentially from the causal relation in which they are involved. As such they cannot simply be considered as categorical, since their identity is essentially bound to their causal manifestation. If property tokens of type F are such because it is their essence to produce property tokens of type G, then the identity of those tokens is critically dependent on their causal manifestation. To make the contrast with our previous example explicit, if fundamental property tokens of type F are defined as tokens that produce tokens of type G, then the causal connexion obtaining between the Fs and the Gs is necessary and does not depend in any respect on what obtains elsewhere in the universe. As a result, Humeanism cannot accept functional property tokens because it denies any necessary link between the essence of property tokens and their causal dispositions whereas functional property tokens are essentially characterized by their causal dispositions. Genuine functional property tokens are therefore excluded and consequently realizer functionalism leads to an elimination of mental property tokens. The problem of the ontological status of functional property tokens leaves us, as functionalists, with an uncomfortable dilemma. On the one hand, causal role functionalism, taking mental property tokens to be functional second order property tokens, faces the supervenience argument, which shows that these second order property tokens, qua being ontologically distinct from the physical property tokens that realize them, are epiphenomenal. On the other hand, although it can accept functional descriptions, realizer functionalism cannot metaphysically admit genuine functional property tokens, as it takes fundamental property tokens to be exclusively categorical. Realizer functionalism is thus arguably put upon the slope to the elimination of mental property tokens. Hence, the familiar duo of bad options appears:

Psychology and neuroscience

131

either mental property tokens are epiphenomenal, or they don’t really exist at all. 4.2.4 Functionalism all the way down to physics Let us recast the terms of this debate. One the one hand, we can’t adhere to both classical causal role functionalism and Kim’s supervenience argument. Against the background of Chapter 3, this suggests taking the referent of our mental descriptions to be identical with the configurations of physical property tokens that realize them, in order to vindicate their causal efficacy. However, to interpret realizer functionalism in this sense is a mistake, because it would amount to identify mental property tokens, whose essence is dispositional, with property tokens whose essence is categorical only. In other words, the dilemma of functionalism arises because of the presupposition of a Humean metaphysical framework that takes for contingent the causal relations in which fundamental physical property tokens are inserted. Obviously, rejecting the claim that the essence of fundamental physical properties is categorical only allows realizer functionalism to admit genuine functional properties. 4.2.4.1 Troubles with Humeanism Independently of the dilemma of functionalism, there are recent and serious objections addressed to the conception according to which properties are categorical only and thereby disconnected from the causal relations in which they stand. As has been noted since Hume’s time, the Humean views on causation are at odd with our experience of human agency, according to which our mental properties brings about our behaviour in such a way that this behaviour would not exist without the occurrence of these mental properties (Esfeld 2007: 211). However, within a Humean framework the fact that, for instance, Rabin’s death is temporally subsequent to the intention of an assassin to kill Rabin does not depend in a strong sense on this intention, since this spatiotemporal contiguity is primitively contingent. Of course, what confers its causal character on this contiguity is the fact that we can find other, similar intentions that are also contingently followed by similar events. The

132

From psychology to neuroscience

problem is not so much that causation is an extrinsic relation,33 but the modal character of this relation. Through the repudiation of necessary connections in favour of an account of taking causal relations to supervene on the overall distribution of intrinsic and categorical property tokens, Humeanism is unable to acknowledge singular causation. Yet our experience of agency is precisely characterized by the fact that we experience our actions as being brought about by our intention, even if there are no comparable pairs of events anywhere in the universe. We experience our mental properties as what grounds the existence of our behaviour, which Humeanism does not admit. This inability of Humeanism to account for our experience of mental causation is important to what we argued in Chapter 2, in which we tried to show that the premise of mental causation is itself firmly grounded (2.2.1 The Premise of mental causation). The overall argument for token-identity would consequently be strongly weakened if we have to give up the conception according to which mental property tokens bring about their effects. Nonetheless, one can still argue that our metaphysical views on the world should not be dictated by pre-theoretical considerations. However, Humeanism has been recently the target of serious theoretical objections, which also focus on the implications of the disconnection between the categorical essence of fundamental properties and the causal relation in which they enter. Accordingly, Humeanists are committed to the unknowability of the categorical essence of the fundamental intrinsic properties, which makes the categorical essence as mysterious as causation is from the Humean perspective. This consequence of Humeanism – that we cannot know about the fundamental properties of nature – has been

33

Hawthorne (2004) argued that the extrinsic character of causation is incompatible with a conception of the mental that imposes causal requirement on it. However, this argument was countered by Weatherson (2007), holding that “it is possible that extrinsic properties are possessed intrinsically”; he uses the case of uncles, who, though extrinsic to local spatio-temporal regions, might still be contained within those regions over the duplication of those regions in other possible worlds. A similar argument applies to causation.

Psychology and neuroscience

133

labelled Humility34 by philosophers commenting on the contemporary revival of the Humean metaphysical framework, in reference to the Kantian humility.35 The argument of Humility against Humeanism runs as follows. First, Humeanism takes fundamental physical properties to be intrinsic and categorical properties. Since the entire distribution of these properties in our world is taken as primitive, relations between fundamental property tokens are contingent and their identity is independent of these relations. Secondly, assume that Tf is a complete and final physical theory delivering a true and complete inventory of those fundamental properties and of their respective role in nature. Analogously to the Ramsification of folk psychology, Tf can be expressed by a Ramsey-sentence describing every possible relation obtaining contingently in our world between fundamental property tokens. This sentence is built by replacing the theoretical terms t1, … tn, of Tf by variables x1, …, xn and contains additional terms of our common language o1, … om.The problem then arises about how we should consider the relation between the categorical essence of a property ti and its role in the Tf. Since this link is contingent, it seems perfectly reasonable to conceive a world w’ such as the one proposed by Bird (2005: 449) in which charge has the causal role associated in the actual word w with gravitational mass, while gravitational mass has the causal role of charge in w. Technically, this derives from the possibility of permuting the theoretical terms ti and tj with respect to the variables standing for them in the Ramsey-sentence. In such a case, we have two possible worlds which are physically different although both satisfy Tf, and since ex-hypothesis Tf is a complete theory of what obtains in both worlds, we have no way of distinguishing between both worlds.

34

See notably Bird (2005) and Black (2000) for different formulation of the argument. Lewis (2009) acknowledged and endorsed the commitment to Humility. We shall here follow his reconstruction of the argument. 35

See notably Langton (1998).

134

From psychology to neuroscience

Why should a Humean recognize that both w and w’ are genuine possibilities? Because the assumption that the categorical essence of properties is only contingently related to the causal relation in which they stand commit him to what is known as quidditism, namely the endorsement of a primitive “whatness” of properties. Yet, given the fact that we acquire knowledge about the world through our causal interaction, or through the causal interaction of our measurement devices with the world, we do not have any means to know if we live in w or w’. As Lewis recognize it, “it is one thing to know that a role is occupied, another thing to know what occupies it” (2009: 204). This is the consequence of Humility of quidditism. As Jackson put it, “the possibility that (i) there are two quite different intrinsic properties, P and P*, which are exactly alike in the causal relations they enter into, (ii) sometimes one is possessed and sometimes the other not, and (iii) we mistakenly think that there is just one property because the difference does not make a difference [lead] to the uncomfortable idea that we may know next to nothing about the intrinsic nature of our world. We know only its causal cum relational nature” (Jackson 1998: 23-24). Obviously, we are in the face of an equivalence between the unknowability of the categorical essence of fundamental properties and causation. Historically, it was the unknowability of the latter that so counted against it from the point of view of classical empiricism. Thus, if we listen to the councils of empiricism itself, we should suspect any account which takes as primitive the distribution of properties whose essence is in principle not accessible to our knowledge. 4.2.4.2 The causal theory of properties The line of reasoning leading to the objection of Humility against Humeanism is essentially based on epistemological considerations going back to the contemporary discussion involving, among others, Shoemaker, who argues that the supposition “that the identity of properties [consists] of something logically independent of their causal potentialities [implies] not merely (what might seem harmless) that various things might be the case without its being in any way possible for us to know that they are, but also that it is impossible for us to know

Psychology and neuroscience

135

various things which we take ourselves to know” (1984: 214-15). This strongly suggests the endorsement of a more substantial relation than pure contingency between the essence of properties and the causal relations in which they stand. Such a stronger conception of this connection often takes the form of a metaphysical acknowledgement of causal powers, which we label “Dispositionalism”. According to a first version of Dispositionalism, “what makes a property the property it is, what determines its identity, is its potential for contributing to the causal powers of the things that have it, [which] means, among other things, that if under all possible circumstances properties X and Y make the same contribution to the causal powers of things that have them, X and Y are the same property” (Shoemaker 1984: 212).36 In other words, what is essential to a property is no longer an intrinsic categorical essence, but rather its causal powers. This amounts to taking the relation between the essence of properties and the causal relations in which they take place as necessary rather than, on Humean lines, contingent. Just as Humeanism can make true functional descriptions without admitting genuine functional properties, Shoemaker’s account can accept categorical descriptions in spite of denying that properties themselves are essentially categorical. For instance, “being triangular” and “being trilateral” should not be regarded as predicates designating different categorical properties having the same causal powers, but rather, as different predicates with different meanings designating the same property. Indeed, it is difficult to conceive how something making the application of the former predicate could fail to make true the application of the latter in virtue of its causal powers. If we give up the idea that there is a distinct property for each predicate we have in language, there is no clear obstacle to consider those predicates as referring to the same property,37 i.e. the same causal power. To put it in 36 37

For similar accounts, see notably Bird (2005), Ellis (2001), Mellor (1991).

This point is extensively discussed in Heil (2003), chap. 1, 5 and 7, where the author criticizes strongly what he calls the naïve “Picture theory” according to which the principle of realism mandates that any true predication “Fx” is true in

136

From psychology to neuroscience

a nutshell, the distinction categorical/causal is not ontological but propositional, that is, it picks among two different kinds of predicates; causal powers can therefore be said to make true the application of categorical descriptions. However, to endorse a metaphysics in which all properties are pure powers, along the line proposed by Shoemaker, brings up some new problems. If we suppose that Fs are nothing but powers to produce G’s, we must also presume that G’s are themselves causal powers to cause H’s, which are in return powers to cause Js, etc. It seems that we obviously enter an infinite regress (Martin 1997: 213). Heil (2003: 98) expresses the same point quite ironically. Consider a line of domino, the first toppling the second, the second the third and so on. Now, if all there is to the first domino is the causal power to topple a second one, the same to the second, and so on, nothing happens, simply because there is nothing to be toppled by the first domino. The domino example forces us to reflect that a metaphysical framework including only purely causal, non-qualitative properties is just as fanciful as a Humean metaphysics. This motivates the metaphysical claim according to which properties are essentially categorical and dispositional at the same time, the dispositionality and categoricality of a given property being strictly identical: for any intrinsic property P of any particular object, P’s causal powers Pd and P’s quality Pq are one and the same thing, such that Pd = Pq = P. Note that we are not concerned with aspects of properties as sometimes claimed, since supposing that we are concerned with aspects would amount to acknowledging that dispositionality of P and qualitativeness of P are second-order properties; in this case, our problems would once more confront us with the general problem of the causal impotence of second-order properties. Before turning to the solution offered by this conception of properties for the dilemma of functionalism, let us make a few remarks about the relation between objects and properties, as we are going to

virtue of the existence of a corresponding and specific property F exemplified by x.

Psychology and neuroscience

137

defend the thesis that properties are both qualitative and dispositional at once. In Chapter 2 (2.1 Objects, events and properties: preliminary remarks), we argued also that an appropriate formulation of the problem of mental causation should be given in terms of property tokens that are defined as the exemplification of a property by an object at a certain time. Objects are concrete particulars, localized in space and time. Properties characterize objects, which are, as well, a matter of certain qualities and certain powers that produce other properties. We shall here consider that properties characterize object qua being modes of the objects, that is, ways of being of particular and concrete objects rather that universals instantiated by objects.38 Property tokens are ontologically basic entities, namely particularised modes of being of objects that are localized in space and time, such as cars, chairs, individuals, but also fundamental space-time points. The particularised modes of being of objects are at once certain qualities and certain causal powers, having the causal disposition to produce further particular objects. The ways objects make true the applications of certain concepts can be either categorical or dispositional; it is this that allows us to classify entities in the world taxonomically. Property types, in our schema, are not universals instantiated by objects but rather concepts, which are descriptive categories labelled by predicates. Property kinds are classes of descriptive categories such as, for instance, the class of psychological types. Let us return to the dilemma of functionalism, which put us in the traces of this metaphysical digression. We have pointed to the problems of constructing a functionalism under a Humean metaphysical framework that takes properties to be intrinsic and categorical standing only contingently in causal relations, as this vitiates the insight of functionalism – grasping the systematic causal aspect of properties. Instead, Humean metaphysics makes it logically impossible to identify straightforwardly functional property tokens

38

Cf. Heil (2003: , chapt. 13), who discusses in details the advantage of such an account over taking properties to be universals.

138

From psychology to neuroscience

with the configurations of physical properties that realize them, since the former are essentially causally characterized whereas the latter are essentially categorical. As we have also explained, if we move to a metaphysic of powers, we can collapse the problematic opposition between functional and categorical properties and thus identify functional property tokens with certain configuration of fundamental property tokens satisfying the right causal specifications. Among the configuration of fundamental of physical property tokens existing in the world, some of them are genuinely powers to bring about certain other configurations of functional property tokens that satisfy the causal specifications of certain functional property tokens. Nothing then precludes the identification of functional property tokens with configurations of physical property tokens,39 from which it ensues that realizer functionalism can admit genuine functional property tokens. A metaphysic of powers does justice to our pre-theoretical notion that our agency is defined by the fact that our intentions bring about our behaviour, since configurations of fundamental property tokens, to which functional mental property tokens are identical, necessarily bring about their effect. Hence, the relation between causes and effects is no longer a contingent one, depending on the occurrence of similar and adjacent causes and effects elsewhere in the world that collaboratively form a pattern of regularity supervening on the primitive distribution of fundamental property tokens. Against appropriate local background conditions, the occurrence of a mental property token is sufficient by itself to bring about a certain behaviour. Finally, let us simply repeat that a metaphysic of powers is immune to the humility objection, as there is, in principle, no primitive “whatness” of properties beyond the limit of our epistemic access. According to this framework, properties are qualities and dispositions at once, and their identity goes beyond their causal manifestation, which constitutes our lone source of knowledge with respect to what properties are. Hence, we avoid affirming differences where there is no 39

Let us be perfectly clear here. The claim is that mental property tokens, which are functionally individuated, are identical to physical property tokens, which are essentially dispositional rather than essentially categorical.

Psychology and neuroscience

139

causal difference that could ever be detected in any situation. From a more general point of view, this argument can be seen as just one of coherence between our metaphysical conception of the world and the requirement limiting our epistemic access to the fundamental physical properties to their causal function. This is obviously nothing but a way of extending the functional conception of properties all the way down to physics. 4.2.5 Summary Before moving ahead, let us summarize the conception of folk psychology we have proposed. Psychology is a theory using a class of systematically interrelated concepts in order to give an epistemological account of the mental phenomena we observe in the world. Psychological descriptive categories are essentially functionally defined, that is, they grasp the typical causal relations of psychological property tokens and property tokens of different kinds. The nexus of these causal obtaining in the world is what psychology, as a higher level theory, sets out to explain. Among different ways of functionally individuating mental property types, common-sense functionalism appears as the most accurate approach to mental properties, since it is the only account that defines functional types in close relation with our ordinary mental categories such as desires and beliefs. Within that framework, the nonmental causes of mental property tokens are characterized in terms of environmental circumstances. The behavioural effects of mental property tokens are defined in terms of conditional environmental impact. Mental property tokens are characterized by the location they occupy in a global network of causal interaction between environmental causes, behavioural effects and mental property tokens of different types. This overall network is described by translating folk psychology into Ramsey-sentences. According to the argument for token-identity, mental property tokens are ontologically identical to complex configurations of physical property tokens, which makes it logically erroneous to think of them as second-order property tokens supervening on physical property tokens. However, taking functional property tokens, which are essentially

140

From psychology to neuroscience

characterized by their causal role, to be identical to physical property tokens, supposes a causal approach of properties, which applies to physical property tokens as well, as shown by our discussion of realizer functionalism. This leads us to conceive of the problem of mental causation in a general functionalist framework, which defines causally property types of any kind all the way down to physics, rejecting thereby a Humean metaphysical framework. It is only at this price that the problem of mental causation can be solved by a genuine identity solution that does justice to our pre-theoretical conception of human agency.

4.3 Neuroscience 40 In Chapters 2 and 3, we argued that the mental property tokens and physical property tokens are ontologically identical, which is the basis of the claim of supervenience. The problem of mental causation is usually formulated in terms of a gap between the mind and the physical world. The latter is characterized the self-sufficient character of physical causation, since physics, qua being causally complete and universal, is in principle able to account for the occurrence of any physical property. In fact, according to our argument of Chapter 3, physics can account for the occurrence of any kind of causally efficient property tokens, since any causally efficient property tokens is a physical property token, or is identical to a configuration of physical property tokens. However, looking at the work of the contemporary science of cognition, one can easily see that it is not described in terms of charges, neutrons or spins (although these may impinge at some level), but rather in such neurobiological terms as neurons, synapses, cells, etc. There are at least two good reasons that physics hasn’t simply taken over the field. The first is that it is not the case that any type physical 40

Our views on neuroscience have been profoundly shaped by the account proposed by Carl Craver in his excellent book, Explaining the Brain (2007). The second part of this chapter is strongly inspired by his mechanistic account, although, as will be pointed out, we disagree with respect of the issues of reduction and of the ontological status of higher level causes.

Psychology and neuroscience

141

systems produces the kind of behaviour we are interested in. Of course, stars, electrons and entangled particles can be causally related to human behaviour in some sense. When I look through a telescope at Saturn’s rings, I am of course carrying out one of the behaviours that make up science. But this is obviously extrinsic to our interest here. We are rather interested in physical systems whose causal contribution plays a key role in the explanation of the extraordinary adaptativeness of human behaviour. This has led the empirical research to investigate the causal mechanisms that tie together sensory inputs and motors outputs of the central nervous system of living beings, especially humans.41 The second reason for why scientists do usually not explain the causal antecedent of human behaviour using the first level theory terminology of physics is linked to the complexity of the human brain. In spite of being in principle explainable in terms of aggregations of fundamental particles, causal relations leading to public human behaviour are quickly enmeshed in a complexity that defies linear analysis. Just to illustrate the point, an adult human brain, weighing on average 1350g, is made, among other type of cells, of roughly 85 billions of neurons (Williams and Herrup 1988) interrelated by something like 100 trillions of synapses (Bickle 1998: 135). Given that a single neuron is an incredibly complex object from the point of view of fundamental physics, that amplified complexity provides more than enough reasons to avoid the idea of starting the study of such systems from there. The vast interdisciplinary research program labelled “neuroscience” can be seen as framed on the one hand by the objective of developing a comprehensive account of the causal origins of both the human and non-human behaviour within the nervous system and on the other by the need of integrating methodological tools from different 41

This should not be read as a rejection of the multiple realization thesis, since 1) there is evidence for multiple realization within living beings and 2) it is still conceivable that non-carbon based individuals can nonetheless exemplify mental properties. Rather, we mean here simply that, contingently, the actual empirical research focus on actual entities exemplifying mental properties.

142

From psychology to neuroscience

scientific disciplines to resolve challenges presented by the complexity of the cognitive mechanisms. Neuroscience typically integrates results originating from cognitive psychology, neuropsychology, neurophysiology, neuroanatomy, electrophysiology, molecular neurobiology, but also from medical disciplines such as neurology or psychiatry and even theoretical disciplines such as theoretical computational sciences. Neuroscientific explanations thus, understandably, span different levels of description, integrate different results of different fields, each of them having their own methodological constraints that serve possibly different goals. In the next section, I will take us through the plurality of levels of description and research fields that make up part of the heterogeneous character of the neurosciences, using the investigation of spatial memory as orienting case.42 4.3.1 The scope of neuroscience 4.3.1.1 Multiplicity of the levels of investigation Neuroscience typically provides an explanation of the critical contribution of the nervous system to the production of behaviour by having recourse to different levels of description: there are at least four in the case of spatial memory.43 Consider a rat exploring a spatial environment. The phenomenon to be explained is the capacity of the rat to produce a behaviour, which attests through its systematicity that it has a memory of the space it has explored. Typically, the phenomenon of spatial memory is studied at the behavioural level by experimental 42

This should not be understood as the claim that there is a unique phenomena of spatial memory, nor that the provided example of explanation is in a final and correct state. Our present purpose is to illustrate the general kind of explanations that neuroscience provides.

43

Their is no general agreement with respect to the number of levels that have to be considered within neuroscience, given that it could notably vary depending on the object of inquiry. We do not need to endorse a precise quantity here. However, we shall focus in more detail on two levels of descriptions in the chapter discussing examples of application of the strategy of epistemological reductionism that is pursued here, namely, neuropsychology and molecular neurobiology.

Psychology and neuroscience

143

cognitive psychology and ethology, each setting up different experiments, such as mazes, surrounded by visual cues, in which different specimens can be inserted. The data studied in such behavioural experiments are for instance the time that elapses in finding a path to arrive at some goal marked by an award (food, for instance). Or the study could be of the impact of the visual cues on the rat’s behaviour. This is typically the first level, where the target of inquiry is entirely defined in terms behavioural characteristics. The exhibition of behavioural patterns characteristic of a specimen that has memorized the spatial structure of its environment depends in some or other way on several parts of the nervous system. However, bilateral lesions of the hippocampus and of some other areas of the frontal cortex impairs space orienting behaviour, suggesting damage to the local brain areas concerned with spatial memorization. Research has shown over time that among the functions of the hippocampus is that of determining the spatial orientation of the individual’s movements across his environment. Accordingly, the hippocampus plays the role of a “cognitive map” of the environment, enabling the animal to navigate through it in a non-hazardous manner. At this second level of description, the hippocampus appears to be central to the brain’s ability to memorize spatial features of the environment and contributes, through its connectivity with other brain areas, to produce the general behaviour of the individual. At a third level of description, evidence stemming from the record of the firing activity of pyramidal cells located in that part of the hippocampus called Ammon’s horn confirmed the key role of the hippocampus in the phenomenon of spatial memory. The so called “place cells” fire preferentially when the rat enter a certain location of a testing platform in a certain orientation (O'Keefe and Dsotrovsky 1971). Over the last thirty years, empirical data from a number of experiments has demonstrated and detailed the implication of hippocampal neurons in the spatial representation that mice have from their environment, notably with respect to other variables such as visual cues or the speed of the rat’s movements. If place cells are likely to maintain their firing locations even if the visual cues surrounding the maze have been removed, the

144

From psychology to neuroscience

relocation of the visual cues appears to cause an adaptation of the firing activity of the place cells. This adaptation is explained at a fourth level of description at which the electro-physiological properties of neurons are considered. The general hypothesis with respect to the stable adaptation of the way in which neurons respond to the firing of afferent neurons is long-term potentiation (hereafter LTP) and its reciprocal process, namely long-term depression (hereafter LTD). LTP is a lasting enhancement of the post-synaptic response to the same pre-synaptic action potential consequent to a repeated and high-frequency stimulation, while LTD is a lasting decrease of the post-synaptic response as the result of a persistent stimulation at a low frequency.44 Finally, both LTP and LTD rely on molecular constitution of the synapses and of the extra-cellular environment. The molecular details of these molecular processes are quite complicated and will be briefly presented later (7.3.2 From molecules to behaviour: some empirical evidence). Letting aside LTD, the LTP process occurs in case of repeated stimulation of the pre-synaptic cells with an interval of 200msecs. This stimulation bypasses the inhibitory potential in the post-synaptic cell that would usually buffer it to be depolarised by a subsequent stimulation of the pre-synaptic cell. The depolarization of the post-synaptic cell membrane causes Mg2+ ions to unblock NMDA receptors, which allows an influx of Ca2+ ions into the cell. This calcium influx initiates a series of biochemical reactions leading to structural changes in the dendritic spines or new synapse formation. Theses structural modifications enhance the propensity of the postsynaptic cell to fire in response to stimulation from the pre-synaptic cell. Research, then, that addresses spatial memory must span multiple levels of analysis, from the behavioural level, at which the studied 44

LTD is not understood as well as LTP. Moreover, there are multiple forms of LTD in various areas of the brain. Nonetheless, LTD occurs in the hippocampus by affecting the synapse linking the Schaffer collaterals and the CA1 pyramidal cells, as a result of a repetitive stimulation of the former over significant periods of time at a low frequency (Purves, Augustine, Fitzpatrick, Katz, Mcmantia and Mcnamara 2002: 450)

Psychology and neuroscience

145

phenomenon is defined, down to molecular processes, in order to produce a valid neurobiological account. This is not, of course, an anomaly of research into spatial memory, but rather a general feature of the neuroscientific explanations. For instance, the investigation of the visual system goes from the general phenomenon of vision to the connectivity of neuronal layers in the primary visual cortex while studying the contribution of different brain areas to the overall phenomena of vision, such as the dorsal stream that enables the online visual guidance of action and the ventral stream, which is involved in the recognition of objects (cf. Goodale 2000). Notorious mental diseases such as Huntington’s chorea or Alzheimer’s dementia are also, even if only incompletely, explained by similar hierarchical investigations spanning multiple levels of descriptions. Apparently, the first rely on an progressive atrophy of the basal ganglia involving a important loss of ACh and GABA neurons (Kolb and Whishaw 2003: 734), whereas the neural correlates of the second are multiple, including among others important alterations of the dendritic ramifications of the cortical pyramidal cells as of the hippocampal neurons (Kolb and Whishaw 2003: 744). There are thousands of such examples. The point is that neuroscientific explanations can and most often do span multiple levels of description. The reason is basically that neuroscientific research investigates the components of the nervous systems in relation to the global behavioural outputs of the organism, which quickly leads to the enormous complexity of the entire system. To deal with this complexity, neuroscience proceeds by functionally decomposing the nervous system into parts and sub-parts at different levels with different specific functions, in order to explain how, as in the case of our example, Ca2+ ions movements through NMDA receptors of synaptic path in the hippocampus may shape certain aspects of publicly observable behaviour. 4.3.1.2 Multiplicity of the fields of investigation Being an interdisciplinary research undertaking, neuroscience brings together a large variety of scientific field of investigation, that are each characterized by their own explanatory goals, concepts, techniques of

146

From psychology to neuroscience

investigations and methodological constraints, both at different levels and at the same level of description. We have seen the material correlate to the plurality of levels in the last section; here, let us borrow a short example to Craver (2007: 237-40) to illustrate how neuroscience integrates results of different disciplines at the very same level of description. The discovery of the phenomenon of LTP is the product of the integration of results originating from both electrophysiological and anatomical researches on the hippocampus, without any regard for the problem of memory. On the one hand, electrophysiologists during the 1950s used the tetanisation of hippocampal cells as a laboratory tool in order to increase the period of time during which the electrophysiological properties of neurons could be studied in vivo. They knew that delivering a short high-frequency stimulus to the hippocampal neurons restores the declining properties of neural circuits along the experiments. But, although it seems retrospectively surprising, these initial researches did not include memory or learning in their area of interest. The hippocampus was not associated with memory, then, but rather with a vast list of other functions. Electrophysiologists were especially interested in the properties of hippocampal cells for two reasons. First, the hippocampus is a relatively simple structure, which is relatively similar across species, in addition to being easily accessible in vivo through the skull. It thus presented a desirable target for investigation by electrophysiology. Second, it seemed to be highly implicated in cases of severe epilepsy, which was already interpreted as an electromagnetical storm spreading in all the brain. From the viewpoint of anatomical research, it was desirable that the structure and the wiring of hippocampal neurons be clarified by means of Golgi stains and terminal degeneration. Obviously, the map of neural connectivity of hippocampus became a precious tool for electrophysiologists, allowing them to multiply the possible locations of electrodes to inject current and measure the cell’s responses. In short, anatomists and electrophysiologists, studying different properties of the same neural population, combined their respective techniques of

Psychology and neuroscience

147

investigations in order to build an account of the relations obtaining between neural wiring and electrical-physiological properties. As this example shows, neuroscience advances by integrating research results originating from different levels of description and from different fields of investigation at the same level. Neuroscience is therefore a highly heterogeneous body of knowledge about the nervous system and its involvement in the production of the behaviour. From the perceptive of reducing psychological descriptions to neuroscientific descriptions, this obviously raises the question of knowing what confers a certain unity to the targeted reduction basis. We shall explore this question in the next section. 4.3.2 The unity of neuroscience 4.3.2.1 Mechanisms and mechanistic explanations As our summarized example of the explanation of spatial memory shows, neuroscience typically builds explanations of high level phenomena by means of a step by step analysis of the elements of the nervous system that are involved in the production of the target of inquiry, aiming to uncover the mechanisms by means of which a cognitive phenomena occurs. We take mechanism here to be defined as a set of entities and their causal activities organized such that they exhibit the phenomenon to be explained (adapted from Craver 2007: 5). The explanandum phenomenon P consists in the production of certain effects and is therefore defined in terms of a certain causal role. For instance, the strengthening of the synaptic connection between two neurons is the kind of phenomenon that can, in principle, be mechanistically explained. The phenomenon is defined by the increasing of the propensity of the post-synaptic cell to fire in response to a pre-synaptic stimulation after a tetanisation. In that sense, the phenomenon is defined in terms of the typical causal role linking the tetanisation of the pre-synaptic cell to the strengthening of the synaptic connection. The individuation of the causal role ψ of P has to include a precise characterization of the conditions in which the phenomenon occurs, or is inhibited, or modulated, as well as its collateral effects that

148

From psychology to neuroscience

are not directly relevant to the phenomenon, but that contribute to its identity (Craver 2007: 128). A mechanism M for P is a collection of organized components X1, … Xm performing activities φ1, …, φn such that M exhibits the causal role ψ of P. Components of M are what we usually refer to as “parts” of M. But it is not the case that every part of M is a component of M. A part of M is a constitutive component of the mechanism M for P if and only if “one can wiggle the behaviour of the whole by wiggling the behaviour of the component and one can wiggle the behaviour of the component by wiggling the behaviour as a whole. The two are related as part to whole and they are mutually manipulable” (Craver 2007: 153). Accordingly, Xi of M is a constitutive component of M 1) if Xi is a part of M, 2) if Xi and its activity φi are causally relevant to the ability of M to execute the causal role ψ characteristic of P, and 3) if some changes in ψ of P impact the activity φi of Xi. The criterion of mutual manipulability is implicitly present in the research strategies that neuroscientists use to validate hypotheses about mechanisms. These strategies are in practice inter-level experiments that aim to show that the respective activities of the components contribute to the overall activity of the mechanism for a phenomenon, and that the phenomenon itself is grounded in the activities of the mechanism’s components. A first experimental strategy relies on the inhibition of the activity or the entire removal of a putative component X of M. If X is a constitutive component of M, then the inhibition of X’s activity or its removal should affect the ability of M to fulfil the causal role ψ. Typically, lesion experiments proceed by removing some brain area in order to evaluate its contribution to the behaviour of the individual in the execution of a certain task. For instance, the bilateral ablation of a rat’s hippocampus impairs its ability to memorize the spatial features of its environment. At a more fine-grained level of description, the knocking-out of certain genes producing CREB protein in rats impairs the late LTP process by means of which the potentiation of synapses get stabilized for days or months, without inducing any change in the early phase of LTP. As a result, the synthesis of CREB proteins

Psychology and neuroscience

149

appears to be indispensable to the induction of durable structural changes in hippocampal synaptic connexions, which tells us that CREB is a constitutive component of the mechanism of memory consolidation (Abel, Nguyen, Barad, Deuel, Kandel and Bourtchouladze 1997; Bourtchouladze, Frenguelli, Blendy, Cioffi, Schutz and Silva 1994: 66). A second bottom-up strategy relies on the stimulation of the components of a mechanism. If X is a constitutive component of M, then the artificial stimulation of X should have some observable outcome on the causal role ψ at the level of the phenomenon. The paradigmatic case of stimulation experiment is the electrical stimulation of the motor cortex to elicit body movements by the subject. This methodology enabled Penfield and Boldrey (1958) to map the connectivity between muscle and motor neurons of the primary motor cortex (Brodmann area 4) of conscious human patients undergoing brain surgery. Using similar conditions and methods, Penfield and Jasper (1954) found evidence for the left-sided lateralization of the main brain areas involved in the production of speech. They observed that, contrary to stimulation of the right hemisphere, stimulation of the left can block the patient’s ability to speak. The criterion of mutual manipulability requires as well that the execution of high-level behaviours induces the activation of the components of the mechanism by mean of which the capacity for those behaviours is achieved. That is to say, if X is a component of the mechanism M that contributes to explain the phenomenon P characterized by the causal role ψ, then there must be a difference in X’s activity that co-varies with the disposition of the subject to fulfill ψ. The overall design of these top-down experiments is about engaging the subject in an experimental task T, requiring the execution of ψ, which allows us to scan for correlate brain specificities. Neuroimaging techniques, such as fMRI or PET, let us observe in real time the brain activity correlate with undertaking a given cognitive task; as, for instance, is seen in the increase of the primary somatosensory and the primary motor cortex activity when hand activity is initiated, for instance by having the subject tap a contralateral finger. The execution

150

From psychology to neuroscience

of a sequence of movements with a finger is correlated with the additional activation of the premotor cortex (Roland 1993). Finally, there is a suppression version of the top-down strategy of investigation. The general idea is that the suppression or the absence of a phenomenon P implies that the constitutive components X of the mechanism M for P will behave differently when the high level phenomenon is suppressed. It is important to note that the scientific assumption, here, is that a component is constitutive of a mechanism if and only if the influence between the manipulation of the phenomena P and those of the putative component X of the mechanism M for P is reciprocal. One way manipulability is not sufficient to secure the constitutiveness of X. None of the four strategies presented above can individually secure the constitutiveness of a component. For instance, in a lesion experiment, the intervention on region A can disrupt the functioning of other regions B and C of the brain by damaging neural projections or interrupting blood supply in direction of those other areas. In such a case, one cannot know which of the consequences of the lesion are due to the impairment of areas A, B or C. It is consequently not possible to know which of them represent constitutive components of the mechanism for the target phenomena. It is only when a hypothesis is confirmed in both directions, or in other words, earns its seal of the mutual manipulability, that we can legitimately say that the component is genuinely constitutive. To sum up Craver’s account of constitutive relevance, the criteria of mutual manipulability tests the claim that a given part of a mechanism is constitutive of the mechanism explaining some higher level phenomenon. This is another way of stating the reason that interlevel experiments have had so central a place in neuroscientific practice, since the different versions of this methodology of investigation ensure that the considered components are constitutive of the mechanisms that explain the explanandum phenomenon.

Psychology and neuroscience

151

4.3.2.2 Mechanistic explanations and functional analysis At this point, one can wonder what distinguishes the mechanistic account of neuroscientific explanations from Cummins’ account of explanation by functional analysis. Accordingly, a phenomenon P having the causal role ψ is explained by analysing ψ into a number of other sub-causal role γ1, γ2, …, γn of P or components of P such that the joint execution of γ1, γ2, …, γn results in the execution of ψ (Cummins 1975: 759), which is very close to Craver’s account of mechanistic explanations. Indeed, both positions agree on the view that explaining a phenomenon should be done via an account of how the causal role characterising the phenomenon is achieved by some sub-elements. To put it in a nutshell, the mechanistic framework is differentiated from Cummins’ account of functional analysis by the addition of empirical constraints on the possible models that we can derive from the functional analysis of the target phenomenon. Neuroscience aims to account for a given phenomenon by providing a model of mechanism that explains how a mechanism actually executes the causal role characteristic of a phenomenon by means of its actual activities engaging its actual components, rather simply how it is possible for this phenomenon to be achieved by mean of possible activities of possible components. Moreover, neuroscience aspires to explain completely how the causal role characterising a phenomenon is produced by components of the mechanism. By completely, we mean that neuroscience aims to build a complete actual model of the mechanism accounting for a given phenomenon P, in opposition to sketch models that do not specify all components and activities of a mechanism (Craver 2007: 112-14). These two normative requirements added to functional analysis lead neuroscience to provide empirically informed explanations of higher-level phenomena using the techniques presented above. Actual components of actual mechanisms are empirically amenable to the scientific analysis of their internal structure and their causal disposition to interact with other components of the same mechanisms, revealed by means of the use of a multiplicity of empirical techniques of investigations. In particular, these techniques have to pass the test of

152

From psychology to neuroscience

being able to manipulate the components in the context of inter-level experiments. The activities of components are causal roles. Their respective characterisation is settled in parallel with the isolation of the characterisation of the component responsible for that causal role. The point is that when scientists initiate top-down or bottom-up experiments, their interventions directed at the entire phenomenon or the component are designed to evaluate and highlight the causal contribution of the component to the phenomenon. In short, the criterion of mutual manipulability is used to justify the isolation of the constitutive character of a component through its causal contribution to the mechanism’s exhibition of the causal role characterizing the target phenomenon. Finally, mechanistic explanations add to functional analysis the criteria of the organization of the mechanism’s components. First, components and their respective activities give rise to the phenomenon, in as much as they are organized to activate on some definite path, in opposition to simple aggregates of components, which are insensitive to the rearrangement of their components. The components of a mechanism interact in a precise way, certain components interacting with certain further components but not with others. Secondly, mechanisms are spatially organized. The position, but also the size and the shape are important features of components in many explanations. For instance, Mg2+ ions can block NMDA receptors only to the extent that they are localized in the right place in the mechanism and that they have the appropriate size to fulfil this role. Third, interactions between components are ordered in time. Synapses, for instance, display plasticity only because components of LTP interact according to a precise temporal sequence, for otherwise, the fact that only certain frequencies of electrical stimulation potentiate the synapses would remain deeply obscure. 4.3.2.3 Unifying neuroscience As explained at the beginning of this section, neuroscience integrates results from both different levels and different fields, which obviously raises the question of whether or not neuroscience constitutes a unified

Psychology and neuroscience

153

body of theory. We shall again reference Craver’s account (2007: chapter 7), according to which the different fields of neuroscience are unified by their respective contributions to the common goal of constructing a general mechanistic model of the production of behaviour by the nervous system. This unification proceeds firstly from the horizontal integration of multiple fields to produce parameters that contain the space of all possible models underlying phenomenon that may be of neuroscientific interest. Secondly, this model mechanism is said to be vertically integrated in the overall neuroscientific project when a) its components are themselves explained mechanistically and b) this model is related to a higher level phenomenon, which can in turn be itself a component of a higher phenomenon. . The horizontal inter-field integration arises when different fields contribute to isolate the mechanisms of a given phenomenon by adding variously their own constraints to restrict the set of the ways the mechanism could possibly be modelled. These constraints concern the possible components, their spatial and temporal organization and their respective activities. The adjunction of constraints from different fields, which have different techniques of investigation and different descriptive resources, ensures that a component is a plausible candidate for exercising a certain activity in a mechanism. The fact that a component is observable from different perspectives provides several independent evidences supporting the hypothesis that it is an actual component of the mechanism. The inter-level integration of mechanisms proceeds from the insertion of a mechanism in a hierarchy of mechanisms, as a component with respect to a higher level of description and as a phenomenon with respect to a lower level mechanism. To illustrate the meaning of this distinction, take the case of a hippocampal neuron n1 having a causal role φ1. n1 is integrated to a unified neuroscientific account of spatial memory firstly if it is a constitutive component organized along with other constitutive components n2, …, nn taking respectively the causal roles φ2, …, φn of a mechanism so that we can explain the functioning of the hippocampus h and its causal role ψ, wherein h is similarly a component of the mechanism for spatial memory. Secondly, n1 is integrated to the unified global account of

154

From psychology to neuroscience

spatial memory if n1 and its causal role φ1 are explained by a mechanism having constitutive components c1, …, cm, which take respectively the causal roles δ1, …, δm and that are organized in such a way that they exhibit φ1. At this point we appeal to the mutual manipulability of components and phenomenon to ensure the consistence of the analysis of the hierarchy of mechanisms to account for the target high-level phenomenon. Mechanistic explanations unify neuroscience by explaining both how a certain entity e and its activity φ at a level LL0 operates as a constitutive component to a phenomenon pL+1 having the causal role ψL+1 situated at a higher level LL+1, and how the activities δ1, …, δm of constitutive components c1, …, cn of a mechanism mL-1 situated at a lower level LL-1 account for e and its activity φ as a phenomenon. The fact that neuroscience aims to account completely for φ in terms of δ1, …, δm and C1, …, Cn raises an epistemological and a metaphysical question. Epistemologically, we want to inquire into the reductive character of mechanistic explanations, while metaphysically, we want to know about the ontological status of the neurobiological property tokens to which mechanistic explanations refer. 4.3.3 Mechanistic explanations and the ontological status of mechanisms 4.3.3.1 Mechanistic explanation as reductive explanations With respect to our epistemological question, let us argue that the fact that neuroscience provides reductive explanations of higher-level phenomena in as much as it gives mechanistic explanations that aspire to explain without reminder higher-level phenomena by way of the organization of the component’s activities within mechanisms at a lower level. An explanation of a property F is reductive if it fulfils the three following criteria:45 firstly, it functionally defines the high level

45

These criteria are common to Chalmers (1996: 42-51), Kim (2005: 108-20) and Levine (1993: 131).

Psychology and neuroscience

155

property F, characterizing the explanandum in causal terms; secondly, it isolates a configuration of underlying subvenient properties such that they fulfil the causal role proper to F, characterizing the explanans in the terms of the more fundamental descriptive modalities proper to these lower level properties; and thirdly, it provides an account of how those lower level property tokens fulfil the causal role characteristic of F and thereby why they are a case of F. Mechanistic explanations spanning multiple levels of description fulfil these criteria. As pointed out previously, mechanistic explanation proceeds by characterizing the higher-level phenomenon to be explained in term of its characteristic causal role, including its precipitating conditions, possible modulations and by-products. It is perfectly reasonable to assume that higher-level phenomena are higherlevel property tokens individuated according to their causal role. The phenomenon of spatial memory having the causal role of orientating the behaviour of an individual in space is analysable as a property having that causal role. The first condition is therefore largely fulfilled. Secondly, mechanistic explanations describe mechanisms, which are defined as entities and activities organized such that they fulfil the causal role characterizing the high level property to be explained.46 Neuroscience aims thereby to uncover the actual mechanisms whereby the high level causal roles are implemented. Here also, it seems perfectly reasonable to assume that components of activities are property tokens, and mechanisms are thereby configurations of property tokens. The mechanisms are described using concepts that are specific to the level of description at which the mechanism’s 46

Interestingly, the literature devoted to reductive explanation refers often innocently to “physical mechanisms” that underlie the exercise of a high level capacity. For instance, Fodor says that the primary goal of scientific explanation is to “explicate the physical mechanisms whereby events conform to the laws of the special sciences” (1974: 107). Similarly, Levine argues that when we are looking for a reductive explanation of some high level property, we are seeking “the underlying mechanisms” (1993: 132). Chalmers argues that in order to explain reductively a high level property, “all we need to do is to show how certain lowerlevel physical mechanisms allow the analysis to be satisfied, and an explanation will result” (1996: 44).

156

From psychology to neuroscience

components are located, with the result that they satisfy the second criterion for reductive explanations. Finally, this description, supplemented by the relevant laws obtaining at this level, explains how the components interact in such a way that they exhibit the causal role which characterizes the phenomena to be reduced and why the mechanism is a mechanism for the phenomenon to be reductively explained. Satisfying this last criterion, mechanistic explanations appear to be reductive explanations, which is of course not a surprise since as we have pointed out, they aspire to completely account for the causal role characterizing the studied phenomenon in terms of an underlying structure of components whose respective activities are organized in such a way that they exhibit the explanandum. As the example of spatial memory shows, neuroscience provides mechanistic explanations that span multiple levels, from high level cognitive abilities down to chemical processes, which means that neuroscience aspires to explanations that have reduced its target phenomena to the activities of their components, with the latter, too, reductively explained in term of the activities of their constitutive components. To continue with our spatial memory example, we have a first reductively approximate explanation by way of reference to the hippocampal structures and their critical involvement in the phenomenon of spatial memory. We then examine this involvement to understand the hippocampal neurons, whose firing depends computationally on the position that the individual occupies in its environment. LTP is the process that accounts for this ability of hippocampus to create a spatial map of a novel environment by adapting to the weights of the synaptic connections, so the LTP process then becomes a target of inquiry, and we reductively explain it in terms of an interaction between the molecular components of both the presynaptic and the postsynaptic neurons and the extracellular environment. The process of neuroscientific inquiry is characterized by the aspiration to account reductively for the neural mechanisms causally involved in the production of the behaviour down to their chemical properties. At this level, components are molecules chemically interacting with other molecular components. To reduce further would

Psychology and neuroscience

157

obviously require a physical explanation of the chemical mechanisms. As a contingent fact, neuroscience usually stops at this point, since the study of the underlying fundamental physical components of these mechanisms does not bring any knowledge that would contribute specifically to the formulation of reductive explanation of mental phenomena. However, there is no formal objection to the formulation of mechanistic explanations down to the physical level. The lone limit is arguably a hypothetical fundamental physical level at which the causal dispositions of properties are not reductively explainable in terms of activities of further constitutive components. Accordingly, at this level we must accept the relation between the component and its activity as primitive, which is perfectly consistent with the causal theory of property proposed above (4.2.4.2 The causal theory of properties). Note that even if one endorses a Humean metaphysical framework, the situation is not so different, since the causal role of fundamental properties is not reductively explained in terms of the causal relations obtaining between more fundamental properties, but rather in non-causal terms, since in such a framework, causal relations supervene on non-causal facts. In both cases, neuroscience, qua formulating mechanistic explanations, provides reductive explanations of higher level cognitive phenomena down to chemical components of the nervous system of a living being, with the codicil that these components can also be reductively explained down to physical fundamental properties. 4.3.3.2 The ontological status of neurobiological properties As we shall see in more detail in the next chapter when the multiple realization argument will be discussed in an epistemological context (5.3 Multiple realization in an epistemological context), there is a controversy about whether or not the fact that a high-level phenomenon may be reductively explained opens the way to the epistemological reduction of this phenomenon. Let us here focus on the question of the ontological status of neurobiological property tokens. Spanning multiple levels of mechanisms, neuroscience deals with various types of entities such as cortical and sub-cortical structures, brains areas, nerves, cells, proteins, genes, ions, electrical potential, etc.

158

From psychology to neuroscience

The disputed point is the ontological relation obtaining between these entities standing at different levels of mechanisms. In the terms of the mechanistic framework, the question is to know what the relation between higher-level phenomena and their underlying mechanisms is. Our present purpose is to argue that neurobiological entities are identical to configurations of fundamental physical property tokens and that levels of mechanisms are levels of descriptions rather than levels of being. Consider the relation between two levels of mechanisms, for instance the hippocampus having a causal role ψ in the brain’s organization and the hippocampus’s constitutive components. According to the mechanistic account, the causal role ψ of hippocampus is explained when neuroscience isolates an actual model of the hippocampal functioning that explains how the organization and the activities φ1, …, φn of constitutive hippocampus components c1, …, cn gives rise to ψ . The crucial point is to note that this implies that these components, their activity and their spatial organization are sufficient to give rise to ψ, which can be expressed in terms of supervenience of ψ on c1, …, cn, φ1, …, φn and their organization o considered as a whole. For anything that has ψ, there is a lower level property, for instance the property ψ# consisting of being composed of c1, …, cn that have respectively causal roles φ1, …, φn and that are organized in a o-way, such that anything that has ψ# has ψ. The supervenience of the phenomena, qua being causally defined, on their respective underlying phenomena has the same theoretical foundation as the supervenience of mental property tokens on physical property tokens, namely the causal completeness of physics. This ensures that no “new” causal powers will appear above the causal powers of fundamental physical property tokens. Phenomena situated at a higher level of mechanism inherit their causal power from the causal powers of their constitutive components organized in a specific way. If this were not the case, then it would be impossible to explain completely the causal role characterizing a phenomenon P by referring to the organization of the causal roles of the constitutive components of a mechanism M for P, as required for a mechanistic

Psychology and neuroscience

159

account. As we argued in the last sub-section, the mechanistic scheme of explanation is reiterable down to fundamental physical property tokens. It follows that the causal powers of any phenomenon P are inherited from those of the fundamental property tokens that are transitively the constitutive components of P. Once it is accepted that a phenomenon P supervenes on the organization of the causal roles of the constitutive components of a mechanism M for P, Kim’s supervenience argument for token identity applies as well to the case of the relation between levels within the mechanistic account, to the extent that we endorse also the premise of completeness of physics and the premise of the absence of causal overdetermination applied in the present theoretical context.47 If a brain

47

In reason of the importance of Craver’s account of mechanistic explanation for our development, we want to make clear that we disagree with Craver with respect to his interpretation of Kim’s supervenience argument. According to him, Kim’s argument does not apply to levels of mechanisms because the activity φ of individual components of a mechanism M for P does not compete with the overall causal role ψ of the phenomenon P (2007: 216). Accordingly, he interprets the physical property tokens symbolised by the Ps in Kim’s writings (Kim 2005: chapter 2; we respect stricly Kim's notation here, although the correspondance with our development of chapter 3 is obvious) as being components of mechanism M for a phenomenon P. Under that interpretation it is obviously not the case that the causal contribution φ of a constitutive component ci of a mechanism M competes with the causal activity ψ of M. However, it seems mistaken to consider Kim’s Ps as representing individual components of a whole, which is sufficient for the effect P*. Kim’s Ps designate instead an entire configuration of physical property tokens, which is comparable to an entire mechanism, having, qua being a configuration of physical property tokens, an internal organization of parts, each playing a causal role in the causal behaviour of the configuration as a whole. There are at least two reasons for that. First, Kim’s physical property tokens Ps should be imagined to be the supervenience basis for a higher level property token M. Obviously, it is generally not the case that higher level property tokens supervene on a single physical property tokens such as an electron, a photon or any kind of such microphysical particle, since the supervenience basis is sufficient for the supervening property. That is why it is more reasonable to assume that Kim’s Ps stand for configuration of physical property tokens. Second, Kim’s argument works using the

160

From psychology to neuroscience

phenomenon P causes a brain event P* when exercising its causal role ψ, the existence of mechanism M for P upon which P supervenes explains completely why P* occurred. Against the background of the presupposition according to which M for P and P are ontologically distinct entities, the premise of absence of systematic overdetermination and the premise of causal completeness of physics suggests that we should prefer M rather than P as being the cause of P*. In other words, Kim’s argument suggests that if the phenomenon and the mechanism are ontologically distinct entities, the phenomenon is causally impotent, the real causal work being done by the mechanism’s components. By extension, any phenomenon above a fundamental configuration of physical property tokens, i.e. the most fundamental components of mechanisms, is an epiphenomenon. We can avoid this conclusion if we reject the ontological distinctness of phenomena and mechanisms and instead identify for ontological purposes any particular phenomenon explained completeness of physics: P is assumed to be a sufficient cause for the effect P* and that is why P competes with the higher level property token M. Moreover, Craver argues that Kim’s argument works only if a lower level property P#, having causal role φ#, and the higher level property M# having causal role ψ, are related by a realization relation such that P# realizes the property M# and that both properties are in competition for the production of the same effect E. Craver defines P# having causal role φ#, as the property of having the various components, the various activities and the various organizational feature of a mechanism. Let us here emphasize firstly that Kim’s argument runs on the basis of the supervenience relation solely. Secondly, the subvenient physical property competing with the supervenient property is not a single isolated microphysical property token, but a configuration of physical property tokens, qua being a sufficient basis and qua being a sufficient cause for the effect E. In this context, a configuration of physical property tokens is a collection of individual property tokens organized so that the respective causal role of the individual property token are a sufficient cause for E and a sufficient supervenience basis for M#. There is consequently no significant difference between the configurations of physical property tokens to which Kim’s Ps refer and Craver’s mechanisms. Given this line of reasoning, it seems rational to consider that the supervenience argument applies as well to the case of the relation between Craver’s phenomena and mechanisms.

Psychology and neuroscience

161

mechanistically by neuroscience with its underlying mechanism. Since the relation of constitutive componency is transitive, it amounts obviously to claim that any phenomenon, at any level of mechanism, is identical to the fundamental constitutive physical components of their various components at different intermediate levels. The hippocampus is for instance identical to its constitutive components c1, …, cn and their organization, but also to the overall organization of the constitutive components of each ci, and so on, down to the fundamental physical constitutive components. The claim that neuroscience spans multiple levels of explanations should therefore be understood as consistent with the epistemically layered view of the world (3.7.5 Token-identity and the layered model of the world). From the point of view of knowledge, there really are multiple levels of organization in the nervous systems. Molecules are organized such as they form cells, which are in turn organized in such a way that they form cells layers, brain areas such as the primary visual area, cortical structures as the hippocampus, lobes, hemisphere and finally the brain. But ontologically, such higher-level entities are identical to configurations of fundamental physical property tokens. Neuroscience spans multiple levels of description in order to explain the active organization of components contributing to the realization of phenomenon observable at a higher level of description, with the whole connected via mechanistic explanations. The levels we speak of in neuroscience are levels of description, not of reality. To this argument, let us add some metaphysical remarks. Craver tries as hard as possible to remain neutral with respect to metaphysics for three main reasons. First, classical models of reduction involve the establishment of inter-level type identities,48 which are not available in the case of the relation between phenomena and mechanisms, since the former are multiply realizable by the latter. Rats, humans and elephants have spatial memory but do not have identical hippocampal structures, which obviously suggests that the path to the same phenomena can be 48

For instance, Craver wrote that “according to classical reduction models, levels are integrated by identifying the kind-terms describing phenomena at one level with the kind-terms describing the phenomena of another” (2007: 256).

162

From psychology to neuroscience

realized by multiple mechanisms. Craver takes this to be an argument favouring non-reductive physicalism (2005: 375). Later, we will deal with the issue of the relation between multiple realization and epistemological reductionism – for the moment, let us simply reiterate that multiple realization does not affect ontological reductionism as adumbrated through the token-identity thesis. Nothing about ontological reductionism makes it impossible that individuals with different cerebral endowments and organizations can manifest a higherlevel ability such as spatial memory realized by different mechanisms. However, each individual has one and only one underlying mechanism, which shows that the multiple realization argument does not affect token-identity. A second important reason of Craver’s metaphysical agnosticism is the causal relevance of higher level phenomena both in top-down and bottom up experiments (2007: 196). As we have seen, such experiments are the core investigation techniques used to isolate the constitutive components of a given mechanism. Obviously, if Kim’s supervenience argument is taken to lead to the epiphenomenality of any non-fundamental property token, which is how Block interprets it (2003: 138), then it is impossible to expect from an intervention on a higher-level phenomenon P to have an impact on the components of the mechanism for P. However, this interpretation is mistaken for the two following reasons. Firstly, as long as higher-level phenomena supervene on physical property tokens, intervening on higher-level phenomenon implies changes at any subvenient level. Supervenience implies that any higher-level change has a corresponding lower-level change (3.3 Supervenience). This means that when an experimenter asks an individual to undertake a task, the undertaking of that task causes changes both at the higher level and at the lower level. As long as supervenience holds, it is simply not possible to intervene on a higher level phenomenon, as pronouncing words for instance, without intervening on the supervenience basis of that phenomenon, inducing some neural change. But more importantly, it is not the case that the token-identity thesis confers the status of epiphenomena to higher level phenomena. As we have argued, the thesis was formulated precisely to avoid this

Psychology and neuroscience

163

epiphenomenal character of higher level phenomena, whether they are mental properties or cognitive abilities such as spatial memory. Tokenidentity excludes the interpretation that higher-level phenomena are epiphenomenal. Against that background, it is clear that if one asks a subject to undertake a higher level task such as pronouncing words, then this undertaking causes changes in the subject both at the behavioural level and at the neurological level. Token-identity does not affect the reliability of inter-level experiments. Quite to the contrary, it claims that any change that is observable from a higher level point of view are observable from a lower-level point of view as well. Finally, Craver argues that reduction provides a mistaken interpretation of the way in which neuroscience makes advances,49 since reductionists ignore certain upward-looking phase in empirical research and are unable to account for intra-level integration processes (2005: 375). An example of the first case is the historical process having led to the idea that LTP is a component of the mechanism of spatial memory and an example of the second is the integration of results from different field to uncover the mechanism underlying LTP. Let us briefly respond in the following way. First, reductionism is classically a theoretical regulative hypothesis developed to satisfy the philosophical insight that science is a united whole, rather than a normative commitment that empirical sciences have to follow or an historical thesis with respect to the becoming of science. In practice, it is clear that empirical research advances in multiple directions and that theories co-evolve in a non-systematic way (Churchland 1986). This includes of course upward looking reflexion, implementing interfield unifications, but this has nothing to do with reduction as a regulative ideal for unity of science. Secondly, as we have shown, token-identity does not imply that the only real causal relations to be discovered are fundamental physical causal relations, since identity is conservative (3.7.2 The conservative character of strict identity). Token-identity has 49

Craver does not clearly distinguish between epistemological and ontological reduction. Given that most of the time he refers to the classical model of intertheoretic reduction, we assume that he takes reduction to mean both epistemological and ontological reduction.

164

From psychology to neuroscience

therefore no such implication with respect to what scientists should be looking for. 4.3.4 Mechanistic explanations as comprehensive explanations As we have shown, neuroscience provides mechanistic explanations of the production of behaviour by the nervous systems, rooting the causal evolution of neurobiological systems firmly in fundamental laws of nature, which grasp the causal dispositions of fundamental physical property tokens. Given this rootedness in the physical word, one might ask why neuroscience does not aspire to account for the physical origins of behaviour directly on the basis of the huge configurations of physical or biochemical property tokens, in order to link brain state types with types of behavioural responses directly by horizontal causal laws. By brain state types, we mean a complete description of a brain at a certain time, which describe the connectivity of its components and their current properties, such as spiking or not if the component is a neuron. Such types are obviously constructed types, made out of large conjunctions of simpler types referring to constitutive components of the system, their activity and their organization. By behavioural types, we mean behavioural responses falling under a description such as “doing action a”. The reason that the neuroscientific explanations do not formulate such horizontal causal laws is tied to the complexity of its target of inquiry. Systems with this kind of complexity impose practical but nonetheless very strong constraints on the possible strategy of empirical investigation available to explain the functioning of complex neurobiological systems such as the human brain. First of all, it is practically not possible to observe and describe entirely a brain state token from the point of view of biochemistry. Secondly, building such a description by observing the same brain state token over and over is a hopeless task, since the fact that the brain of a given individual is in constant evolution and combinatorial considerations suggest that it would be wildly improbable to encounter twice the same token of brain state type in the history of the universe. If we consider in a simplified way that the description of brain states we are looking for take into account the right number of neurons and whether or not there are

Psychology and neuroscience

165

firing, the number of possible brain state types describing a nervous system composed of 85 billions of neurons is something near 2 to the power of 85 billions. Third, even if the observation of such a huge number of variables would be feasible in some way, the corresponding descriptions would be simply of no use for any practical purpose, being too large to give rise to any kind of concrete computation or prediction. Fourth, if the availability of such descriptions is already more than seriously compromised, the fact that there are so many possible brain state types prevents the practical possibility of inferring function f(x1, … xn) → {A, B}, where (x1, … xn) is the vector activation of a neuronal network of n neurons and A and B designate behavioural responses such as choosing reward A or reward B, because it is not possible to observe all brain states types and to know which one codes for which behavioural response.50 Finally, given that each particular brain within a single species is different, having a slightly different number of neurons and different synaptic connections, it would be necessary to find the function f that is specific to each individual, or even worse, if we take into account that the brain of a given individual is in permanent evolution. The situation is obviously even worse is we look for more detailed neurobiological description, biochemical or even physical description of brain states. The mechanistic explanations of neurosciences are not intended to discover the complete set of mapping functions that will describe the entire cerebral system, from the atomic level to that of high level functions. Accounting for cognitive phenomena by isolating actual models of the mechanisms underlying them, neuroscience thereby considers the causal relations obtaining the different constitutive components a mechanism in order to explain their contribution to the

50

We assume here that the all neuron in the net are causally relevant to the production of the behavioural response. If this is not necessarily the case for the human brain, the complexity problem arises even is number of relevant neuron is small. For instance, if this number is 100 neurons, there are 2100 possibilities mapping on two possible behaviours, what is already largely too many to complete the observations of brain state relevant types, and infer the mapping function f.

166

From psychology to neuroscience

target phenomena and relate only indirectly to the components to the behaviour by considering the phenomena itself as a constitutive component of a higher level model. For instance, if one is interested in why receptor fields on the retina react to changes of luminescence, one investigates the connectivity between the different neurons that are constitutive of a single receptor field in order to subsequently build a model of the functioning of this receptor field. On that basis, this model can be replicated in order build a more general model of the receptor fields of the retina and it can be assembled with an accurate model of the connectivity of the visual system, and other models, until a hypothetical global model of the brain is constructed step by step. Note that within this framework, brain states types are available, not by any kind of direct observation, but rather through this process of combination of restricted models explaining the individually functioning of components of the brain. This research strategy contrasts advantageously in two ways with the objective of considering directly the relation between brain state tokens and behaviour. First, it is not the case that the scientist has to observe brain state tokens as a whole, nor that they have discover the mapping functions connecting different types of systems to their overt behaviour. We avoid the difficulties enumerated above while preserving the ability of neuroscience to provide genuine explanations of the functioning of the brain. Neuroscience enables us understand how the brain, as a highly hierarchized collection of interacting mechanisms, may exhibit behaviours as rich as the human one. This includes complex abilities such as speech or face recognitions, even if, of course, current neuroscience is not a complete theory. Finally, a note on the concept of comprehensive explanations. Qua providing mechanistic explanations, neuroscience enables us to understand how a behaviour is causally produced by the brain and how this causal relation is rooted in fundamental physical causal dispositions of brain’s constitutive components, but without formulating laws that directly link types of brain state and behaviour. Nonetheless, mechanistic explanations, although bypassing the practical problem of observing in detail brain states as wholes and of formulating general laws linking these types of brain state with behaviour, does not contain within itself

Psychology and neuroscience

167

any constraints in principle that block the theoretical possibility of explaining completely how fundamental components of the brain bring about the behaviour, since the causal disposition of these components is firmly rooted in physical laws.

4.4 Summary and transition In this chapter, we firstly argued that theories are networks of interrelated concepts that aim to account for the relation obtaining between occurrences of property tokens of a certain kind. Theories take entities in the world and place them in ordered taxonomies according to some consistently developed essential criteria. In the context of ontological reductionism, property types are descriptive categories we use to classify property tokens against the background of a given taxonomy. There are currently two main theories that explain the production of human behaviour, namely psychology and neuroscience. We argued that psychology is a functional theory of the mind and that mental property types have to be functionally defined with respect to their contribution to the production of the behaviour and of further mental property tokens in certain environmental circumstances. Examining different versions of functionalism, we found that only common-sense functionalism can account for the causal insertion of ordinary mental properties in the natural world and grasp the subtle interactions by mean of which a plurality of ordinary mental property tokens produces the behaviours in question. With respect to the ontological status of mental properties, we argued that causal role functionalism, which takes mental property tokens to be second order properties, is an untenable position, given our argument in Chapter 3. Realizer functionalism, which considers that mental descriptions refer directly to configurations of physical property tokens, is a tenable position to the extent that a causal theory of properties is assumed, for which we found independent reasons. Therefore, folk psychology is a theory that classifies entities that are identical to configurations of fundamental physical property tokens according to the criteria that are formalized by common-sense functionalism.

168

From psychology to neuroscience

Neuroscience explains how the nervous system produces the behaviour. Due to the complexity of such systems, it is neither possible to observe brain states directly in a sufficiently detailed way, nor it is possible to infer general causal laws linking these brain states directly to the behaviour. Neuroscience aims rather to build mechanistic explanations spanning multiple levels of mechanisms and binding together multiple fields of investigations. Within that framework, higher level phenomena are explained by their respective underlying mechanisms, such that the organization of the activities of the constitutive components of a mechanism tells us how the mechanism operates for our target phenomena. By putting empirical constraints on the set of possible mechanisms that must be considered in an account of a given phenomena, neuroscience aims to build actual models of mechanisms that account completely for them. Multiple levels of mechanisms are unified when an entity is integrated as a component in a mechanism for a higher level phenomenon, on the one hand, and on the other, when the activity of this entity is explained by a lower level mechanism as a phenomenon. We argued that mechanistic explanations are reductive explanations, since the causal behaviour of a whole is explained by the causal behaviour of its constitutive components and their organization. Cognitive phenomena are consequently reductively explained down to the causal interactions obtaining between biochemical property tokens, which may be reductively explained by the causal interaction obtaining between the fundamental physical property tokens if requested. The causal behaviour of higher level phenomena is therefore firmly grounded in physical laws: the ultimate components of mechanism are fundamental property tokens and their activities are nothing but the fundamental causal disposition of those tokens. Since the causal behaviour of components is sufficient to explain completely the behaviour of mechanisms as wholes, and since the fundamental components of mechanisms are ultimately physical, higher level phenomena supervene on fundamental physical property tokens. This opens the way to the application of the argument for token-identity, with that result that higher level phenomena are identical to configurations of physical property tokens. Finally, we argued that

Psychology and neuroscience

169

neuroscience provides comprehensive explanations of the production of behaviour by the nervous system in the following sense. The complexity of such systems forbids building complete models of the production of actual behaviours by actual brains on the basis of which detailed behavioural previsions can be done; but nonetheless, neuroscience provides models that allow us to understand how complex configurations of physical property tokens may give rise to a behaviour as rich as the human one. Against that background, we shall turn to the question of the relation obtaining between those two theories that provide two different kinds of explanations of the production of the same phenomena, namely the human behaviour.

Chapter 5 EPISTEMOLOGICAL REDUCTIONISM In this chapter we will begin with a discussion of the weaknesses of the main contemporary accounts of epistemological reductionism in order to highlight the requirements that the account, which will be proposed in the main section of this chapter, has to satisfy. We shall therefore start by introducing some of the positive motivations for reducing psychology to neurosciences. In a nutshell, in so far as both psychology and neuroscience explain the genesis of human behaviour, it is natural to expect that these descriptive modalities are closely related. We shall turn, then, to the classical Nagelian account of intertheoretic reduction, which defends the view that psychological laws are ideally completely derivable from the lower-level laws. This account faced heavy criticism in the 60s as the multiple realization argument came into play, since the derivability of laws supposes the establishment of bi-conditional bridging principles between the theoretical terms of each theory. As we shall see, both the New Wave account of inter-theoretic reduction, which can be seen as a weaker version of the classical reductionism, and functional reduction, which tries to turn the idea of a reductive explanation into a model of intertheoretic reduction, pay a heavy tribute to the multiple realization argument in the form of being driven to eliminate psychology as a genuine and legitimate science. Indeed, the multiple realization of mental properties raises an uncomfortable dilemma. On the one hand, when viewed in the light of the project of ontological reductionism, the multiple realization thesis suggests that the patterns of regularities that are supposed to make true psychological law-like generalization do not exist at all, which suggests eliminating psychology from the list of sciences, where it is a condition

172

From psychology to neuroscience

that the entities coming under the same description are somehow identical. Multiple realization infers thereby that psychological taxonomy is ill formed. On the other hand, if one tries to vindicate the scientific legitimacy of psychology as grasping genuine identical causal relations between entities that are physically different, one has to consider the psychological properties exemplified by these entities as non-identical to their physical properties. But this amounts rejecting ontological reductionism with all the consequences that we have seen in previous chapters: the violation of the premise of the closure of physics or epiphenomenalism.

5.1 Why epistemological reductionism? Epistemological reductionism aims to establish systematic relations between any two theories, A and B, such that, one of them, here B, is able to explain, on the basis of its own explanatory principles, the explanatory principles of the other. As I shall show in more detail below (5.2.2.1 Deductibility of laws), such explanatory relations between theories usually take the form of the deduction of laws. A theory A is reducible to a theory B to the extent that the laws of A are deducible from the laws of B. Accordingly, our present concern is to know whether or not the law-like generalizations of psychology are deducible from the law-like generalizations of neuroscience and physics. Before discussing in detail the logical aspect of the reducibility of psychology, let us introduce three motivations for the claim that psychological descriptions, laws and explanations are reducible to neurobiological explanations. Firstly, consider an entity e1 that exemplifies both physical and mental properties. According to the claim of strong supervenience, the physical properties that e1 exemplifies determine its mental properties completely. Now, if e1 falls under a certain complete physical description P1 in virtue of its physical properties, and if these physical properties determine the mental properties of e, the complete physical description of P1 of e1 implies that e1 falls under a certain mental description, M1. Note the emphasis on the phrase, a complete physical description of e1. It is not the case that mental properties supervene on isolated physical

Epistemological reductionism

173

properties such as “having a certain mass” or “having a charge”. Mental properties only supervene on complex configurations of physical property tokens that are a sufficient supervenience basis for the exemplification of the mental properties in question. By a “complete physical description of e1”, we mean a description of e1 that contains any physical predicate that is true of e1.51 Inversely, suppose that two entities e1 and e2 differ with respect to the mental descriptions they make true. Since strong supervenience implies that a mental difference indicates a physical difference, if e1 makes true mental description M1 and e2 makes true mental description M2 in virtue of their respective mental properties, then e1 and e2 must fall under different complete physical descriptions P1 and P2 as well. The asymmetric character of supervenience has its counterpart with respect to the descriptions. In the context of strong supervenience, the physical properties exemplified by e1 determine its mental properties, which means that the latter are dependent on the former. But the relation isn’t necessary symmetrical:52 the mental properties of e1 don’t determine its physical properties. Therefore it is also not the case the mental description of e1 implies that e1 falls under a precise physical description P. Strong supervenience just implies that e also falls under some physical description, without entailing anything more with respect to the precise content of this description. The fact that the physical description of an entity e1 determines the mental description of this entity strongly suggests that the psychological laws can be deduced from those physical laws that grasp the causal behaviour of e1. Firstly, according to the arguments in Chapter 4 (4.2.4.2 The causal theory of properties), a physical property token simply is the power to cause further property tokens. Configurations of physical property tokens can therefore be construed 51

We shall slightly restrict that claim later (6.3.1 Causal similarities and the functional individuation of psychological types).

52

Strong supervenience is a determination relation designating the determination of a kind of properties by another. Thereby, it is prima facie asymmetric. Although strong supervenience does not imply symmetry, it is nonetheless compatible with it.

174

From psychology to neuroscience

as more complexly structured powers to cause further configurations of physical property tokens. The laws of physics grasp the causal behaviour of the physical property tokens. Laws of psychology are supposed to grasp the causal behaviour of the mental tokens, which are functional property tokens, i.e. powers to cause both further mental property tokens and configurations of physical property tokens to occur. Now assume an entity e1, which falls under the complete physical description P1 and the mental description M1. We have just argued that P1 entails M1. If the fact of falling under M1 implies from the psychological point of view that e1 produces a behaviour b1 coming under description B1, how could it be in the general case that the description P1 of e1 fails to imply that e1 produces a behaviour b falling under B1? For it not to be the case, we would have to jettison our reasons for considering the mental properties of e1 as supervenient on its physical properties, since, after all, we accepted the strong supervenience argument precisely because mental properties and physical properties of e1 have the very same effect (3.3.1 Varieties of supervenience). Secondly, in the context of ontological reductionism, this suggestion is even more justified. If mental properties of a particular entity e1 are token-identical to its physical properties, then the causal relations obtaining between the mental property of e1 and its corresponding behavioural effects and between the physical property of e1 and these very same effects are ontologically speaking one single causal relation. In that case, the laws of physics explain why e1, qua falling under the physical description P1, caused behaviour b. However, the causal relation obtaining between e1 and b is also grasped by psychology, which explains why e1, qua falling under the mental description M1, caused the behaviour b. Given that physical laws and explanations are more complete and have a larger domain of application, whereas psychological laws and explanations are less complete and have a smaller domain of application (3.7.5 Tokenidentity and the layered model of the world), it suggests that physics explains in a more complete way, with reference to more general laws, the reason why e1 caused b as well as the reason why e1 falls under

Epistemological reductionism

175

mental description M1. Thus, physics explains why the causal relation between mental properties of e1 and b obtain. Finally, the irreducibility of psychology would lead to a contradiction if we premise that the argument for ontological reductionism is true. Consider the steps in the following argument, which proceed from assuming, (1) There are true law-like generalizations in psychology that cannot be reduced to the laws of physics. Psychological law-like generalizations are true in virtue of causal relation obtaining between entities in the world. However, if it is not possible to deduce the law-like generalization of psychology from the laws of physics, it means that physics cannot appropriately grasp those causal relations. (2) Psychology grasps by true law-like generalizations causal relations that physics is unable to grasp. If physics is unable to grasp a causal relation, then at least one of the relata of the causal relation is not suitably describable by physics. Consequently, (3)They are entities in the world that physics cannot appropriately describe. The third step pushes us to accept either that there are physical property tokens for which physics cannot give an suitable description, or that these entities are not themselves identical to configurations of physical property tokens. Note that the first option is not an empirical observation about the current state of physics, but rather a statement that physics is in principle unable to generate an appropriate description of those physical property tokens, which seems absurd. This suggests obviously the second option, namely to consider that property tokens are not identical to configurations of physical property tokens and that they are essentially non-physical. (4)There are entities that are not identical to configurations of physical property tokens.

176

From psychology to neuroscience

Obviously, if they are causally efficacious property tokens that are not identical to configurations of physical property tokens, then ontological reductionism is false. (5) Therefore, ontological reductionism is false. This last conclusion contradicts the premises with which we started out, thus establishing that the reasoning leading to (5) should be rejected. This rejection is, in effect, the rejection of the premise (1). And this means that either there are no law-like generalizations in psychology, or that they are reducible in principle to the laws of physics. If we reject the elimination of psychological law-like generalizations (cf. 5.4.1.3 NWR and eliminative materialism), we are left with epistemological reductionism, as, at the very least, a theoretical possibility.

5.2 Classical reductionism and the requirements of reduction 5.2.1 Inheritance of the DN-Model of explanation Ernest Nagel (1961) articulated the classical account of inter-theoretic reduction, which became the standard for epistemological reductionism. The Nagelian model of reduction can be interpreted as the natural descendent of the deductive-nomological model of explanation (hereafter DN-model) in the philosophy of science. According to the DN-model, an explanation is composed of an explanandum, a sentence “describing the phenomenon to be explained” and an explanans, which is “the class of those sentences which are adduced to account for the phenomenon” (Hempel and Oppenheim 1965: 247). The central insight of the DN-model is that the sentences of the explanans logically entail the explanandum just as the steps of an argument logically entail its conclusion. The Nagelian reinterpretation of the DN-model in terms of inter-theoretic reduction considers that for two theories, theory T1 and T2, assuming that the laws of the latter are supplemented by correspondence principles linking systematically the terms of each theory, then if the laws of T1 can be reduced to the laws of T2, the latter logically entails the laws of T1. Let us examine in more

Epistemological reductionism

177

details the Nagelian requirement of deducibility and the role of these correspondence rules. 5.2.2 Requirements for reduction 5.2.2.1 Deductibility of laws The first requirement for epistemological reduction is the deducibility of the laws of the theory from the laws of the more fundamental theory. Assume that we have a true psychological law-like generalization Lpsy asserting that, ceteris paribus,53 property tokens falling under the mental description M1, for instance “perceiving a tiger”, cause property tokens falling under the mental description M2, for instance “being afraid of the tiger”.54 To reduce Lpsy we have to show that there is at least one physical law implying that the entities falling under the description M1 cause entities falling under description M2 as well as the fact that those entities can be retranscribed in a physical vocabulary. The fact that such a law exists is secured by the token-identity thesis, which asserts that any mental property token is identical to a configuration of physical property tokens and may therefore be physically described. A configuration of property tokens coming under the description M1 will therefore come under a physical description, P1. The same holds for property tokens coming under M2, which allows for physical description, P2. As argued previously (3.6 The causal argument for the token-identity), token-identity implies that there is in such a case one single causal relation between two different configurations of property tokens, each of them coming under both a mental and a physical description. Now, since the causal relation is

53

Psychological laws are subject to important exceptions and are therefore said “ceteribus paribus”.

54

In what follows we shall make the simplifying assumption that mental types are defined just by their typical effect, rather than by being defined by a disjunction of typical causes, a disjunction of relation to other mental properties and a disjunction of possible behavioural effect, which is the more precise but cumbersome definition we explored in Chapter 4. This has the advantage of allowing us to avoid unnecessary complications in our present discussion.

178

From psychology to neuroscience

always an index of a nomological relation (3.4.1 Completeness of physics), it follows that there is a law of physics, say Lphys, which grasps the causal relation between P1 and P2. Note that it is not necessary that Lphys pre-exist in the more fundamental theory. If there is currently no physical description of a given entity coming under description M1 and therefore no law describing its specific causal behaviour in physics, it is nonetheless always possible to construct a description such as P1 out of the conjunction of simpler physical descriptions already existing.55 The same holds for the physical description of the entity coming under description M2. From such conjunctive descriptions, we obtain a physical law that governs the causal relation obtaining between both entities, described respectively by P1 and P2. However, the whole question of the need for bridging laws arises from the following worry: the fact that it is possible, for any entity making true both a mental description M1 and a physical description P1, to construct an appropriate physical law, is not sufficient for the deducibility of Lpsy from Lphys. The reason is that, although the causal behaviour of the tokens coming under the physical description P1 is governed by both Lpsy and Lphys, this does neither mean that Lpsy applies to all entities whose causal behaviour is grasped by Lphys, nor that the reverse relation holds, namely that Lphys applies to all entities to which Lpsy applies. The reason is that each theory is formulated in a specific vocabulary whose terms are neither semantically nor extensionally equivalent. Much of the discussion about reduction concerns the fact that Nagel’s theory supposes the deduction of the higher level laws from the introduction of correspondence principles between psychological and physical descriptions (Nagel 1961: 347). It is common to refer to these correspondence principles using the expression “bridge laws”. But supposing bridge laws isn’t the same as giving the rules for constructing them, a topic to which we now turn.

55

Hooker (1981) explains in detail how to construct physical concepts in order to reduce special sciences description, laws and theories.

Epistemological reductionism

179

5.2.2.2 Requirements for bi-conditional bridge laws Let us consider in more details the deductive relation obtaining between the laws we introduced in our simplified example. Formally, we can formulate both laws as follows: Lpsy : ∀x (M1x → M2x) Lphys : ∀x (P1x → P2x) Lpsy expresses the claim that, ceteris paribus, for any entity x, if x comes under the mental type M1 then x causes an entity coming under the mental type M2. Lphys similarly expresses the claim that for any entity x, if x comes under the physical type P1 then x causes an entity coming under the physical type P2. We can deduce Lpsy from Lphys if Lphys entails Lpsy. However, this entailment holds only if we can show that any entity coming under description P1 comes under description M1 as well, and vice-versa, namely that any entity coming under P2 comes under description M2 as well. Formally, Lphys entails Lpsy just in case of the validity of Bridgepr1, which secures the co-extensionality of the descriptions P1 and M1, and of Bridgepr2, which secures the coextensionality of the descriptions P2 and M2.56 Bridgepr1 : ∀x (P1x ↔ M1x) Bridgepr2 : ∀x (P2x ↔ M2x) Bridgepr1 and Bridgepr2 can be reformulated as: Bridgepr1# : ∀x (P1x → M1x ∧ M1x → P1x) Bridgepr2# : ∀x (P2x → M2x ∧ M2x → P2x) The co-extensionality of descriptions is a necessary condition for the deducibility of laws. Assume for instance that Bridgepr1 is false. There are here two possibilities. The first possibility is that ∀x (P1x →

56

Nagel himself did not explicitly argue in favour of the bi-conditional requirement, since he focused on the strict deducibility of laws. However, it is customary to interpret this model of reduction as requiring bi-conditional principles (Kim 1998: 90), since, as will be argued later, one-way conditional are insufficient to vindicate the generality of higher level laws and explanations (5.3.2 What is left out by one-way conditional approaches).

180

From psychology to neuroscience

M1x) is false. This possibility has been already ruled out since we accepted that the physical description of an entity determines its mental description (3.7.4 Token-identity and supervenience). Therefore, each time an entity x come under description P1, not only Lphys holds for that entity, but this entity comes under description M1 as well, with that result that Lpsy holds as well. The same goes for Bridgepr2. All of this is unproblematic. Suppose now that either ∀x (M1x → P1x) is false, as claimed by the multiple realization argument (3.2 Multiple realization in an ontological context). In this case, there is at least one entity x that comes under description M1 and that cause this entity to come under M2 according to Lpsy but that fails to come under description P1. Ontological reductionism tells us that there is a physical description P1* such that x comes under P1* and thereby there is a physical law Lphys* such that any entity x coming under the physical description P1* causes this entity to come under a physical description P2*, which itself implies that this entity comes under M2 as well. In this context, there are different laws from which Lpsy can be deduced. More precisely, there is a disjunction of physical laws that enables the deduction of the psychological law-like generalizations. We will consider whether this case is a knock-down objection against reduction in the next section (5.3.2 What is left out by one-way conditional approaches); for now, assume that bi-conditional bridging principles are required for epistemological reduction. Let us finally adumbrate more precisely the meaning of bridge laws in the Nagelian account. Firstly, it is important to remember that the bridge laws supplementing the reducing theory in the reduction of another theory are embedded in our metatheoretical discourse, and are not additional laws of nature. Given ontological reductionism, they do not link distinct property tokens as laws of nature. Rather, they are operational principles that are constructed to establish a correspondence between descriptions of different sciences. Yet they are law-like in that they are nomologically necessary, since, if the correspondences grasped by bridge laws were accidental, then the causal completeness of physics and ontological reductionism would fail. To illustrate this, take our previous example. Against the

Epistemological reductionism

181

background of token-identity, both mental property tokens are describable in the terms of physics, respectively by descriptions P1 and P2. Suppose now that the bridge law linking descriptions M2 and P2 is accidental, meaning that an entity falling under M1 and P1 could cause an entity coming under M2 but not under P2 to occur. Since bridge laws are supposed to be bi-conditional, it means that second entity does not come under any physical description. In this case, obviously ontological reductionism and the causal completeness of physics collapse, due to the fact that there is an entity which is completely explained by a previous mental property tokens, but which is not itself physical. Secondly, the bridge laws supplementing a more fundamental theory T2 should not be considered to redefine in any way the descriptions of the less fundamental theory T1. It is not the aim of epistemological reductionism to reduce the meaning of descriptions of one theory to the one of the descriptions of another theory. Presumably, even in successful cases of the classical inter-theoretic reduction, the meaning of the terms of the reduced theory is preserved. For instance, although thermodynamics has been shown to be reducible to statistical mechanics, this has not changed the meaning of the concept “temperature”. Most importantly for our argument, to conceive bridge laws as providing definitions of the theoretical terms of T1 within the vocabulary of T2 would logically imply arguing in favour of the elimination of T1. If T2 is a more complete theory, and if the entire vocabulary of T1 can be redefined in the terms of T2, then it seems that we can completely eliminate T1. Let us sum up classical reductionism. According to the Nagelian account, to reduce a theory T1 to a theory T2 means that the laws LT11, … LT1n of T1 are deducible from the laws LT21, … LT2m of T2. Given that deduction operates on the formal aspects of these laws, it is necessary to introduce bridge laws connecting the theoretical terms of each theory in order to ensure that each pair of reduced and reducing laws refers to the same entities in the world. Bridge laws take therefore the form of nomologically necessary bi-conditional correspondence principles. The establishment of such bridge laws does not require that a pre-existing theoretical term in T2 correspond to each theoretical term of T1. Such

182

From psychology to neuroscience

theoretical terms can be conjunctively constructed on the basis of previous resources of T2, with the result that law-like generalizations of psychology are embedded in physics.

5.3 Multiple realization in an epistemological context 5.3.1 Multiple realization against classical reductionism In Chapter 3, our discussion of the implication of the multiple realization argument for the type-identity thesis largely opened the way for evaluating the use of this famous argument in the debate about epistemological reductionism. To put it in a nutshell, the multiple realization thesis suggests that the bi-conditionals required by classical reductionism are counter-indicated by the possibility that different type of configurations of physical property tokens are associated with the same mental property token. The multiple realization argument works in an epistemological context therefore as an argument against the epistemological reduction of psychology to physics. Let us come back to our example of the last section. A psychological law such as Lpsy can be reduced to Lphys if and only if conditions Bridgepr1 and Bridgepr2 are satisfied. Lpsy :

∀x (M1x → M2x)

Lphys :

∀x (P1x → P2x)

Bridgepr1 : ∀x (P1x ↔ M1x) Bridgepr2 : ∀x (P2x ↔ M2x) The multiple realization argument attacks the bridging principles by showing that it is not necessarily true that ∀x (M1x → P1x) and ∀x (M2x → P2x). This means that both conditions Bridgepr1 and Bridgepr2 fail, and that Lpsy cannot be deduced from Lphys. Rather, the multiple realization argument asserts that at least certain, if not all, mental descriptions are co-extensional with the open disjunction of physical descriptions. Accordingly, Bridgepr1 and Bridgepr2 should be rewritten as follows.

Epistemological reductionism Bridgepr1 MR :

∀x (P1ax ˅ P1bx ˅ … ˅ P1nx ↔ M1x)

Bridgepr2 MR :

∀x (P2ax ˅ P2bx ˅ … ˅ P2mx ↔ M2x)

183

There are three immediate remarks to be made in order to get clearly into focus what grounds the multiple realization argument and the difficulties it raises for epistemological reductionism. First of all, the question is to know what exactly grounds the claim. As previously argued (3.2 Multiple realization in an ontological context), there is actual empirical evidence for multiple realization. Mental descriptions actually apply to entities that are physically different. Different people, with different brains are taken into account psychologically by the same mental description. For instance, the mental description “desiring to drink something” can not only, on a trivial level, involve localized activity in the different brains of different people, but it can also involved different localized activity in the brains of different individuals, for instance those exhibiting important neurological abnormalities such as brain lesions, neurological disease and so one. Even over the course of the life history of a single individual, it seems that we have to acknowledge that the same individual can fall under the same mental description at different time in spite of the fact that his brain changes over time. It is consequently at least credible to speak of multiple realization over time within the same individual.57 Furthermore, the cognitive processes of non-human animals are also such as to make it possible that simple mental descriptions such as basics desires, beliefs or emotions occur not only in their different brains, but are localized in their brains differently, according to both their evolutionary and individual trajectory. We must of course include ourselves among these animals that have different trajectories, exemplifying identical mental properties in spite of our neurological differences. All of this is consistent with the account of mental properties and descriptions we developed in the previous chapter.

57

See Horgan (Horgan 1993) for a defence of this claim and Bickle (Bickle 1998: section 4.3 ) for a counter-argument. Whether or not this obtains does not affect our position, since this one is designed to overcome any kind of multiple realization.

184

From psychology to neuroscience

But as importantly, this functionalist account of mental properties also derives support from the ascription of mental descriptions to hypothetical entities completely lacking in the carbon-based neurobiological materials that are common to individuals coming under mental descriptions. As we have argued, what is determinant for an entity to fall under a mental description is not a given intrinsic physical characteristic, but rather the fact of being inserted in an appropriate network of causal relations to environmental stimulations, to other mental property tokens, and to behavioural patterns. These formal traits theoretically include individuals who are radically different from human or other animals from the point of view of their physical composition and organization. It is for that reason that we have noted that mental descriptions are co-extensional with an open disjunction of physical descriptions. The second remark is tied to the fact that in spite of their physical dissimilarities, every entity falling under a mental description M falls necessarily under a given physical description, even if all them, or even most of them, do not fall under the same physical description. This is ensured by ontological reductionism, according to which any mental property token is identical to a configuration of physical property tokens and comes thereby under a certain physical description as well. To take an extreme example, any entity coming under description M satisfies the condition of having a non-null mass. If mental property tokens are a), identical to the configuration of physical property tokens and if b) any physical entity has a mass, it follows that any entity which is psychologically describable must satisfy the description, “has a nonnull mass”. From this broadest of perspectives, it is therefore possible to construct a physical description that is satisfied by all of them. Thirdly, the multiple realization argument does not concern the fact that there is no one applicable physical description making true a certain mental description tout court. The multiple realization argument is rather grounded in the claim that psychology proposes interesting generalizations about the causal relation obtaining between entities that have nothing interesting in common from the point of view of physics (Fodor 1974: 103-04). This general formulation is supposed to describe two constraints with respect to anything counting as an applicable

Epistemological reductionism

185

physical condition on every entity making true a given mental description that might possibly play a meaningful role in a reductive program. The first constraint is that the applicable physical condition has to be specific to entities coming under a mental description M. In this respect, the simple fact of having a non-null mass is not a condition that can be inserted into a bridge law as Bridgepr1, simply because there are a lot of entities that fulfil this condition but that fail to come under the mental description M or any other mental description. Formally, the co-extensionality requirement for reduction imposes that such a physical condition has to be a necessary and sufficient condition for coming under M. The multiple realization thesis claims that such a condition is not available. The second constraint is bound up with our expectation that the applicable physical condition does provide or contribute to a reductive explanation with respect to the question of why all and only the entities satisfying this condition also come under the functionalized mental description M. Suppose, contra multiple realization, that we find that all and only the entities coming under M satisfy two physical conditions C1 and C2 that are apparently causally independent but perfectly correlated. Now, if only C1 provides an explanation for why the entity satisfying both conditions makes true the mental description M in particular, it seems perfectly reasonable to take only C1 into account in order to build a psychophysical bridge law. The upshot of this point is to emphasize that such a condition Ci has to be not only necessary and sufficient, but also explicative to play an adequate role in epistemological reduction. At this point, it should be clear that the multiple realization thesis constitutes an important objection to classical reductionism. First, if it is not possible to construct a physical description that captures a necessary and sufficient condition relative to each possible mental description, then our bi-conditional psychological bridge laws must be thrown out. Second, if bridge laws can’t be constructed, then according to classical reductionism itself (which requires the bridge law supplements) it should not be possible to deduce the laws of psychology on the sole basis of the laws of physics. It is therefore not possible to deduce a law such as Lpsy : ∀x (M1x → M2x) from a law

186

From psychology to neuroscience

such as Lphys : ∀x (P1x → P2x) by means of bi-conditional bridge laws. Since it follows that is not possible to embed psychology in physics, the project of classical reductionism must be called a failure. 5.3.2 What is left out by one-way conditional approaches The objection posed by the multiple realization thesis is to bridge laws in so far as they are supposed to present a bi-conditional form. Accordingly, the conditionals going from mental descriptions to physical description are not available. As we have seen, however, conditional links from complete physical descriptions to mental descriptions are perfectly acceptable and one could argue that they should be sufficient to meet the Nagelian criteria of deducibility of laws. After all, the fact that several laws of physics are available for deduction is not equivalent to the case where there are no such physical laws, which suggests that taken on its own, the multiple realization argument is not a knock-down objection against epistemological reductionism. In fact, some contemporary approaches precisely pointed out that one-way bridging principles are sufficient to meet the Nagelian requirement of deducibility (see for instance Richardson 1979). Suppose we have to following. Lpsy :

∀x (M1x → M2x)

Bridgepr1disj :

∀x (P1ax ˅ P1bx → M1x)

Bridgepr2disj :

∀x (P2ax ˅ P2bx → M2x)

Lphysa :

∀x (P1ax → P2ax)

Lphysb :

∀x (P1bx → P2bx)

In such a case, we argue that Lpsy can be deduced from Lphysa and Lphysb supplemented by Bridgepr1disj and Bridgepr2disj. So the question is to know exactly what is left out by the epistemological reductionism. In order to explain why approaches taking the one-way conditional to be sufficient for epistemological reductionism are not conclusive, let us introduce the concept of “homogeneous explanations” (Sachse 2007: 77). Suppose that the behaviour of Jack and John are explained by reference to property tokens of the same psychological type, for instance the desire to drink a beer. In such a

Epistemological reductionism

187

case, psychology explains homogenously the behaviour of both individuals since it explains it by describing both individual with the same description and by subsuming both cases under the same general causal law. Suppose also that in spite of their psychological similarity, Jack and John are different from the physical point of view and that they fall consequently under different physical descriptions. Here, physics explains how the physical properties of Jack and John bring about their respective behaviour by using different descriptions and by subsuming each case under a specific causal law, thus explaining heterogeneously the behaviours of Jack and John. Our present point is that, although ontological reductionism tells us that any mental entity exemplifying mental properties may be explained by physical causal laws, strategies of epistemological reductionism that privilege one-way conditionals are unable to account for the ability of psychology to provide homogeneous explanations of phenomenon that are heterogeneous from the point of view of physics. More precisely, if we adhere to ontological reductionism, we need to answer the question of the nature of the relationship between different physical entities, with different causal dispositions, grasped by different physical laws and the mental descriptions, laws and explanations that they might come under. One can regard the situation in the following way. Assuming ontological reductionism, there are configurations of physical property tokens that are unified only by the fact of coming under the same mental description, and share no other significant characteristics. One may be silicon-based, while the other may be carbon-based, for instance. So if these entities come under the same mental description, it seems that there is no other justification for the psychological classification than a contingent brute fact, which is obviously not theoretically satisfying. As a result, it appears that approaches based on one-way conditionals fail to explain the source of psychology’s capacity for providing abstract homogeneous explanations of the behaviour of entities that are physically heterogeneous, or, in other words, to provide true homogeneous psychological generalizations. Let us insist on the problem, here: against the background of ontological reductionism, the absence of an explanation of the ability

188

From psychology to neuroscience

of psychology to provide homogeneous explanations of phenomena that are physically heterogeneous is philosophically anomalous. It is perfectly reasonable, given the tendency of the argument we have outlined above, to wonder what does provide the ground for psychological law-like generalizations. Indeed, if two mental property tokens falling under the same mental description are identical to two different configurations of physical properties, then it seems prima facie as if they have nothing in common except the fact that they fall under the same psychological description. Following this argument, we could conclude that psychology fails to grasp any objective feature of reality. It seems that we have cut ourselves off from any reason to be realist with respect to mental property tokens. On this reading, the oneway conditional solution to the problem of the ontological status of the truthmakers of special sciences law-like generalization seems to overtly eliminate psychological property tokens as objective ontological features grasped by psychology. From that point of view, it appears that it is an indispensable feature of any acceptable account of epistemological reductionism to explain the justification for the way psychology abstracts from the physical differences between different physical properties to make them come under the same psychological description, on penalty of disqualifying psychology as a science grasping objective features of reality. This is why bi-conditional links are required for classical reductionism, and why the multiple realization thesis introduced a crisis into that approach by blocking its account of epistemological reductionism. 5.3.3 Multiple realization in the current debate A large part of the current debate with respect to the reducibility of psychology and the special sciences in general is structured by responses to the anti-reductionist consequence of the multiple realization thesis. As a result, the issue has shifted from the question whether or not we can reduce these sciences to the more fundamental level of physics to the question of whether or not the co-extensionality of mental and physical descriptions is even achievable. It is possible to distinguish between two main classes of positions.

Epistemological reductionism

189

On the one hand, some philosophers claim that this coextensionality is not reachable and, therefore, that psychology is not reducible. First of all, the arguments tied to the conceivability of nonactual entities failing to have anything in common with biological individuals continues to ground the irreducibility of psychology. For instance, we can imagine mental properties that are expressed in silicon-based forms of life instead of carbon-based forms of life (Braddon-Mitchell and Jackson 1996: 43), with that result that the epistemological reduction is not a reasonable objective. However, under the pressure of sceptical attack with respect to the multiple realization of mental properties, the recent debate has shifted to examination of actual cases in which mental and physical description are not co-extensional. For instance, Aizawa and Gillet (Azaiwa and Gillet 2009), but also Bregant et al. (2010), argue empirically in favour of multiple realization. Referring explicitly to the work of Bickle on memory consolidation (2003), they point out that at the neurobiological level, different mechanisms explaining memory consolidation can be found in nature. Mice and Aplysia, Drosophila present differences in the amino acid sequences of the regulatory and catalytic subunits that Bickle takes to be the unique neurobiological basis of memory consolidation. Thus, at the neurobiological level, different proteins may be involved in the phenomenon of memory consolidation, which gives rise to stable memory by inducing structural changes in the synapses. It follows that “these different amino acid sequences will contribute different properties and relations toward the realization of higher level properties. Thus, insofar as any given psychological property is realized, in part, by the properties and relations of such proteins, then that property will evidently be multiply realized at the biochemical level” (Azaiwa and Gillet 2009: 200). On the other hand, various strategies have been suggested in the philosophy of science to obtain the co-extensionality required by epistemological reductionism by rejecting the empirical grounds of the multiple realizability thesis. Taking this argument, the further claim is that psychological properties are univocally realized, or that what seem to be different realizers of the same higher level type are in reality realizers of different higher level types. An example of the first line of

190

From psychology to neuroscience

argumentation has been advocated by Bickle, who, as we saw above, takes the process of memory consolidation to be univocally realized by a single class of molecular processes across all known living beings (2003: Chapter 3, especially p. 157). This allows him in the New wave framework to speak of smooth reduction of that psychological phenomenon to a single molecular process. A similar strategy consists in arguing that manifest differences between realizers are not relevant to distinguish different types of realizers. Shagrir, an advocate of this line, claims that “tokens of neural states and processes located in different areas may fall, after all, under the same neural type […]. Thus, psychological states and processes could be realized in different areas of the brain without being multiply realizable” (1998: 448). He bases the credibility of his claim on an analogy with the fact that computational modules can execute the same work at any location. Within the second line of argumentation, some authors try to show that different realizers of what may seem to be instances of the same special science property fall in fact under different and higher level types. Polger (2008) argues that even neural plasticity, the fact that, for instance, a psychological function can migrate through the cortex to a new location following a brain trauma, does not support multiple realization, because there are constraints on relocalization of cognitive functions and, most importantly, because “the preserved function is not maintained at the same level of performance as prior to plastic change”(Polger 2008: 467). In other words, a difference at the psychological level reflects a difference at the neurological level. Therefore, this is not a case of multiple realization of the very same psychological phenomenon. Shagrir (1998) goes exactly in the same direction. In addition to treating the localisation as a non-relevant criterion for the individuation of neurological types, he also argues that “the current consensus among neuroanatomists of brain damaged people is that damaged brains do not really recover the lost function, but generate substitutive operations that have some gross or superficial resemblances to the lost function, [therefore] the function they achieve is different from the function achieved by healthy people”(Shagrir 1998: 448).

Epistemological reductionism

191

A different way of employing the same argumentative line consists in relying on the computational approach of individuation of high level types. Still following Shagrir, if psychological types are conceived as computational types, it can turn out that “two processes that fall under different lower-level types, and whose behavioural antecedents (inputs) and consequences (output) are the same, may still fall under different higher-level types” (1998: 449-50), because both of them achieve the behavioural response by different computational means. In other words, if we individuate higher level types in terms of computational function, then it seems unlikely that we have to acknowledge on empirical grounds that the same higher level description applies to different computational devices. Similarly, Churchland (2007: 34) argues that “genuine identity [between computational types] is too much to ask for”, since no two people display the same configuration of synaptic connections and synaptic weights. Therefore, “no two people on the planet will be computing exactly the same function, for no two people share the same matrix of synaptic connections”. As this brief survey shows, the multiple realization argument is highly contested on empirical grounds. Either it is argued that mental states are univocally realized, at least in our world, or, if there are differences at a lower level, then we are facing two different higher level types. In both cases, it seems that multiple realization evaporates, and, consequently, that the way is open for epistemological reduction of psychology. Shapiro (2000) goes even one step further, by criticizing the very idea of multiple realization on conceptual grounds. According to him, differences in physical composition of realizers imply differences in realization only if one can show that the differences among purported realizations are causally relevant differences for the production of the effect characterizing the higher level type. This leads to the following dilemma:58 “either the realizing kinds truly differ in their causally

58

Shapiro’s dilemma of multiple realization is to be distinguished from the dilemma that will be drawn in the last section of this chapter.

192

From psychology to neuroscience

relevant properties, or they do not. If they do not, then we do not have a legitimate case of multiple realizability, and MRT [multiple realization thesis] in this given instance is false. If the realizing kinds do genuinely differ in their causally relevant properties, then, it seems, they are different kinds. But if they are different kinds, then they are not the same kind, and so we do not have a case in which a single kind has multiple realization” (Shapiro 2000: 647). In other words, there is no theoretical space for multiple realization. For each case in which the multiple realization thesis is claimed, either type-identity holds directly, or type-identity is vindicated by fragmentation of the higher level classification. Such a claim is obviously at odds with the idea that entities with varying physical structures and compositions can instantiate the very same high level functional property. This is a disputable conclusion, and Aizawa & Gillet (2009: 196) have argued explicitly against Shapiro that this account is not consistent with Shapiro’s own interpretation of functional analysis. Moreover, they provide convincing examples of multiple realization in the psychological and biological domains. Last but not least, apart from Gillet & Aizawa, let us emphasize the following point. On the one hand, type-identity seems to be difficult to establish empirically. Attempts such as Bickle’s memory consolidation case and its failure under the press of empirical results reveal such difficulties. On the other hand, any account aiming to establish co-extensionality by fragmentation of the higher level classification faces the problem of explaining the capacity of the special sciences to provide homogeneous explanations of phenomena that are highly heterogeneous from the point of view of physics, a criterion we have pointed to above. Any reductionist account should provide such an explanation on penalty of undermining realism with respect to our mental properties. As these brief examples shows, the inference from multiple realization to the irreducibility of psychology is taken for granted by a non-negligible part of the proponents in both sides of the debate over the reality and the meaning of the multiple realization thesis. Let us now turn in more detail to two reductionist approaches that claim to overcome the multiple realization argument without disputing the

Epistemological reductionism

193

empirical evidence for it, namely the New wave account of intertheoretic reduction and Kim’s strategy of local reductions.

5.4 Overcoming multiple realization 5.4.1 The New Wave account of inter-theoretic reduction The so-called New Wave account of inter-theoretic reduction (hereafter NWR) proposes an alternative strategy of epistemological reductionism, which can be interpreted as a weakened version of the Nagelian classical account. One the one hand, it is strongly related to the Nagelian account though the important role it confers to deduction, although this time around we are not dealing with Nagel’s model of a higher level theory that is deduced from a base theory, but rather with an image theory which is formulated within the vocabulary of the base theory. On the other hand, NWR tries to avoid any commitment to biconditional bridge laws relating theoretical terms of both theories. The main proponents of NWR are Hooker (1981), Churchland (1979; 1981; 1985) and more recently Bickle (1998; 2003). 5.4.1.1 The NWR framework The central claim of the NWR framework is to introduce a third theory T1*, in the process of reducing a high level theory T1 to a base theory, T2 . T1* is constructed as an image of T1 specified within the vocabulary of T2. To quote Churchland, a “successful reduction ideally has the outcome that, under the term mapping effected by the correspondence rules, the central principles of [T1] (those of semantic and systematic importance) are mapped onto general sentences of [T2] that are theorems of [T2]. Call the set of such sentences [T1*]. This set is the image of [T1] in [T2]” (1979: 81). A theory T1 is hence reduced to a base theory T2 as follows (adapted from Churchland 1985: 10): T2

(supplemented by boundary conditions and limiting assumptions)

entails T1* (a set of theorems of [restricted] T2), e.g., (x) (Ax → Bx)

194

From psychology to neuroscience (x) ((Bx ∧ Cx) → Dx) which is relevantly isomorphic with T1

(the older theory)

e.g. (x) (Jx → Kx) (x) ((Kx ∧ Lx) → Mx) Accordingly, reduction can be regarded as a two stage-process. In the first stage, T1* is deduced from a base theory, T2 that has been supplemented by limiting assumptions or counterfactual boundary conditions specifically chosen in order to entail a T1*, which is as close as possible to T1. For instance, in the case of the reduction of thermodynamics to statistical mechanics, one assumes that the molecules of the gas are characterized only by their mechanical energy (Churchland 1985: 9). Through the addition of such external constraints to the base theory, NWR takes up both ends of Schaffner’s proposal, which states that deduction of the high level theory must obey one of two methodological rules: either the higher level theory must be constructed in a corrected version that can be deduced from the true base theory; or the base theory must be modified in order to entail the actual reduced theory (Schaffner 1967). This point is rooted in the fact that a true theory cannot entail a false one. Modus ponens imposes that if the higher level theory is only approximately true, then the premises of the deduction have to be approximately true as well. Such an approximation is obtained by adding external constraints to the base theory. In the second stage of the reductive process, we create an intertheoretic mapping between T1* and T1. In the best case, both theories share a common syntactic structure and it is possible to build a set of ordered pairs, whose function is to indicate, which term in T1* corresponds to which term in the original theory T1. In the second stage, NWR is obviously quite close to the Nagelian model, with the ordered pairs of theoretical terms playing the role of identity statements. In the worst case, no such ordered pairs are available and T1 cannot be reduced in any reasonable sense to T2, with the result that T1 has to be deeply revised or even simply eliminated (Churchland 1986:

Epistemological reductionism

195

284). Inter-theoretic reduction appears therefore within the NWR framework as a mechanism operating on a continuum between smooth reduction and radical elimination, where it ‘picks out’ the true sciences. The location of a particular case of inter-theoretic reduction on this continuous spectrum depends on the strength of the analogy between the syntactic structure of T1* and T1 and on the amount of external constraints that have to be applied to T2 in order for it to imply a theory T1* which is relevantly isomorphic to the original T1 (Bickle 2003: 1819). Proponents of NWR argue that their account of inter-theoretic reduction presents a decisive advantage over its competitors,59 since it does not require inter-theoretic bridge laws and is consequently immune to the multiple realization argument. Given that the formal deduction obtains between two theories that are formulated within the same vocabulary, namely the one of the base theory, there is no need to establish correspondence principles between two different taxonomies across the premise and the conclusion of the deductive process. Correspondence principles appear only in the second stage of the NWR reductive process and thereby those “components are only ordered pairs of terms that indicate the substitutions in [T1*] that yield the actual generalization of [T1], or approximations of the actual generalizations if that is all that a given case permits. […] these ordered pairs imply neither synonymy (sameness of meaning), nor coextension (sameness of reference) of terms, nor material identity.” (Bickle 2003: 18). NWR seems, therefore, to have bypassed the central objection to the classical account of inter-theoretic reduction. 5.4.1.2 NWR and the dispensability of psychology Let us highlight two principal difficulties faced by the NWR account of inter-theoretic reduction. If we temporarily ignore the multiple realization argument, this approach firstly suggests that the higher level theory T1 will be either replaced by its equipotent image T1* or simply eliminated (Endicott 1998: 58). Let us suppose that T1 can be smoothly 59

For a critical discussion of other virtues of NWR, see notably Eck, Jong and Schouten (2006) and Endicott (1998).

196

From psychology to neuroscience

reduced to a base theory T2 in the prescribed manner, using an imagetheory T1* inferred from T2 that is perfectly isomorphic to T1. In such a case, one can construct pairs of theoretical terms that establish substitution rules with the result that the image theory T1* has the resources to perfectly mimic the explanatory role of T1. T1 and T1* have consequently the very same explanatory power. This implies that everything that can be explained by T1 can be explained by T1* as well, which seems to make T1 dispensable. This dispensability of T1 becomes even more clear once we observe that T1* is a set of theorem inferred from T2, which is, as a lower-level theory, more complete, with a larger domain of application. There is therefore a significant asymmetry between T1 and T1*. If everything that can be explained by T1 can be explained by T1*, it ought to be explained on the basis of T2 as well. Moreover, T2 can explain things that are not explained by T1. This is what leads Churchland to claim that “the intra-theoretic deduction […], and the inter-theoretic mapping, constitute a fell-swoop demonstration that the older theory can be displaced wholesale by the new without significant explanatory or predictive loss” (Churchland 1985: 11). From that point of view, a case of ideally smooth reduction does not really entail retaining the old theory, but rather replacing it by its image, since it turns T1 into a sub-theory of T2 that simply uses a distinct vocabulary. On the other hand, in any other case than the one of perfectly smooth reduction, the NWR approach suggests that the image-theory should be preferred to the original theory, since, in such a case, T1*, being constructed within a more complete and general theory, optimizes the explanatory power of the original T1. Hence, it seems that, contrary to the original claim of its proponents, NWR implies theory replacement not only at the bumpy end of the continuous spectrum of inter-theoretic reduction, but on both extremities of this continuum (Endicott 1998: 58). So much is the case if we suspend belief in the multiple realization thesis. Let us now consider the consequences on the NWR framework if we do accept the multiple realization thesis, which implies that entities coming under different physical types can nonetheless come under the very same mental type. Hence, NWR

Epistemological reductionism

197

would have to recognize that there are possibly several theories T2-1, …, T2-n, from which an image T1*i of T1 can be inferred, with that result that there are several image-theories T1*1, …, T1*n of T1, each of which may be formulated within the vocabulary of a different base theory, all of them simulating the explanatory role of T1. From that point of view, NWR tends to construct local and kind-specific reductions, just as the Kim’s model of functional reduction does (5.4.2.2 Multiple realization and functional reduction). Bickle goes straightforward in this direction, acknowledging that usually scientific reductions are precisely kindspecific and arguing that this does not per se block the way to reduction (Bickle 2003: 22). However, our previous argument applies here to each of the image theories T1*1, …, T1*n of T1. Again, we face a choice: either a theory T1*i is ideally isomorphic to T1, in which case T1*i has the same explanatory power as T1 while being embedded in a more general theoretical framework, which suggests that the image should be preferred to the original theory, or T1*i is not ideally isomorphic to T1 and the fact that the former is embedded in a more general and complete theory suggests that the explanations provided by T1*i are more optimal than explanations provided by T1. In either case, the original theory is replaced by the image theory T1*i . 5.4.1.3 NWR and eliminative materialism The situation becomes concretely difficult for NWR when it comes to the question of the ontological status of the truthmakers of psychological law-like generalizations. We will proceed as above, first suspending multiple realization, and then allowing it. In the case of ideally smooth reduction considered while suspending the multiple realization thesis, pairs of theoretical terms play the role of identity statements. From the ontological point of view, it means that mental property tokens are type-identical to appropriate configurations of physical property tokens. This is the only instance in which NWR can consistently retain the ontology of the original theory T1, i.e. mental property tokens do exist, and they are tokens and type-identical to configurations of physical property tokens. However, since, as we have shown above, smooth reduction leads to the elimination of psychology in favour of the more general and more complete character of the base

198

From psychology to neuroscience

theory T2, ontologically, the referents of psychology must be seen as nothing but a sub-set of the entities that are explained by T2. Now, since T1 cannot provide any explanation that is not already provided on the basis of T2, it seems that we may arbitrarily take property tokens of T1 to be type-identical to configurations of configuration of property tokens of T2 or take the descriptions of T1 to refer to tokens of T2. In case of extremely bumpy reduction, where there are no pairs of theoretical terms that are available to play the role of identity statement, the ontological situation is even more skewed to the base theory. Since the base theory is more complete, and as such provides better explanations, it is reasonable to assume the ontology of the truthmakers for the higher level theory that is being replaced by the base theory should also be replaced by the ontology of the base theory. Thus, even putting aside multiple realization, NWR suggests in the best case in the dispensability and, in the worst case, the elimination of the ontology of psychology. Consider what happens if the multiple realization thesis is claimed. In this case, it is even more difficult to see how the NWR reduction could preserve the ontology of folk psychology. As we explained above, the multiple realization argument suggests that there are possibly different base theories T2-1, …, T2-n to which psychology can be reduced. This means that there are different image-theories that can mimic the explicative power of the original theory while being embedded in better theories. This obviously suggests the dispensability of the original theory. Take the case ensuing when each theory T1*i is not ideally isomorphic to the original T1. Then, the base theory provides more powerful explanations for the phenomenon to be explained. This obviously infers that each T1*i and each T2-i will give us a better grasp of the ontological make up of the world. On the other hand, even in case of perfect isomorphism between each T1*i and the original T1, it remains the problem that the entities making true the lawlike generalization of T1 are not systematically identical from the point of view of the lower-level theories. Against the background of ontological reductionism, this fact suggests that the law-like generalizations in question fail to grasp objective similarities existing in the world, leading again to the conclusion that T1 is mistaken when it

Epistemological reductionism

199

posits commonalities between certain entities, since a more close examination reveal that these commonalities fail to exist. NWR in consequence suggests that folk psychology could theoretically be replaced (and its ontology be eliminated) by neuroscience and other bases-theories. In logically tending towards this position, NWR collapses into eliminative materialism, which is the thesis according to which “our common-sense psychological framework will not enjoy an inter-theoretic reduction, because our common-sense psychological framework is a false and radically misleading conception of the causes of human behaviour and the nature of cognitive activity. On this view, folk psychology is not just an incomplete representation of our inner natures; it is an outright misrepresentation of our internal states and activities. Consequently, we cannot expect a truly adequate neuroscientific account of our inner lives to provide theoretical categories that match up nicely with the categories of our common-sense framework. Accordingly, we must expect that the older framework will simply be eliminated, rather than be reduced, by a matured neuroscience” (Churchland 1981: 67). There are two important points to be noted here. First, Churchland argues that the taxonomy of ordinary psychology will be eliminated and replaced by a future neuroscientific taxonomy. In making this move, he implies that the class of creatures exemplifying mental properties is restricted to those exemplifying neurobiological properties as well. In fact, he explicitly endorses this claim in his Neurophilosophy at work, claiming that what unifies the class of cognitive creatures is not the fact that they are capable of fulfilling similar causal roles or computational functions, but rather the fact that all of them fulfil some causal roles or computational functions by means of configurations of neurobiological property tokens (Churchland 2007: Chapter 2). This claim obviously rejects the multiple realizability of mental properties, according to which it is logically possible that non-biological entities could exemplify mental properties. Churchland’s neurobiological-centrism clashes with the unchauvinistic functionalism that we have been positing here (4.2.1.3 Common-sense functionalism).

200

From psychology to neuroscience

Secondly, let us consider eliminative materialism with regards to the problem of mental causation. According to eliminative materialism, it is not the case that there are mental property tokens out there in the world as posited by ordinary psychology: such things as ordinary beliefs or desires do not exist in the world. Consequently, there can be no ordinary mental causation; my beliefs and my desires, supplemented by accessory conditions, do not bring about my behaviour. Contrary to ontological reductionism, eliminative materialism not only repudiates the premise of the problem of mental causation, according to which mental and physical property tokens are distinct. By asserting that there are no such mental property tokens, but it also rejects the premise of mental causation. As we have previously shown, however, this is an unacceptable result. NWR, by flirting with eliminative materialism, essentially solves the problem of mental causation by giving up the mental. Moreover, let us emphasize one of the chief reasons that NWR collapses into eliminativism: it fails to provide an account of the ability of psychology to abstract from lower-level differences. We can see this clearly when we do posit the multiple realization thesis and ask about the ontological status of the truthmakers of the psychological law-like generalization. Against the background of ontological reductionism, the absence of well established pairs of theoretical terms suggests that there is no physical counterpart to the similarities upon which psychology focuses. Now, if we have no explanation of how psychology comes to group together and explain homogeneously entities in the world that physics can only assert heterogeneously, then it seems reasonable to assume that we should simply favour our best scientific theories and their respective ontologies, even at the price of the premise of mental causation. After all, our experience of mental causation could simply be mistaken and we could learn to live without ordinary mental property tokens. At this stage, it seems clear that NWR collapses into eliminative materialism because it sacrifices the ontology of the psychology in the face of the relatively complete character of neuroscience and other more fundamental theories. However, as we shall argue in the next chapter, we don’t need to make such a hefty sacrifice if we can provide an account of the ability of psychology to

Epistemological reductionism

201

explain homogeneously the behaviour of physically different entities. The first step in the direction of such an account is Kim’s account of functional reduction. 5.4.2 Kim’s account of functional reduction 5.4.2.1 Functional reduction Both the classical and the NWR accounts of inter-theoretic reduction are independent of the precise way in which the theoretical terms of the theories taken into account are defined. In contrast, Kim’s account of functional reduction requires a functional approach of psychology. Kim’s purpose is to transform the general idea of reductive explanation into an account of epistemological reduction. The three steps of this reductive program are, a) to provide a functional definition of the property type to be reduced; b) to discover what physical entities should be accepted under this property type; and c) to develop a reductive explanation to account for why entities coming under a certain lower level type comes under the higher level property type as well (Kim 2005: 101-02). Let us review these elements one by one. The first step consists in the functionalization of the mental descriptions to be reduced. Common-sense functionalism provides the theoretical resources for such a functionalization, although Kim’s himself remains relatively neutral by confining himself to the general notion of the causal role alone. Accordingly, for an entity, to come under a mental description means nothing more than that it fulfils the typical causal role specified by the functional definition of the psychological property type. In other words, any placeholder inserted in a causal network in such a way that it makes true a certain functional description is a mental property token of that type. For instance, if a mental property of type M is defined as a property fulfilling the causal role C, then any entity fulfilling the causal role C is a property token of type M. The second step of functional reduction consists in discovering what physical entities fulfil the causal role in question. Against the background of ontological reductionism, to discover such physical realizers means nothing but to provide descriptions of the entities

202

From psychology to neuroscience

satisfying the causal role C that are formulated within the vocabulary of a lower level theory, ultimately physics. The possibility of providing a physical description of an entity making true a mental description is secured by supervenience and ontological reductionism. If we assume the supervenience of mental property tokens on configurations of physical property tokens, then any entity making true a mental description falls under a certain physical description as well, since the physical properties of this entity determine its mental properties, which in turn make true the description in question. This availability is even clearer against the background of token-identity. After all, if any causally efficacious mental property token is identical to a configuration of physical property tokens, then any such token is describable in physical terms. The third step of Kim’s strategy of functional reduction consists in explaining how the configurations of physical property tokens described at the second step perform the task C, whose execution qualifies them as realizers of the mental property type M. This explanation accounts for why these configurations of physical property tokens are mental property tokens of type M. The theoretical availability of such an explanation depends upon ontological reductionism and the causal, nomological and explicative completeness of physics. Let be a mental property token m1 be identical to a configuration of physical property tokens p1, and assume as well that it fulfils the causal role C by causing a complex configuration of property tokens p2, which come under the physical description P2. Against the background of the causal completeness of physics, there is a law-like generalization according to which any configuration of physical property tokens coming under the description P1 causes a configuration of physical property tokens coming under P2 such that p1 explains why p2 occurred. Such a physical explanation is evidently more detailed, given the current advanced state of molecular neurobiology, than the psychological explanation linking m1 and the occurrence of p2. The psychological explanation treats the mental property m1 as a whole, whose essential feature is to fulfil the causal role C and it is in virtue of this essential feature that m1 is a token of the functionalized mental

Epistemological reductionism

203

property type M1. In contrast, the physical explanation, using the constructed description P1 of p1, does not treat p1 as whole but rather as a configuration of simpler physical property tokens, i.e. constitutive components, organized in such a way that they fulfil the causal task C.60 This kind of physical explanation takes into account the components, the activities and the organization of the configuration of physical property tokens coming under description P1 to account for the fact that the causal task C is achieved by p1. In addition, such a physicalist account is more complete than the psychological explanation of p2, since it is embedded in a theoretical framework that is more complete than psychology. At this point, Kim’s account of functional description is sufficient to explain reductively how any configuration of type P1 fulfils the causal task that essentially defines a certain mental property type M1; in so doing, it explains why entities coming under the physical description P1 come under the mental description M as well. The payoff to Kim’s model of functional reduction is that it offers an account of epistemological reductionism on the basis of the idea of a reductive explanation. Let us emphasize the following points with respect to this strategy of functional reduction. First of all, let us make clear in which sense it provides an account of inter-theoretic reduction. After all, in their simplest form, reductive explanations concern individual entities. The lower-level properties of this entity explain why and how it fulfils the causal role, which is characteristic of a certain type of higher-level properties. By contrast, epistemological reductionism aims to establish systematic relations of derivation between the laws of different theories. As argued with respect to the Nagelian account (5.2.2.2 Requirements for bi-conditional bridge laws), this systematicity is spelled out by discussing the relations obtaining between the property types of each theory. The core idea of Kim’s strategy is to provide an account of epistemological reductionism by using the bottom up 60

Without entering explicitly into the debate about mechanisms, Kim has suggested that the description provided at the second step of his account might describe the mechanism by mean of which the causal task C is achieved (Kim 2005: 102).

204

From psychology to neuroscience

conditionals provided by reductive explanations in order to connect the types of different theories with the force of nomological necessity. After all, as long as laws of physics are held to be complete and constant in our world, all entities coming under the same physical description will have the same causal dispositions and will fulfil the same causal role, from which it ensues that it is nomologically necessary that they come under the same psychological description. Kim’s account of epistemological reductionism supposes therefore bottom up conditionals. In addition to bottom up conditionals, Kim’s account of functional reduction also requires top down conditionals linking mental property types to physical property types in order to achieve epistemological reductionism (Esfeld and Sachse 2007; Marras 2002; Marras 2005; Sachse 2007). After all, epistemological reductionism requires the deducibility of laws and we have seen that this requires the co-extensionality of the theoretical terms proper to each theory. In the absence of top down conditionals, it is not possible to deduce a given psychological law-like generalization on the basis of a particular physical law (5.3.2 What is left out by one-way conditional approaches). In this respect, Kim’s account of functional reduction does not differ essentially from the Nagelian classical account, since it also requires bi-conditional bridge laws as well.61 Secondly, in spite of sharing this feature, Kim’s model of functional reduction enjoys an important advantage over the Nagelian account because it does provide implicitly reductive explanations of higher level theories. By starting with the functionalization of the higher level description, it can explain, against the background of the laws of physics, how physical causal processes, obtaining between components of the entity coming under the functional description, bring about the psychological effect in question.

61

Morris (2009), focusing on the notion of realization, argues that functional reduction does not require bridge laws. However, his position does not distinguish between the notion of a reductive explanation and functional reduction, which is supposed to be an account of epistemological reduction.

Epistemological reductionism

205

In contrast, Nagelian bridge laws do not provide any comparable reductive form of explanation. In spite of being law-like and thereby nomologically necessary, they are not explanatory in their own right. The classical account of reduction explains higher level phenomena by means of the bottom up deduction of explanatory concepts. That is to say, once one has constructed a correlation between a given mental concept and a physical one, it becomes possible to fully deduce the explanatory psychological description from the physical description of the phenomenon. The problem is that this kind of deduction fails in its essential task, which is to account for how a particular configuration of physical property tokens brings about a certain psychological effect. It doesn’t give us a satisfactory explanation of why this configuration is identical to a mental property token of a certain type. Note that, for reasons parallel to those explained in the section above, this holds for the pairs of theoretical terms within the NWR framework as well. Assume that the multiple realization thesis is false. Then let’s assume it is possible to construct within physics an image theory T1* that is perfectly isomorphic to psychology. Even so, NWR does not provide by itself a genuine explanation for how certain physical configuration of physical property tokens bring about their psychological effect, or why it is a token of that precise mental type. Just as in the classical account of inter-theoretic reductionism, the fact that the NWR account does not explicitly appeal to the functionalization of higher level types prevents it from formulating genuine reductive explanations. 5.4.2.2 Multiple realization and functional reduction Let us now consider the consequence of the multiple realization argument for Kim’s account of functional reduction. The multiple realization argument makes it impossible to establish bi-conditional bridge laws. Accordingly, it is not the case that all descriptions of psychology are co-extensional with constructible physical descriptions. The multiple realization thesis is so discrediting to the Nagelian account of inter-theoretical reduction because the latter works essentially by establishing bi-conditional principles between property types. Within Kim’s account, it is however possible to apply functional

206

From psychology to neuroscience

reduction to entities making true both a mental and physical description by taking them individually. That is, for each property token of psychology, it is in principle possible to develop a reductive explanation proper to this token. Assume that the essential feature of mental property tokens of type M is to fulfil the causal role C. The mental type M is thereby functionally defined and applies to any entity fulfilling C. Now, for each particular entity e coming under the description M, ontological reductionism ensures that there will be a physical description P such that it will provide a detailed and complete explanation of how e brings about the effect resulting from the execution of the causal role C, within the framework of the laws of physics. The fact that it is possible in theory to provide a reductive explanation of any mental property token is independent of the multiple realization argument. It does not matter here whether all property tokens of type M come under the same physical descriptions or not. If some property tokens of type M comes under the physical description P1 while other comes under the physical description P2, it is still possible to explain reductively each property token of type M either by P1 or by P2. The crucial point is here that one-way bottom up conditionals are sufficient to provide reductive explanations of any particular mental property tokens, since with them we can explain how the causal role C of each mental property token of type M is achieved solely on the basis of the physical type of the configuration of the physical property token that it is identical to. However, as already pointed out, although they are sufficient for reductive explanations, one-way conditionals are not sufficient for epistemological reductionism. Given multiple realization, it is not the case that all property tokens of a certain mental type are always reductively explained by the same physical account, from which it ensues that that there is no unique homogeneous reductive explanation of all tokens of the same mental type. This raises the following problem. Given that we now have, ex hypothesis no relevant physical commonality between the different types of configurations of physical property tokens that may come under the same mental description, then there is no explanation of the capacity of psychology to offer these

Epistemological reductionism

207

homogeneous explanations, which undermines the legitimacy of psychology as a theory grasping objective similarities in the world, assuming ontological reductionism. In order to overcome the multiple realization argument, Kim argues in favour of so-called “local” or “species-specific” reductions (1998: 93-95; 2005: 25-26). Accordingly, given any mental description M, it is possible to construct an intermediate description Mx that is coextensive with the physical description Px within a particular species S. These intermediate descriptions are individuated in a hybrid way to represent the causal role of the original mental description along with an additional specification representing the species or kind of entities to which they will exclusively apply. As a result, Mx is co-extensional with the physical description Px of the entities coming under Mx. For instance, we can construct a species-specific type Mh of the type psychological M, “feeling pain”, which is co-extensional with a neurobiological description Ph and that applies only to humans. And we can also construct another species-specific type of pain event for a Martian, Mm, that will be co-extensional with a physical description Pm proper to the Martian feeling of pain, and so on for any other possible kind of entity fulfilling the causal role of pain. The upshot of this move is to provide bi-conditional bridging principles that are specific to certain kind of physical entities in such a way that they allow us to formulate local but nonetheless homogeneous reductive explanation specific to these kinds of physical entities. Whether or not this move is conclusive is a disputed issue, since it seems to depend on “how much” there is multiple realization in our world, which is a highly contingent fact. In the best-case scenario, there is no intra-specie multiple realization and homogeneous reductive explanations can be given for each mental property relative to its instantiating species. In the worse case scenario, there is intra-specie multiple realization or even intra-individual multiple realization distributed over time, with the result that the local reduction could turn out to concern only the relation between mental and physical descriptions applicable to one single entity at a particular time. From this it ensues that there is contingently no more than one single entity that comes at the same time under a certain physical description P1 and

208

From psychology to neuroscience

a certain mental description M. In this case, Kim’s model requires us to introduce an intermediate description Mx operating for each particular entity. On this account, there can be no homogeneous reductive explanation of different entities coming under the same mental type. The general problem here is that none of those scenarios seems to be very satisfactory. In the second scenario presented above, there is no homogenous explanation at all, and here the strategy of local reduction collapses into the very simple idea of a reductive explanation on a caseby-case basis. As we have already argued, what is missing in this kind of account is an explanation of how psychology can build homogeneous explanation of entities that fail to share any relevant physical feature. This situation suggests that psychology fails to grasp any objective similarity obtaining in the world, putting into doubt its scientific legitimacy. Unfortunately, the situation is not any better in our first scenario. Although co-extensionality between physical types and hybrid sub-types is reached, it is only those sub-types that are reduced rather than the original types. But the introduction in the individuation of these sub-types of some lower-level specifications does not advance our goal of explaining the ability of psychology to abstract from the physical difference distinguishing the entities coming possibly under the same mental type. The central problem we set out to answer remains untouched. There are no homogeneous lower-level level explanations that allow us to put psychology on a scientific basis for explaining the phenomenon of general mental description that occur in spite of physical differences. Against the background of the tokenidentity thesis, the legitimacy of psychological taxonomies seems either intrinsically mysterious or illegitimate. From this point of view, Kim’s introduction of hybrid sub-types fails to bridge the gap opened by the multiple realization thesis, which is to explain how psychology comes to group physically different entities under the same psychological type. Thus, Kim’s account of functional reduction fails to achieve epistemological reductionism because it simply displaces the initial problem of multiple realization from the relation between the original psychological types and physical types to the one of the relation between these original psychological types and the introduced hybrid species-specific types.

Epistemological reductionism

209

Why is this failure important? Psychology is supposed to grasp objective commonalities between numerically distinct mental property tokens, which come thereby under the same psychological description. The supervenience argument for token-identity implies that these tokens are individually identical to certain configurations of physical property token. However, the multiple realization thesis has shown that it is not necessarily the case that every psychological token of a certain type is identical to a configuration of physical property token of the same specific type. This being the situation, the problem is how psychology can possibly showcase commonalities that are invisible from the point of view of physics. The absence of such an account raises the spectre of the elimination of psychology. After all, ontological reductionism implies that the causal relations in which any causally efficacious mental property token stands are completely explained by the laws of physics. In spite of its pragmatic utility, psychology seems therefore highly dispensable, because it can simply be eliminated by reduction to the lower level theory of physics, which provides a more complete explanation of seemingly mental phenomena and thus seems to grasp in a better way the ontological make up of a world in which these mental phenomena are said to occur. The lesson of multiple realization, according to which there is no one physical counterpart to psychological commonalities, suggests, on this account, that these commonalities in actuality fail to have more than a nominal existence, from which it ensues that there are no mental property tokens out there in the world. Psychological law-like generalizations are false in their own terms, by lack of a real referent. In order to block the eliminativist move, one has to argue in favour of the indispensability of psychology, emphasizing that it is only through psychological descriptions that we can see ontological commonalities between entities that are irreducibly different from the point of view of physics. After all, our bottom up point of view makes it impossible to showcase the commonalities that psychology is supposed to grasp, because physics cannot bracket the differences distinguishing the entities coming under the now eliminated psychological descriptions. If we accept this as the better explanation of the world, we have eliminated, as well, the basis for the multiple

210

From psychology to neuroscience

realization thesis, since it would mean that we can make do entirely with physical commonalities shared by entities grouped together under the now discredited same mental description. However, we pay too heavy a price for this picture of the world. Psychology does, after all, seem able to grasp common feature of entities by abstracting from the physical details distinguishing them. Outside of the eliminativist move, the problem is to know exactly what is meant by this abstraction from physical details. In the next chapter, we will give our account of how psychology can provide homogeneous explanations of entities that are only heterogeneously explained by physics. Thus, it is obviously a major problem with Kim’s account that it leads to the elimination of psychology. In fact, Kim’s introduction of hybrid sub-types, which is motivated by the desire to avoid the problems of classical reductionism and NWR, in actuality simply displaces the problem of the non-co-extensionality of physical types and mental types to the one of the non-co-extensionality of these same original mental types and their hybrid sub-types, which have been constructed as the basis for local reductions. Given that physics cannot be used to justify the way in which psychology classifies the entities to which it refers in our world, the introduction of hybrid sub-types does not address the issue, but simply refracts it. We still want to know what different entities coming under the same psychological type but under different sub-types have in common that would justify their grouping. A simple answer would obviously be that they share the same causal role. This is a trivially true, being the equivalent of saying that this is the reason they come under the same psychological type. However, as we shall argue in the next chapter (6.2.1 Physical differences and causal differences), physical differences imply causal differences as well, and that thus, entities coming under different sub-types will fulfil different causal roles as well. Therefore, the question raised by the multiple realization re-emerges from the causal point of view as well: how is it possible for psychology to see causal identities where physics only sees causal differences? Kim’s account gives us no direction

Epistemological reductionism

211

towards any kind of answer, thus collapsing into the elimination of psychology.62 To sum up our reconstruction of Kim’s account of functional reduction, let us highlight the fact that Kim’s account is motivated by the project of turning reductive explanations into an account of epistemological reductionism. Against the background of tokenidentity, which is secured the Kim’s own supervenience argument, it is possible here to provide a reductive explanation of any causally efficacious mental property token. However, the introduction of intermediate sub-types that are individuated both on a functional and a physical basis is not sufficient to vindicate epistemological reductionism. By failing to provide a genuine explanation of psychology to provide homogeneous explanations of phenomenon that are heterogeneous from the point of view of physics, this account suggests the elimination of psychology due to the combined strength of two arguments. On the one hand, we can completely explain any causally efficacious entity by reference to physics. Psychology then seems to be dispensable as a matter of theoretical economy. On the other hand, against the background of ontological reductionism, the failure to construct physical descriptions that are co-extensional with mental descriptions and the lack of any other account of the ability of psychology to abstract from physical differences distinguishing the entities coming under the same psychological description undermines the legitimacy of psychology as a science that can grasp objective similarities. This explains why Kim’s account is driven to the elimination of psychological descriptions, explanations and tokens.

62

Kim recognizes this consequence of his account: “Unless two realizers of E show significant causal/nomological diversity, there is no clear reason why we should count them as two, not one. It follows then that multiply realizable properties are ipso facto causally and nomologically heterogeneous. This is especially obvious when one reflects on the causal inheritance principle. All this points to the inescapable conclusion that E, because of its causal/nomic heterogeneity, is unfit to figure in laws, and is thereby disqualified as a useful scientific property. … The conclusion, therefore, has to be this: as a significant scientific property, E has been reduced – eliminatively” (Kim 1999: 17-18).

212

From psychology to neuroscience

5.5 The general dilemma of multiple realization To illustrate the general problem raised by multiple realization in the context of epistemological reductionism, let us briefly summarize the consequences of the argument that has run through the varieties of epistemological reductionism we have been considering. First of all, the classical account of epistemological reduction radically fails in the face of the multiple realization thesis, since the bi-conditional bridge laws that we need to deduce the laws of psychology are logically undermined. More precisely, given the unavailability of bi-conditional bridge laws, the Nagelian account is unable to deduce the generality of psychological law-like generalizations. Thus, multiple realization leads to the collapse of the classical account of inter-theoretic reduction. On its ruins arose the NWR account. NWR, postulating an image-theory that is relevantly isomorphic to psychology, a higher level theory, claims to reduce it to the lower level theory, physics, in as much as the image-theory is completely explicated through physics terminology. In this case, it is possible to obtain pairs of theoretical terms between both theories that establish systematic links between psychology and physics. However, again the effect of the multiple realization thesis is ruinous to this neat arrangement, for to get around multiple realization, NWR is forced to construct different imagetheories, each of them being more or less isomorphic to psychology. Even if we assume that there will be several image-theories that are relevantly isomorphic to psychology, the NWR account does not provide any explanation of the ability of psychology to provide homogeneous explanation of phenomenon that are heterogeneous from the point of view of physics. The inevitable result is that that collapses into the elimination of psychology. Finally, Kim’s account, which sets out to avoid the fate of the two previous forms of reduction, falls victim to the same logical fate. Although the functional account of epistemological reductionism allows implicitly for reductive explanations of psychological phenomena at the token level, the strategy of local reduction just displaces the problem of multiple realization from the relation between mental and physical descriptions to the one between psychological

Epistemological reductionism

213

types and hybrid sub-types that are supposed to enable local but nonetheless homogeneous reductions. As a result the same problem occurs again. Kim’s account also fails to provide any explanation of the general ability of psychology to provide homogeneous explanation of phenomena that are heterogeneous from the point of view of physics. The outcome is again an eliminativist one. In each case, the inability of these strategies of epistemological reductionism to account for the legitimacy of psychology to provide homogeneous explanations of phenomena that are heterogeneously explained by physics strongly suggests the elimination of psychology. The problem comes from the ontological status of the patterns of regularities that are supposed to make true psychological law-like generalization. Ontological reductionism implies that the same entities in the world make true both psychological law-like generalizations and certain physical law-like generalizations. Thereby, ontological reductionism ensures that any causal relation obtaining between two mental property tokens can be grasped by some or other physical law. However, the multiple realization thesis implies that the set of entities making true the psychological law-like generalizations contains entities that are physically different. Provided that physical differences imply causal differences (6.2.1 Physical differences and causal differences), this implies that what psychology may take for identical causal relations appear to be different from the point of physics. The question is therefore to know how psychology is allowed to see causal identities whereas physics see causal differences. In the absence of such an account, it seems reasonable to assume that the patterns of regularities grasped by psychology either do not exist, or there is something not physical about them. Let us call this the dilemma of multiple realization. On the one hand, the multiple realization thesis, when conjoined with ontological reductionism, suggests that causal identities that are supposed to make true psychological law-like generalizations do not exist. The endorsement of a realist view of psychology seems therefore to be deeply compromised. If psychological law-like generalization fails to have truthmakers, then it is perfectly well grounded to say that there are no mental property tokens out there in the world, at least not in the

214

From psychology to neuroscience

sense of property tokens that fall under the same mental description in virtue of being perfectly similar. This is why, against the background of ontological reductionism, the multiple realization thesis suggests an eliminativist attitude with respect to psychology, favouring its replacement by theories capturing genuinely identical causal relations in the word. On the other hand, if one wants to vindicate the scientific legitimacy of psychology as grasping genuine identical causal relations in spite of the fact that they fail to have a counterpart at the physical level, it seems that the only option left is to consider that the causal relations making true psychology are not identical to physical causal relations. This obviously amounts to rejecting ontological reductionism and plunging us into all the notorious problems of property dualisms, which we surveyed in Chapter 2. In sum, the argument from multiple realization drives the debate with respect to the relation between mental and physical properties either to the elimination of psychology or to property dualism.

5.6 Summary and transition Given that the main results of this chapter have already been summarized in the last section, let us here be relatively brief and add the following comments with respect to the state of the art of the debate about epistemological reductionism. Fundamentally, through the establishment of bi-conditional bridge laws aiming to enable the deduction of psychological laws on the basis of more fundamental laws, epistemological reductionism is driven by the isolation of some physical invariant, which is shared by every entities falling under the same mental description and specific to them. This would ground the legitimacy of psychological categories within physics. To put it in a nutshell, two entities falling under the same mental description should share a physical necessary and sufficient condition, which would in turn explain why they are identically classified within the psychological point of view. However, the core idea of the multiple realization thesis is to show that such a condition does not exist, since it is conceivable that

Epistemological reductionism

215

entities lacking any relevant physical commonalities can none the less fall under the same higher level type. But how is it possible that psychology takes entities to be causally identical, explaining them homogeneously, whereas physics takes the same entities to be causally different, explaining them heterogeneously? The reductionisms we have examined run aground on this question. This requirement of causal identity is the key factor that grounds the dilemma of multiple realization. On the one hand, one can argue that the lack of physical commonalities suggests that the identical causal relations that psychology is supposed to grasp by mean of its law-like generalizations simply fail to exist. On the other hand, one can argue these identical causal relations do exist, but they are simply not physical. This amounts obviously to giving up ontological reductionism. As we saw, several approaches address the multiple realization argument and its anti-reductionist consequence. A first line of approach tries to negate the premise of multiple realization either by trying to vindicate type-identity or by fragmentation of the psychological taxonomy. However, empirical results resists suggests that typeidentity is out of range and the fragmentation of the psychology taxonomy does not solve the problem of knowing how it is possible for psychology to provide homogeneous explanations of entities that are heterogeneous from the physical point of view. A second line of approach tries to overcome the multiple realization argument. Unfortunately, it turns out that both the New wave reductionism and the Kim’s strategy of functional reduction collapse into the elimination of mental property tokens, just as it is the case with any strategy trying to overcome the multiple realization argument by fragmentation of psychological types. At this stage, we are in the face of a fundamental desideratum, the solution to which gives us a standard from which to judge whether epistemological reductionism has succeeded. In other words, we need a coherent account for the ability of psychology to provide homogeneous explanation of entities that physically heterogeneous. To draw such an account is the larger target of the next chapter. As we shall see, such an

216

From psychology to neuroscience

explanation should start by losing the idea that psychology grasps causal relations that are identical in its homogeneous explanations.

Chapter 6 REDUCTION BY MEANS OF FUNCTIONAL SUB-TYPES Our purpose in this chapter is to design a strategy for epistemological reduction that will satisfy the constraints that we have explored in the previous chapter. The central claim of our account is that physical differences implies functional differences as well, which opens up the possibility of functionally defined sub-types of psychological types that will be co-extensional to constructed physical types. This claim is grounded in the observation that entities which are homogeneously classified by psychology, but are physically heterogeneous, must be ipso facto causally heterogeneous. This causal heterogeneity entails specific differences in the way that physically different entities bring about the characteristic effect of a certain mental type. We shall in turn argue that these specific differences can be made salient enough to be grasped by psychology given appropriate environmental circumstances. As a result, functional descriptions can be fine grained enough to construct a fully functionally defined sub-type for each physical type of entity coming under a certain psychological type. Thereby, coextensionality between psychological sub-types and physical constructed type is in principle achievable. On this basis, one can firstly reduce psychological sub-types to constructed physical types by means of homogeneous reductive explanations. Secondly, any sub-type of a given mental type is individuated by the functional specification of the abstract mental type in question, plus some additional contextual functional specifications. Therefore, if one brackets these additional functional specifications, one is able to retrieve the original mental type, which means that multiple realization is turned into an intra-theoretical issue relative to

218

From psychology to neuroscience

the degree of specificity of our psychological definitions, rather than an inter-theoretical problematic issue. We can then explain in a conservative manner the ability of psychology to abstract individuals coming under the same psychological description from the differences that are picked out by the lower-level theory. Moreover, it is only by recourse to psychological description that we can bring into relief the features that link a relatively large set of physical entities, vindicating the scientific indispensability of psychology in as much as this higher level theory is necessary to grasp objective patterns of similarities in the world.

6.1 What should be expected from any account of epistemological reductionism? Before discussing in detail the reductive strategy we are pursuing here, let us briefly recap what is required to thread the needle, so to speak, by a reductionism that at the same time avoids the elimination of the reduced theory, psychology. First of all, the laws of psychology have to be deducible from the laws of some lower-level theories, ultimately physics (5.2.2.1 Deductibility of laws). As we have seen in the last chapter, this has been formed the major theme of Classical reductionism, which has tried to create a framework for making intertheoretic deduction possible by finding bi-conditional bridging principles by which to reduce the higher level laws (5.2.2.2 Requirements for bi-conditional bridge laws). The New Wave account of inter-theoretic reduction focuses the deducibility of intermediate image-theories, which can be more or less isomorphic to psychology (5.4.1.1 The NWR framework). This move is motivated by the problems with Nagelian reduction associated with the multiple realization thesis. Finally, Kim’s account of functional reduction emphasizes the causal role which is thrust upon the entities in our explanations. Thus, an appropriate physical description of an entity e coming under the mental description M, supplemented by laws of physics must necessarily imply that e fulfils the causal role R defining M. Thereby, Kim’s account aims to construct a model of epistemological reduction on the basis of the notion of reductive explanation, by introducing hybrid sub-types, which are co-extensional

Reduction by means of functional sub-types

219

with constructed physical types. As argued, these co-extensional subtypes are logically reduced to physical types by mean of bi-conditional bridge-laws, with that result that Kim’s model appeals to deduction in that context as well. The general lesson to be drawn is that epistemological reductionism requires that the lower-level laws imply – with some degree of strength – the higher-level laws. Moreover, even in the case of reductive explanations, it is necessary that the physical description of a mechanism M for a phenomenon P gives us information about the causal characteristics of P. Without being able to deduce the higherlevel descriptions, we are left with neither epistemological reduction of laws nor simple reductive explanations. Secondly provided that epistemological reductionism requires the deducibility of laws, it also requires the establishment of systematic links between the vocabularies of the higher and lower-level theories. This supposes in turn the co-extensional pairing of theoretical terms coming from each theory, what usually takes the form of bi-conditional bridge principles. Without bottom up conditionals, there is obviously neither deduction of laws nor reductive explanation. Without top down conditionals, the legitimacy of the higher-lever theory collapses. As we have shown, if there is nothing common from the point of view of physics about the heterogeneous physical entities coming under the same psychological law, and if physical property tokens are all there is, than psychology fails to grasp any objective feature of the world and should be struck from the list of sciences. The multiple realization thesis shows, given the history of the arguments we have rehearsed so far, that top down conditionals between psychological types and an ensemble of disjunctively joined physical types has lead the debate to two alternative and unsatisfactory termini: either psychological property tokens must be eliminated, or ontological reductionism and the consensus scientific world picture (in which physics is complete) must be rejected (5.5 The general dilemma of multiple realization). To escape from this dilemma, we must thirdly require that a good account of epistemological reductionism should preserve and account for the capacity of psychology to legitimately provide homogeneous explanations of phenomena that are only hetero-

220

From psychology to neuroscience

geneously explained by physics. From this point of view, we obviously must reconsider the legacy of the multiple realization thesis and the role it has played in creating the unsatisfactory outcomes of previous attempts at reductionism. Meeting these three criteria is the major desideratum governing the argument of this chapter.

6.2 Starting point: an implication of multiple realization 6.2.1 Physical differences and causal differences The classic formulation of the multiple realization thesis asserts that it is possible that entities that lack any relevant physical commonalities can nonetheless fulfil the same causal role under the very same psychological type. This is usually taken to mean that certain entities differing with respect to their physical composition nonetheless fulfil identical causal roles. This is the relevant feature that allows them to be identically classified within the psychological taxonomy. Contra this view, let us argue that different configurations of physical property tokens differ from the causal point of view as well, even though they come under the same psychological type. The first justification for this claim is tied to the relation of dependency between a causal relation and its relata, because the causal relation is, after all, defined by its relata. The same holds from the point of view of the types. Given a causal relation of type R, obtaining between any two entities coming respectively under the types P1 and P2, any causal relation obtaining between entities grouped under other physical types will be different as well, because the relational type R is defined as obtaining between relata of types P1 and P2. Our second justification is empirical. Molecular configurations of different types have different causal dispositions to react to their environment. Think for instance about heavy water (D2O), whose composition differs from normal water (H2O) by the addition of a neutron in the nucleus of each of its hydrogen atoms. This sub-atomic compositional difference has impressive consequences on the causal disposition of these molecules. Heavy water boils at 101.4°C rather

Reduction by means of functional sub-types

221

than 100°C, freezes at 3.82°C rather than 0°C, has a different refractive index, and so on. In the third place, this claim is generalizable on the basis of Schoemaker’s epistemological argument, which we already mentioned (1984: section V). After all, everything we come to know with respect to an entity comes from our experience of this entity and this experience is mediated by the causal interaction we engage with it. In particular, causal interactions are what we use to classify an entity as being of a certain type. This holds for composition as well. We know that an entity is composed of different stuff from another one because the two react differently to their environment or our measurement instruments. If this weren’t the case, if both entities behaved exactly the same in every possible experimental circumstance, how would we know, experimentally, that the two entities were composed of different stuff? There would, in this case, be strictly speaking no evidence at all for considering that they have different compositions and that they should thereby fall under two different physical types. The point is the same one we made earlier with respect to any metaphysical position that disconnects the essence of properties and their causal role (4.2.4.1 Troubles with Humeanism). Such positions are committed to the claim that there are ontological differences that in principle are not accessible to our knowledge, in as much as they forever resist our empirical investigations. There are, in other words, differences that make no difference at all, which begs the question of what, exactly, their difference consists in. To put it in a nutshell, if two entities e1 and e2 have exactly the same causal powers in all conceivable situations, there is no justification for dividing them between a type P1 (which is identified with this same causal power) and a type P2, rather than taking them as both being under type P1 and eliminating type P2. Finally, the causal approach of properties that we developed above (4.2.4.2 The causal theory of properties) entails that differences in composition imply causal differences. The fact that it is by the causal dispositions of an entity we ultimately know, empirically, what its composition is, implies that the function projecting from the causal powers of the parts of an entity and their organization to its causal powers as a whole is injective: if a part p of an entity e is replaced by a

222

From psychology to neuroscience

part p’ of another type, or if two part p and p’ of different types are exchanged in e, then the causal capacities of e as a whole – how it will react in all situations into which it is placed - are altered. Suppose that a component is removed without changing anything to the causal power to the entity as a whole, we would then have good grounds for claiming that the removed part was in reality not a genuine part of the entity that enables it to fulfil a certain higher-level causal role, but rather an independent entity e’ accidentally located within the spatiotemporal region occupied by the entity e. The same holds in both our other cases, in which the absence of any effect would suggest that the removed part of e wasn’t really part of e. 6.2.2 Implications for the multiple realization argument The fact that physical differences necessarily imply causal differences has deep consequences for the multiple realization thesis. Suppose that two entities e1 and e2 are grouped under the very same mental description M, but come under the physical descriptions, respectively, of P1 and P2; e1 and e2 are therefore different configurations of physical property tokens, which implies that they have different causal power and that they consequently fulfil slightly different causal roles as well. In other words, the multiple realization thesis implies that the physical realizers causally differ and fail to implement the very same causal role tout court. To express it in terms of property tokens, imagine that property tokens m1 and m2 are both tokens of the same psychological type M and that they are respectively identical to configurations of physical property tokens p1 and p2. If p1 and p2 belong to two different constructed physical types, say P1 and P2, they must differ causally, which means that m1 and m2 differ as well. This is obviously consistent with our previous conclusion, according to which two mental property tokens of the same mental property type can be respectively strictly identical to two different configurations of physical property tokens that fail to be perfectly similar, implying that the former are not perfectly similar either (3.7.1 Remarks on the concept of ‘identity’), including from the causal point of view.

Reduction by means of functional sub-types

223

This means that there are differences in the respective ways in which p1 and p2 fulfil the causal role R, which defines the mental type M under which they come. In other words, configurations of property tokens bring about the characteristic effect of the psychological type M in a specific physical way. In fact, this is the reason that, according to the multiple realization argument, two different reductive explanations of the mental property tokens m1 and m2, must refer respectively to p1 and p2. Against the background of ontological reductionism, the causal relation obtaining between m1 and its effect and the one obtaining between p1 and this effect is one and the same causal relation. Let us label this causal relation r1. The same is true of the causal relations obtaining between m2, p2, and their respective effects. We are, here, in the face of another single causal relation, say r2. The central question is now to explain how it is possible that r1 and r2 could appear as identical from the point of view of psychology, whereas they appear as different from the point of view of physics. An intuitive way of making sense of this situation is to say that psychology ropes together entities with a high degree of causal similarity, to the neglect of other causal details that do not matter from its point of view. This point takes up the criticism of the multiple realization argument developed by Bechtel and Mundale (1999), according to which the problem of multiple realization occurs because psychology uses coarse-grained descriptions whereas physics uses more fine-grained ones. It is this granular difference that gives rise to the fact that mental descriptions have a larger extension than the physical ones. Entities that are significantly similar from the causal point of view can thereby come under the same mental description in spite of their differences, because psychology neglects these compositional and related causal differences. Psychology grasps, through its law-like generalizations, causal relations that are sufficiently similar for its classificatory purposes. Against the background of ontological reductionism, this is the ultimate consequence of the multiple realization argument. As a corollary, when we make the grain of psychological descriptions as fine as the grain of the physical ones, multiple

224

From psychology to neuroscience

realization vanishes, and it is possible to reach the co-extensionality between sharpened functionally defined psychological descriptions and physical descriptions. This is the central feature of our strategy of epistemological reduction by means of functionally defined subconcepts (Esfeld and Sachse 2007; Sachse 2007; Soom, Sachse and Esfeld 2010). If causal differences distinguish the entities coming under the same higher level type, then there must, in principle, be differences in the causal role they fulfil. Now, if these causal differences can somehow be grasped functionally, i.e. using the descriptive resources of common-sense functionalism, then it is possible to construct fully functionally defined psychological sub-types that are co-extensional with physical types. As we shall see, the fact that those functional types are fully functionally defined provides us with a working explanation of psychology’s capacity to explain homogeneously entities that are heterogeneously explained by physics. Let us enter into more detail here, starting by examining the claim that that any causal difference between entities coming under the same psychological type can be grasped using the descriptive resources of folk psychology.

6.3 Reduction by means of functionally defined subtypes 6.3.1 Causal similarities and the functional individuation of psychological types Consider again the consequence of the multiple realization thesis in the context of ontological reductionism. Entities e1 and e2, coming under the same mental description M but under different physical descriptions P1 and P2, have necessarily different causal dispositions. However, the fact that they come under the same mental description guarantees that each of them fulfils the requirement of the functional definition of the psychological type M. If this were not the case, these entities would simply not count as tokens of the psychological type M. To preserve the simplicity of our exemplification, let us here simply take the mental type M be defined by the execution of the causal role C, although, as we saw in Chapter 4, the functional definitions of mental properties are

Reduction by means of functional sub-types

225

pretty complex (4.2.2 Common-sense functionalism pursued). But nothing here will depend on it. Given that both entities are different, each of them fulfils the causal role R in a specific way. This just means that each of them fulfils slightly different causal roles, say the causal role of e1 is R’ and the causal role of e2 is R’’. Since both of them fall under the same mental types, R’ and R’’ are nonetheless similar. This suggests analysing the situation as follows. e1 fulfils the causal role R’ = R + some causal specificity S1 whereas e2 fulfils the causal role R’’ = R + some causal specificity S2. This opens the way to two central observations. Firstly, it is evident that these specificities S1 and S2 in the respective causal roles of e1 and e2 can be observed using the descriptive resources of physics. After all, it is precisely because these entities react differently to certain environmental conditions that we are able to distinguish them – otherwise, we would have no reason to think that they are physically different. Consequently, there are environmental circumstances in which different entities coming under the same psychological type are distinguishable using the descriptive resources of physics. The specificities S1 and S2 are therefore graspable in physical terms. As we shall see later, this is exactly what happens when a neuroscientist observes through neuroimaging the patterns of neural activities resulting from the execution of certain behavioural tasks (7.2.2 Example of complex neuropsychopathology: Schneiderian schizophrenia). The second one is that any such causal difference is in principle detectable using the descriptive resources of psychology. (Sachse 2007: 139-40). To see this, assume that there is a world w1 similar to our, i.e. in which the distribution of mental property tokens and physical property tokens is identical to the one that obtains in our world. Accordingly, there is multiple realization in w1. Mental property tokens and physical property tokens are consequently distributed in w1 at a time t1 so that there are entities that can be described using the same psychological description whereas they come under different physical descriptions. Let us compare w1 with another possible world w2, which we characterize at first by the fact that the distribution of mental property

226

From psychology to neuroscience

tokens at t1 is identical with the distribution of mental property tokens in w1 at t1. However, w2 is distinct from the physical point of view in not allowing multiple realization. This means that in w2, all entities that come under a given mental description come under the same constructed physical description. As a result, mental and physical descriptions are co-extensional in w2. It is clear that in such circumstances, w1 and w2 are not distinguishable from the psychological point of view at t1. However, it is easy to imagine that after a certain length of time, at t1+n, both worlds will nonetheless diverge from the psychological point of view. In w1, entities coming under a given mental description fulfil, due to their physical differences, slightly different causal roles, which generate some differences with respect to the causal evolution of w2 that may be important enough to lead to differences within the distribution of mental property tokens that would be picked up by psychology. This example assumes that at some point of time, a divergence within the respective distributions of mental property tokens in w1 and w2 will occur. It means that at this point of time, an entity coming under a mental description M and a physical description P1 in w1 and its counterpart in w2, coming under M as well but under P2 from the physical point of view, will react differently to their environment, giving rise to a psychological difference between both worlds. There are therefore environmental circumstances in which entities coming under the same mental description but under different physical descriptions differ causally in such a way that this causal difference is graspable within the vocabulary of psychology. It is important at this point to emphasize that the claim is neither that any pair of physically different entities coming under the same mental description can be discriminated from the psychological point of view in every possible environmental circumstance, nor that the same set of environmental conditions can be used to make this crucial distinction with respect to any pair of entities coming under the same psychological type. It is rather that, for any pair of entities coming under the same mental description, but under different physical descriptions, we can imagine at least one set of environmental conditions such that these entities can be distinguished from the point of view of psychology. Let us assume

Reduction by means of functional sub-types

227

that for any physical difference between two entities coming under the same mental description, there is a set of environmental circumstances such that the causal specificities proper to each entity are graspable from the point of view of psychology. Let us now integrate our conclusion with our previous consideration of the individuation of the causal role of two entities, e1 and e2, differing from the point of view of physics but not of psychology. As said above, both of them fulfil the psychological causal role R and fulfilling this causal role is a sufficient condition for e1 or e2 to be qualified as tokens of type M: ∀(e) (Me ↔ Re) where R stands for “fulfils the causal role R”. Now, it is customary to assume that any entity of the mental type M fulfils its typical causal role in any environmental context c, as long as this context matches the domain defined by the ceteris paribus clauses proper to the psychological laws in question. Note that we are here confronting an implicit assumption of the classical interpretation of the multiple realization thesis, according to which all realizers of a mental property type M fulfil without any difference the causal role assigned by M, whatever may be the context within the domain allowed by the ceteris paribus clauses. ∀(c) ∀(e) (Me ↔ Re) with c belonging to the relevant ceteris paribus domain. But according to the consequence of the multiple realization thesis in the context of ontological reductionism, entities e1 and e2 are only imperfectly similar from the causal point of view. They must fulfil slightly different causal roles, say R’ and R’’, which can be conceptualized as the fulfilment of the causal role R with some particular specificities S1 and S2: ∀(e) ((Me ∧ P1e) ↔ R’e) where R’ = R + S1 ∀(e) ((Me ∧ P2e) ↔ R’’e) where R’’ = R + S2

228

From psychology to neuroscience

Now, if we stick to the claim that there are environmental circumstances c in which the causal specificities S1 and S2 distinguishing two entities that come under the same mental description can be grasped from the psychological point of view, it is possible to build two new functional definitions M’ and M’’, each of them being specific to one physical type of entity such that it fulfils the overall causal role R (where c is compatible with the ceteris paribus clauses proper to F): ∃(c) ∀(e) ((M’e) ↔ (Re ∧ cS1e)) ∃(c) ∀(e) ((M’’e) ↔ (Re ∧ cS2e)) where S1 and S2 are specified using the descriptive resources of folk psychology. A more detailed example of such close-grained functional descriptions will be presented in the next chapter (7.1.2 Application of the sub-types strategy). Before exploring the benefit in plausibility that accrues from this move to epistemological reductionism, let us note three precautionary conditions. First of all, while our guiding idea is to show that we can make functional psychological descriptions more fine grained, we don’t have to describe in the vocabulary of psychology the environmental conditions in which different entities coming under a given psychological description can be distinguished. All that matters here is that the precise way in which these entities fulfil their respective causal role can be described fully functionally under particular environmental circumstances. These environmental circumstances themselves exist as the external complements to the exhibition of the causal role. Their very possibility and the fact that they may in principle be described using psychological or physical vocabulary is sufficient to give a purely functional characterization of the causal role of any possible entity coming under the psychological concept M. Secondly, providing an example of environmental conditions in which physical differences among entities coming under the same psychological type do not lead to causal differences that are salient from the point of view of psychology is not sufficient to refute our present point. In order to refute it using a counter-example, one as to

Reduction by means of functional sub-types

229

prove that there are two entities, coming under the same psychological type but under different physical descriptions, such that they behave in every possible environmental circumstance in a way which is similar enough to exclude that the causal differences distinguishing them is not graspable using the descriptive resources of common-sense psychology. To reiterate, what is required here is the possibility of providing one set of environmental circumstances under which physically different entities coming under the same mental type can be distinguished from the psychological point of view. As we shall show later (7.2.1 Neuropsychology and experimental design), a large part of the empirical research in neuropsychology is built on the assumption that neurological specificities, be it brain lesions, neurotransmitter dysfunctions, or specificities in connectivity, are reflected in behaviour and that appropriate experimental set-ups enables to exhibit the behavioural impact of these physical specificities. Neuropsychological experimental designs perfectly illustrate our point according to which physical differences leads to psychological differences as well to the extent that appropriate environmental circumstances are provided. Finally, one may adduce as a counter-example to our argument that certain physical differences among entities coming under the same psychological type, whether or not they exhibit certain differences in one or another given environment, do not matter for the execution of the ordinary causal roles that concern us; they fail to lead to any relevant functional difference that could be observed from the psychological point of view. Although examples of this kind can easily be imagined, this does not affect the point under discussion. Following Shapiro (2000: 647), we can construct an answer as follows. Either such a difference concerns the physical properties of an entity e that are causally relevant for the execution of the causal role R, which is produced under the psychological type M, or not. If it is the case, then this difference affects the way in which e fulfils the causal role R with that result that our argument holds. If it is not the case, then the physical difference in question is not a relevant feature of the entity relative to distinguishing between two different physical types of entities fulfilling the causal role R. Such irrelevant differences should therefore not be taken into account here, since the multiple realization

230

From psychology to neuroscience

thesis concerns lower-level differences among physical characteristics that play an active role in the implementation of R. Let us reformulate Shapiro’s point in terms of reductive explanations. Either the differences in question concern the physical properties of the entity that are causally implicated within the execution of the causal role or not. If it does, then they are several reductive explanations of the property M. In this case, the different entities fulfil the causal role R in a different way and our previous point applies. If the difference does not concern the execution of the causal role in question, then there is no reason to consider that there are several reductive explanations, with the result that this difference is not sufficient to motivate a distinction between different physical types of entities realizing the same causal role since the physical properties exemplified by e relatively to the execution of the causal role R are identical. In other words, such a difference would even fail to motivate the claim that we face a case of multiple realization. This point obviously matches Craver’s point with respect to the causal relevance of constitutive mechanisms that we already have considered (Craver 2007: Chapter 4, especially p.153). If one is looking for a reductive mechanistic explanation of a phenomenon functionally defined by the causal role φ, then one is only interested in the parts of the entity fulfilling φ whose manipulation is relevant for φ, and vice-versa. Given the criteria of mutual manipulability in the context of mechanistic explanation of the execution of a certain causal role, a difference such as the kind of counterexample proposed above does not enter into consideration. Let us sum up this point. Any physical difference between entities coming under the same mental description entails that both entities differ causally as well. They fulfil therefore slightly different causal roles in fulfilling the same psychological causal role with its causal specificities. These causal specificities in the way the psychological causal role is executed can be observed are easily captured by physics, but they are also apparent from a psychological point of view, to the extent that appropriate environmental circumstances are provided.

Reduction by means of functional sub-types

231

6.3.2 Fully functionally defined sub-types and co-extensionality This theoretical detectability by psychology of physical difference among entities coming under the same mental description opens the way to the construction of psychological sub-types that are fully functionally characterized and co-extensional with the physical description of these entities, in terms that the following argument will unfold. Our first move is to take for granted the multiple realization of psychological types in our world. Thus, there are entities coming under the same mental description, but under different physical descriptions in virtue of certain physically distinguishing specificities. This implies that they have different causal dispositions, which will be expressed in some unknown number of possible environmental circumstances. They fulfil consequently slightly different causal roles. Second, comparing our world with a world in which there is no multiple realization, it is strongly suggested that these causal specificities are in principle graspable from the psychological point of view, as long as the environmental circumstances in which the causal difference manifests itself in a sufficiently salient way are taken into account. This means that psychology is in principle able to construct functionally defined, sharp-grained sub-types of any psychological type that will allow for psychology to grasp these causal specificities. Third, since it is possible to construct a pair of functionally defined sub-types to distinguish any two physically different entities coming under the same mental type at the psychological level, it is possible as well to construct a functionally defined sub-type for each relevant physical type of entity. It follows therefore that coextensionality between physical types and psychological sub-types can be reached by the procedure of making psychological types sharper grained that narrows their extension. Consider once again the individuation of functional psychological types and sub-types. The starting point is the functional definition of a mental type M, grasped by the following definition: ∀(e) (Me ↔ Re)

232

From psychology to neuroscience

Given multiple realization, it is possible that the different entities coming under the mental type M come under different physical descriptions, here P1 and P2. Consequently they differ causally so that the description constructed as a conjunction of the psychological type and the physical type applying to e are extensionally equivalent to a description grasping the precise causal role fulfilled by e. ∀(e) ((Me ∧ P1e) ↔ (R’e)) ∀(e) ((Me ∧ P2e) ↔ (R’’e)) Provided that any such difference can be grasped in psychological terms as explained above, it is possible to construct the following functionally defined sub-types of M: ∃(c) ∀(e) ((M’e) ↔ (Re ∧ cS1e)) ∃(c) ∀(e) ((M’’e) ↔ (Re ∧ cS2e)) where S1 and S2 are specified using folk psychology. This way of describing the causal role of entities coming under the physical description of P1 and P2 respectively is supposed to express the ability of psychology to grasp the causal dispositions that each physical type of entity manifests in some particular context. Now the crucial point is simply to observe that this procedure is available for any possible physical entity coming under the psychological concept M. It follows consequently that for any possible physical type of entity coming under a certain psychological type M, we can theoretically construct an adequate functional sub-type so that co-extensionality is obtained between physical types and psychological sub-types. ∃(c) ∀(e) ((M’e) ↔ (Re ∧ cS1e) ↔ P1e) ∃(c) ∀(e) ((M’’e) ↔ (Re ∧ cS2e) ↔ P2e) … and so on for any possible type of physical entity coming under the mental type M. Finally, let us relate the possibility of constructing fully functionally defined sub-types that are co-extensional with physical type to the issue of psychological law-like generalizations. If these psychological sub-types are available, it is de facto possible to construct more fine-grained psychological law-like generalizations. Let

Reduction by means of functional sub-types

233

us return to our previous examples of law-like generalizations, considering once again the law Lpsy. Lpsy : ∀x (M1x → M2x) Assume now that the causal behaviour of physically different entities is grasped by this law. As explained, these entities come under different physical laws such as: Lphys 1: ∀x (P1’x → P2’x) Lphys 2: ∀x (P1’’x → P2’’x) … and so on for any other physical types P1i and P2i of entities coming respectively under types M1 and M2. Now, if it is possible to construct psychological sub-types M1’, M1’’, …, M1n of M1 that are co-extensional with P1’, P1’’, … P1n and psychological sub-types M2’, M2’’, …, M2n of the psychological type M2 and the physical types P2’, P2’’, …, P2n, respectively, then it is also possible to formulate detailed psychological law-like generalizations such that there is a detailed psychological law-like generalization that is co-extensional with each physical law-like generalization grasping the causal behaviour of the physical entity coming under the psychological type M. Thus, for any Lphysi describing the causal behaviour of entities coming under the physical type P1i, there is a detailed psychological law-like generalization Lpsyi describing the causal behaviour of entities coming under the psychological sub-type M1i, such that Lphysi and Lpsyi have the same extension. 6.3.3 Relation between psychological types and sub-types Our strategy up to this point has been to show how the causal specificities that divide the different physical type of entities that may be found under a single psychological type allow us to construct functionally defined sub-types in order to reach co-extensionality between these psychological sub-types and physical types. At this point, it is crucial to consider in detail the relation obtaining between the introduced psychological sub-types and the original psychological types. After all, it could be argued that our proposed strategy simply reformulates the fragmentation of the psychological taxonomy in just the

234

From psychology to neuroscience

way we have criticized with some other mentioned positions do (5.3.3 Multiple realization in the current debate), in which case we would be driven to eliminativism or to the rejection of ontological reduction. Simply splitting each psychological type into several sub-categories does not provide any particular help in overcoming the problem of the multiple realization thesis if there is no further explanation of the process by means of which psychology homogenizes these different descriptive sub-categories in return (c.f. our discussion of NWR and Kim’s account of functional reduction). To put it in a nutshell, the approach under discussion in the present chapter provides a pretty natural way of understanding the relation between psychological types and their respective sub-types in term of abstraction from sub-types to types by bracketing the functional details distinguishing the different physical types of entities coming under the same psychological type so that the psychological types and law-like generalization are preserved as abstract but nonetheless perfectly well grounded scientific categories. Let us recapitulate these elements for the sake of our further argument. First of all, let us consider the difference between abstract psychological types and their respective functionally defined subtypes. The original abstract psychological types are defined by a typical causal role. Though we are simplifying the description of such causal roles to that of a unique variable for the sake of our exposition here, we recognize their practical complexity as captured in the discourse of common-sense functionalism, as discussed in Chapter 4 (4.2.2 Common-sense functionalism pursued). The functionally defined subtypes of an abstract psychological type M are in comparison more detailed, since they take into account both the causal role which is proper to the abstract type M and further contextual causal specificities differentiating them among themselves. Abstract psychological type M: ∀(e) (Me ↔ Re) Two possible sub-types M’ and M’’ of M: ∃(c) ∀(e) ((M’e) ↔ (Re ∧ cS1e))

Reduction by means of functional sub-types

235

∃(c) ∀(e) ((M’’e) ↔ (Re ∧ cS2e)) Our second observation is that any functionally defined sub-type is characterized by its conjunctive character, since it integrates both the causal role of the abstract type and at least one contextual causal specificity, by which it is distinguished from other psychological subtypes. Since a detailed sub-type can be seen as a complex conjunction of simpler causal specifications, the fact that a certain entity comes under a certain sub-concept entails that it comes as well under the abstract relative psychological type since the latter contain only the principal causal role present in the former. From this it ensues that abstract psychological types can be logically deduced from their subtypes in as much as a true conjunction allows us to deduce the truth of one of the conjuncts. At this point, the question is to know why we should privilege the causal role R rather than some specificity Si of a given sup-type Mi within this deductive process. Firstly, our proposal starts by considering higher level abstract types to build on that basis coextensive sub-types. From that point of view, the approach is driven by higher-level categories. Secondly, the behavioural specificities S1, …, Sn proper to each sup-type Mi, … Mn of the abstract type M are only manifested in very precise environmental conditions, which are very rare outside neuropsychological laboratories. Thereby, if one shorts these sub-types according to the most often occurring functional features of their instances, it is pretty natural to privilege the characteristic causal role R, which is always manifested. This deductive relation assures that we can reduce the level of functional details taken into account in order to deduce abstract psychological types from more detailed psychological sub-types while remaining within the scope of psychology. By reducing the level of detail taken into account, psychology neglects certain functional specificities that occur only for certain entities under certain environmental circumstances, providing thereby more encompassing explanations that focus on more common cases. As our example of the simplified definitions of the subtypes M’ and M’’ of the psychological type M shows, if one brackets the causal differences of the physical

236

From psychology to neuroscience

entities coming respectively under one or the other of the subtypes, then what remains in these definitions is nothing but the definition of the abstract psychological type M. However, the fact of focusing on more common cases in such a context does not undermine the scientific legitimacy of abstract psychological types. Given that both the abstract type and its sub-types are formulated within psychology, the only difference between an abstract type and its sub-types is the scale of the functional detail. Since this abstraction process remains within the scope of psychology, there is nothing illegitimate about using abstract types to precipitate out the commonalities of certain entities. Let us express this point in terms of extension of types and subtypes. Assume two sets of entities E’ and E’’ containing respectively entities coming under M’ and M’’. If the causal specificities S1 and S2 that entities in E’ and E’’ respectively present in context c are bracketed, it is natural to regroup these entities in a larger set E, which is the union of E’ and E’’, since all of them fulfil the causal role R, which is nothing but what defines the abstract psychological type M. Provided that the types and sub-types are defined within psychology, the question of focusing simply on the causal role R or R plus certain functional specificities is nothing but a matter of conceptual leeway. Finally, let us consider this relation of abstraction in the context of the law-like generalizations of psychology. As argued previously, the fact that it is possible to construct, for any physical type Pi of entity found under a given psychological type, a functionally defined psychological sub-type Mi such that Mi is co-extensional with Pi, entails that there is a psychological law Lpsy i, which describe the causal behaviour of the same set of entities as Lphys i. The crucial issue is here to describe the relation obtaining between a psychological law Lpsy grasping the causal behaviour of entities coming under the psychological type M, and the psychological laws Lpsy1, …, Lpsy n describing the causal behaviour of the entities coming under each subtype M1, …, Mn of the psychological type M. Firstly, the detailed psychological law-like generalization Lpsy 1, …, Lpsy n are all about the production of the same type of effect, as all

Reduction by means of functional sub-types

237

of them contain implicitly the functional specification that is characteristic of the abstract psychological type M (aside from the causal specificities that fall under their scope). Secondly, on the basis of these more detailed law-like generalizations, it is always possible to retrieve the abstract law-like Lpsy by abstracting from the details concerning the different ways the entities falling under the scope of different sub-types produce the effect that is characteristic from the abstract psychological type M. Thirdly, given that abstract types, their sub-types, abstract law-like generalization and detailed law-like generalization are all formulated using a psychological vocabulary, this abstraction process is nothing but an intra-theoretical variation in the degree of details that will be considered by psychology. Since from a psychological point of view all entities coming under one the detailed psychological law-like generalizations Lpsy 1, …, Lpsy n are entities causing an effect corresponding to the abstract law-like generalization Lpsy, Lpsy and Lpsy1, …, Lpsy n grasp an objective feature of reality without endangering the legitimacy of psychological laws. Crucially, the difference between Lpsy and Lpsy 1, …, Lpsy n is an intra-theoretic question, which means that multiple realization has become an intratheoretic issue as well instead of an inter-theoretic problem. We shall examine the benefits of such a move below in detail, by a systematic comparison with the New Wave reductionism and the Kim’s account of functional reduction. Let us conclude these considerations with respect to the relation between abstract psychological types and their respective sub-types by the following remarks. Firstly, abstract types and detailed psychological sub-types essentially differ in regard to the amount of functional details expressed in each, with the latter taking into account additional functional specificities that the former brackets. Secondly, in reason of the conjunctive character of the sub-types, the abstract psychological types are deductible from any of their sub-types. Thirdly, the possibility of both constructing psychological sub-types and deducing abstract types on their basis grounds the intra-theoretic process by means of which psychology is able to consider more functional details or less in order to throw into relief objective commonalities between certain entities, that is, either the fulfilment of

238

From psychology to neuroscience

a certain causal role or, if we attend to a finer grained level of description, the particular characteristics potentially differentiating these entities as they fulfil their causal role in specific environmental conditions. Last but not least, let us add here that we are dealing with what grounds the theoretical indispensability of psychology – a criterion that, as we have seen in previous chapters, is not obtained by other reductionist methodologies - since it is only by the appeal to abstract types that we are able to bring out what relatively large set of entities have in common from the functional point of view. This grounds the importance and the legitimacy of psychological law-like generalizations, which are built on the basis of these abstract types, which are themselves the product of the abstraction process by mean of which psychology comes to bracket the contextual causal specificities of different sub-types of entities as they execute the same general causal role. 6.3.4 Relation between functionally defined sub-types and physical types Having clarified the relation between abstract psychological types and functionally defined sub-types in the last sub-section, let us consider in detail the relation between these functionally defined sub-types and the physical types to which they are co-extensional. As explained, it is possible, for any physical type of entity falling under the domain of a certain psychological type M, to construct a corresponding functionally defined psychological sub-type such that the latter is co-extensional to the former. Against the background of the classical Nagelian model of epistemological reductionism, this means that it is possible to construct bi-conditional bridge-principles correlating pairs of types of each theory. Therefore, any law-like generalization formulated using psychological sub-types can be deduced from a physical law-like generalization formulated using appropriate physical types. In other words, within the Nagelian framework, any detailed psychological law-like generalization can be epistemologically reduced to an appropriate physical law.

Reduction by means of functional sub-types

239

However, as we have argued, Nagelian bridge-laws are not explanatory per se. They do not explain why entities of a certain physical type fall under a certain psychological type as well. Provided that the psychological sub-types under discussion are fully functionally defined, our approach may justly claim to inherit the positive outcome of Kim’s account of functional reduction vis-a-vis reductive explanations. Premising the completeness of the laws of physics, they allow us to explain how certain entities, falling into the domain of both a certain psychological sub-type and its co-extensive constructed physical type, fulfils the causal role which is characteristic of that functionally defined psychological sub-type; thus, we discover why the entities in question are tokens of the considered psychological subtype. Moreover, this physical reductive explanation is in principle able to explain in detail how the entity fulfils the causal role in any possible circumstances, which means that our account must include reductive account of the differentiating contextual functional specificities between sub-types of the same psychological abstract type. This point is evidently consistent with our criteria for mechanistic explanations (4.3.2.1 Mechanisms and mechanistic explanations), according to which the physical description of a mechanism refers to the organization and to the causal dispositions of those of its constitutive components that are causally relevant for the execution of the causal role defining the phenomenon. In this way we construct, as well, an explanation that concerns the precipitating conditions, the possible modulations and the by-products of the phenomenon. Given the possibility of constructing reductive explanations of each functionally defined psychological sub-type and their coextensionality with appropriate constructed physical types, it follows that the latter provide homogeneous reductive explanations of the former. In other words, as long as psychology is restricted to its most fine grained sub-types, psychology can be epistemologically reduced within Kim’s framework of functional reduction. Psychological functionally defined sub-types and the corresponding law-like generalization can therefore be epistemologically reduced to physics both in the Nagalian sense and through homogeneous reductive explanations without damage to their scientific integrity.

240

From psychology to neuroscience

6.3.5 Overview of the strategy of epistemological reductionism Against the background of these elements, we can formulate the general strategy of epistemological reductionism that is pursued here in five steps, grounded in our argument so far. Like Kim’s account of functional reduction, our account of epistemological reductionism by means of functionally defined subtypes starts with the functionalization of the psychological property types. As we have argued, common-sense functionalism provides the theoretical resources to obtain the necessary functionalization as long as we take the causal role of mental property tokens falling under the types to be their essential feature. The functional definition of a mental property type M typically contains a disjunction of possible causes, a disjunction of possible effects and a list of possible relations obtaining between tokens of type M and other types of mental property tokens. Given the complexity of these functional definitions, let us simplify our argument by representing them by short variables. Let, for instance, R be the causal role which is essential to the mental property tokens of type M, whose full functional definition has the form discussed in Chapter 4 (4.2.1.3 Common-sense functionalism). The second step consists in making physical descriptions of the different entities that may fulfil the causal role R. Our assurance that these descriptions can be constructed is guaranteed by the supervenience of mental property tokens on physical property tokens (3.3.1 Varieties of supervenience) as well as the principle of ontological reductionism (3.7 The token-identity thesis as ontological reductionism), according to which any entity that falls in the domain of a mental description must also fall under a physical description. These descriptions are so constructed as to capture all the physical features of the entities coming under M relevant to the execution of the causal role R, bracketing all other features. Thirdly, the multiple realization thesis shows that entities of different physical types, P1, … Pn, are capable of fulfilling the causal role R, which implies that there is physics cannot give us a homogeneous account of the mental property tokens of type M. Instead, there is such reductive explanations in correspondences with each

Reduction by means of functional sub-types

241

physical type of entities that falls in the domain of the psychological type M (6.3.4 Relation between functionally defined sub-types and physical types). Fourth, the multiple realization thesis implies that the different types P1, … Pn of entities that disjunctively fall within the domain of the psychological type M must differ causally to some degree or another, fulfilling slightly different causal roles R1, … Rn. This only means that descriptions P1, … Pn describe different mechanisms that are causally responsible for the execution of R, each of them bringing about the effect resulting from the execution of R with a distinct causal specificity Si. For any physical type of entity Pi fulfilling the causal role Ri, there is a set of environmental conditions c in which Pi manifests Si in a functionally salient way, i.e. that can be understood using a folk psychological description (6.3.1 Causal similarities and the functional individuation of psychological types). It is therefore possible to build a functionally defined sub-type Mi for any entity of physical type Pi, picked out of the types P1, … Pn falling in the domain of the psychological type M, such that the functional specificities Si of Pi relative to appropriate environmental conditions c can be accounted for in the psychological vocabulary. Accordingly, we can construct functionally defined sub-types M1, …, Mn that are co-extensional with the physical types P1, … Pn (6.3.2 Fully functionally defined sub-types and co-extensionality). Provided that each of these psychological sub-types M1, …, Mn has a conjunctive logical form, including both the causal role of the psychological type M and the contextual functional specificities proper to the physical type of entities to which they are respectively specific, our intra-theoretic problem is a technical one, in order to select the appropriate level of detail that will allow us to bring out objective causal similarities that are shared by physically heterogeneous entities. Psychology is therefore able to provide homogeneous explanations by bracketing the contextual functional specificities S1, …, Sn of the entities coming under the psychological sub-types M1, … Mn according to some well grounded criteria, since all functionally defined psychological sub-types must include the causal role R defining the abstract psychological type M.

242

From psychology to neuroscience

Fifth, since the psychological sub-types M1, …, Mn are coextensional with the physical types P1, …, Pn, it is possible to reduce the former to the latter using the tools of Nagelian epistemological reductionism, i.e. bridge laws. Furthermore, given that the functionally defined sub-types are entirely functionally defined, we can also use homogeneous reductive explanations tools as provided by Kim’s account to reduce M1, …, Mn to P1, …, Pn.. Each of these homogeneous reductive explanations explains hence using a physical type Pi, how certain entities fulfil the causal role Ri, which is specific to the sub-type Mi, and thereby why these entities come under the psychological subtype Mi. Finally, given the fact the psychological sub-type Mi is in part defined by the causal role R, this reductive explanation also explains how these entities fulfil this causal role, and thereby why they come under the abstract psychological type M (6.3.4 Relation between functionally defined sub-types and physical types). Epistemological reductionism is thereby achieved by means of functionally defined sub-types. The construction of functionally defined psychological sub-types is the cornerstone of this strategy of inter-theoretic reduction for two reasons. Firstly, given that types stand in laws-like generalization as nomological predicates, the coextensionality between psychological sub-types and physical constructed types enables the reduction of the detailed psychological law-like generalizations to physical laws. Second, qua being fully functionally defined, these psychological sub-types constitute the starting point of an explanation of the ability to provide homogeneous explanations of phenomena that only heterogeneously explained by physics. Indeed, psychology is able to focus on relatively large set of entities by bracketing the contextual functional specificities that would merely impede our vision of their objective commonalities. To put it in differently, although the entities falling under the same abstract psychological type are identical neither from a physical nor from a causal point of view, they are nonetheless similar enough because all of them fulfil a certain causal role with some contextual specificities that psychology is free to take into account or not. The following figure illustrate the general features of our account of functional reduction by means of functionally defined sub-types.

Reduction by means of functional sub-types

Abstract mental type

M Abstraction M1

243

M2

Functional sub-types

e2

Truthmakers

P2

Physical constructed types

Applies to e1 Applies to P1

Overview of the sub-type strategy. The functionally defined sub-type M1 and the constructed physical type P1 apply to every entity in the set E1. M1 is therefore coextensive to P1. The same holds for M2, P2 and E2. The abstract type M applies to all entities, and only to them, contained in the union of E1 and E2 and it abstracts from the functional contextual details distinguishing M1 from M2.

6.3.6 Conservative reduction of psychology Our proposed strategy of epistemological reductionism is conservative. It leads neither to the elimination of psychological types nor to that of mental property tokens. To the contrary, it vindicates the scientific indispensability of psychology. Let us clarify these claims by contrasting the present account in this regard with those of New Wave Reductionism and Kim’s account of functional reduction. Within the NWR framework, the elimination of psychology comes in two steps. First, as a result of the multiple realization thesis, there are several base theories from which several image-theories mimicking the explanatory role of psychology can be inferred. Secondly, NWR is unable to do justice to the ontology of psychology, whatever the degree of isomorphism between the each constructed image-theory and psychology. In this way, NWR undermines the

244

From psychology to neuroscience

scientific legitimacy of psychology to grasp objective feature of reality. In the best case, the constructed image-theories are perfectly isomorphic to psychology with that result that each of these imagetheories has the same explanatory power as psychology. However, they are embedded in a more complete theoretical framework, which can be applied outside the domain of psychology, suggesting that each of these theories stands, as re-described in physics, in a closer relation to what is out there in the world, especially in the face of the multiple realization thesis, according to which the physical entities corresponding to mental properties have, theoretically, nothing significant in common. Qua favouring the ontology of the basetheories, NWR is thereby driven to the claim that psychology fails to grasp any objective feature of reality. The situation is even more straightforward in the case where there is no isomorphism between the image-theories and psychology. These image-theories simply possess better explanatory power than the higher level theories of psychology, and the latter should simply be replaced as superfluous. In both cases, psychology lacks truthmakers from the point of view of the more fundamental theory with that result that its taxonomy is undermined and its ontological standing is collapsed. But this is a move too far even by the standards with which NWR sets out, which was to reduce, and not eliminate, psychology. In contrast, our strategy of epistemological reductionism does not lead to the replacement of psychology by lower-level theories, nor to the elimination of its ontology, since we aren’t basing our project on the possibility of finding some lower-level necessary and sufficient conditions that would root mental property tokens in the physical make up of the world. Rather, it starts by considering the functional individuation of abstract psychological types and, in a second step, analyzes the causal heterogeneity of the different physical types of entities falling within the domain of a given psychological type in order to build co-extensional functionally defined psychological sub-types. Given this co-extensionality, these sub-types can be reduced to constructed physical types without any injury to psychology’s scientific status while constituting, qua being a conjunction of functional specifications, the base for the abstraction leading back to the original

Reduction by means of functional sub-types

245

psychological types. Against that background, certain entities in the world are mental property token of a certain type because they fulfil similar causal roles, with such contextual differences that may or may not be taken into account by psychology, according to the level of its descriptive focus. But nothing here suggests that the taxonomy of psychology is mistakenly constructed or that there are no mental property tokens out there in the world. A comparison with Kim’s account of functional reduction raises the question of why Kim’s species-specific types do not play a similar role in giving us a method to conservatively reduce psychology. Admittedly, our account and Kim’s are similar. Both start by considering higher level descriptions and both build sub-types in order to obtain co-extensionality with physical types, which in turn provide homogeneous reductive explanations of the introduced sub-types. Just as we claimed that psychology abstracts from the functional specificities distinguishing the different sub-types of a certain abstract psychological type, one could argue that psychology similarly abstracts from the physical specifications distinguishing Kim’s hybrid speciesspecific types. However, Kim’s account disallows such a move for the simple reason that the relation between the species-specific types and the psychological types remains an inter-theoretic relation while the process of abstraction characterizing our approach is intra-theoretic. The problem can be stated as follows. Firstly, in order to be reducible by means of homogeneous reductive explanations, a Kimian speciesspecific type H1 has to be nomologically co-extensional with an appropriate constructed physical type P1 with the result that both types apply to entities fulfilling exactly the same causal role R1, from any possible point of view. Secondly, multiple realization implies that different physical types P1, …, Pn of entities, which fulfil thereby different causal roles R1, …, Rn, come under the same psychological type M, which is itself characterized by a general causal role R that is obviously not identical with the causal roles R1, …, Rn. Fourth, the question is to know whether or not it is not useful to introduce speciesspecific types H1, …, Hn as hybrid sub-types of the type M in order to explain the relation of similarity obtaining between the causal role R1,

246

From psychology to neuroscience

…, Rn that is fulfilled by the entities falling within the conjunction of P1, …, Pn, and the causal role R, which defines the psychological type M. Unfortunately, this move is not conclusive, at least for two reasons. Firstly, as long as these hybrid sub-types are individuated by both a functional criteria and an additional criteria specifying the physical type of entity fulfilling R, we have missed the challenge of multiple realization, which is to explain why the entities coming under P1, …, Pn, fulfilling the causal role R1, …, Rn, are sufficiently similar from a purely psychological point of view in order to come under one single psychological type defined by the causal role R. And even if the additional physical criterion embedded in each hybrid species-specific type is somehow converted into a physical causal role Ri, so that they are individuated by a psychological causal role R and a physical causal role Ri, the second problem is that we still don’t have our explanation. For in this case, the species-specific types just contain the same information in two different vocabularies. This heterogeneity of vocabulary undermines any standard for evaluating the degree of similarity between R, R1, …, Rn. Against the background of their physical dissimilarities, the absence of any functional criteria taking into account the causal differences differentiating the entities of the physical type fulfilling R suggests that the causal commonalities grasped by psychology do not exist. It follows that, from the scientific point of view, psychology must be eliminated as ungrounded. This is why Kim’s account, qua failing to provide an account of the relation between the hybrid species-specific types and the original psychological types, is an elaborate exercise in displacing the multiple realization problem rather than solving it. Instead of placing the difficulty raised by the multiple realization thesis between physical and psychological types, it places it between these psychological types and the species-specific types. This is problematic because the relation between the latter and the former remains an inter-theoretic relation. By contrast, the present approach provides the standard necessary to evaluate the degree of similarity of the different physical types of entities falling within the domain of the same mental type. Given that the sub-types are entirely formulated within the vocabulary of

Reduction by means of functional sub-types

247

psychology, it is possible to evaluate their degree of similarity, which would not be possible if the sub-types were individuate in a hybrid way. To put it in a nutshell, if two physical types of entities differ significantly from the psychological point of view only under very rare circumstances, then it is legitimate to take them as sufficiently similar to come under the same psychological type. If they differ most of the time, our standard tells us that the psychological taxon in question is not well grounded. The strategy of epistemological reduction by means of functionally defined sub-types avoids the devastating eliminative consequence of multiple realization because it fixes, on the one hand, the inter-theoretic problem by the construction of psychological subtypes that are co-extensional with physical types, while dealing with the multiple realization issue, on the other hand, as an intra-theoretic issue concerning the degree of focus of the psychological description. It should be noted that this approach not only integrates the multiple realization thesis but also uses it to establish the theoretical indispensability of psychology. As argued earlier, it is only through recourse to psychological abstract types that we can see that certain entities, which otherwise fail to have any relevant physical feature in common, are nonetheless sufficiently similar in falling under the domain of the same higher-level type once we bracket their respective contextual functional specificities. To sum up the most important point here, it is only by means of this strategy of epistemological reductionism that we can conserve the legitimacy of psychological classification, independently of the fact that the mental properties may be enacted through heterogeneous physical types. In other words, treated as an intra-theoretical issue, multiple realization does not motivate the elimination of psychology, but rather establish its theoretical indispensability.

6.4 Summary and transition We started this chapter by enumerating three fundamental desiderata that must be met by any account of epistemological reductionism aiming to avoid the elimination of psychology: first, the deducibility of laws; second, the establishment of bi-conditional links between the theoretical terms of each theory; and third, the necessity of explaining

248

From psychology to neuroscience

the ability of psychology to provide homogeneous explanations of phenomena that are heterogeneous from the point of view of physics. Let us conclude this chapter by recapping how our account meets these three constraints. The strategy of epistemological reduction by means of functionally defined sub-types starts by considering the implications of the multiple realization thesis against the background of ontological reductionism, namely the fact that different physical types of entities falling within the domain of the same psychological type also differ with respect to the causal role they fulfil. After all, if two mental property tokens are respectively identical to two different configurations of physical property tokens and if these latter are different, by definition, in being causally divergent in some theoretical situation, then it follows that these mental property tokens must also be causally different, at least in some theoretical situation. These causal differences open the way to the construction of psychological subtypes, taking into account the causal specificities of each physical type of entities falling in the domain of some given abstract psychological type in a purely functional vocabulary. Provided that a psychological sub-type can be constructed for each physical type falling within the domain of some given abstract psychological type, we obtain a coextensionality between psychological sub-types and physical types. On that basis, it is possible to reduce the former to the latter either by means of classical reduction or by means of homogeneous reductive explanations, which extend down to the descriptions of the fundamental physical mechanisms responsible for the production of behaviour. Given the close relation between types and laws, this approach obviously enables us to deduce the laws embedding the appropriate psychological sub-types on the basis the laws of physics. As explained when we discussed the completeness of physics (3.4.1 Completeness of physics), causal relations between entities implies nomological relations as well. It is therefore possible to associate a dedicated physical law for each physical type of entity coming under the same abstract psychological type and the same approach is available for each nomologically co-extensional psychological sub-types, so that we can

Reduction by means of functional sub-types

249

achieve the golden reductionist standard of deducing the psychological detailed laws within the Nagelian framework by mean of bi-conditional bridge laws. Finally, this strategy of epistemological reduction resolves the problem of how psychology can provide homogeneous explanations of phenomena that are heterogeneously explained by physics by dealing with multiple realization as an intra-theoretic problem. Given the fact that the psychological sub-types are co-extensional with physical types, they constitute the sharpest degree of psychological descriptive focus possible. Now, through their fully functional individuation, which takes into account the fulfilment of the original causal role R of a certain psychological abstract type M with certain contextual specificities S, it is possible to abstract from these contextual specificities in order to retrieve the original psychological type M. In short, working with psychological abstract types or some of their sub-types is a matter of scientific leeway, depending on the explanatory interests. Last but not least, the interpretation of multiple realization as a result of an intra-theoretical variation in the grain of psychological description confers on psychology its theoretical indispensability, since it is only by the appeal to abstract psychological types that it is possible to bring out the commonality between otherwise disparate entities. Our account shows why it is valid to believe that these physically different entities are similar enough to fall within the domain of one abstract psychological type, because from a psychological point of view, what distinguishes them is only a matter of particular way of manifesting their causal dispositions in very precise environmental conditions. These environmental conditions will be examined in the next chapter, which aims to provide examples of applications of this strategy with respect to the relation between psychology and neuroscience.

Chapter 7 REDUCTION OF PSYCHOLOGY TO NEUROSCIENCE: CASES STUDIES The account presented in the previous chapter depends crucially on the claim that for any two individuals falling under the scope of the same psychological description while differentiated from the point of view of a lower-level descriptive modality, one can always conceive a set of environmental circumstances that would enable us to distinguish them on the basis of purely behavioural considerations. This means that it is possible to build purely functionally defined sub-types in order to achieve a conservative reduction. In this chapter, we examine empirical examples that illustrate this claim. Our first example illustrates how environmental conditions may allow us to construct fully functionally defined sub-types. The relative simplicity of the example chosen, namely the case of the disconnection syndrome, raises the question of the applicability of our claim to more complex cases. In order to show why the degree of complexity is, ultimately, irrelevant, we shall consider in more detail both the object and the methodology of neuropsychology, as applied, as an instance, to understanding Schneiderian Schizophrenia. Our reflection will show that it is possible to find experimental set-ups that allow for the distinction between patients and normal individuals who differ only subtly from the neurological point of view. Finally, we shall consider some empirical results showing that this link between behavioural specificities and neurological specificities extends down to neurobiology, which vindicates our position that even the fundamental biochemical components of the nervous system can be encompassed by our theory.

252

From psychology to neuroscience

7.1 How it works: guidelines to reducibility in principle This section introduces our first example of applied epistemological reductionism by means of functionally defined sub-types – that of inter-hemispheric disconnection – which we highlight to present every features of the reductive strategy under discussion in the light of well known and easily understood empirical results. This simple example will allow us to organize our argument without any distracting collateral explanations. We shall then turn to more complex cases in the rest of the chapter. Before moving on, let us here clarify the role that these examples are intended to play in our argument. First of all, our general concern is to reject the anti-reductionist consequences that have been derived from the multiple realization thesis by showing that physical differences among individuals that are sufficient to speak of different types of individuals necessarily imply causal differences that can be grasped from the psychological point of view under some particular environmental circumstances, with the result that co-extensive functionally defined sub-types are constructible, permitting us to make a conservative inter-theoretic reduction. Since we have already defended these claims on conceptual grounds, we are working here to provide concrete examples to support our arguments, and thereby to illustrate how our proposal of a reductive strategy can be applied. Secondly, the fact that we have chosen examples of pathological cases shouldn’t be interpreted as a limitation of the applicability our strategy. Our claim is more general than arguing that one can, for any mental property type, build a sub-type, which is specific to normal individual, and a series of sub-types that are specific to certain types of neurological impairments. Instead, we claim that any specificity within the constitutive components of the mechanisms enabling any individual to fulfil the mental causal role can be used to construct a corresponding functionally defined sub-type. We are focused here on pathological cases because we currently lack the empirical data relative to differences among normal individuals to explain reductively their

Reduction of psychology to neuroscience: cases studies

253

behavioural specificities, as well as because the behavioural impairment at play due to pathology illustrates at best our proposal. Thirdly, given this illustrative aim, we shall focus solely on the features that are relevant to our illustrative purposes. Although our account in principle should comprehend any behavioural or neurological specificity of any type of individual exemplifying mental properties, we have neither the space nor the expertise to discuss extensively the literature relative to these neuropsychopathologies and the total complexity of these cases. Instead, we will simplify, with the codicil that actual cases within the current scientific literature exemplify many features outside of our account. Finally, we have also to remember that neuroscience is currently not a closed theory of the production of the behaviour by the brain, especially in the two following respects. Firstly, although our knowledge both with respect to the fundamental neurobiological mechanisms that are involved in the implementation of higher level cognitive abilities and with respect to our overview of the functional specialization of brain areas is constantly advancing, important blanks remain in our knowledge. Secondly, in spite of being able to bring out correlations between lower-level neurological properties and higher level psychological phenomena, a large part of the theoretical accounts that are supposed to explain why those correlations obtain are at best incomplete and disputed hypotheses. To put both points in terms of mechanistic explanations, we do not dispose of a global mechanistic account of the production of behaviour by the brain. In the best case, we use incomplete models of mechanisms for cognitive abilities such as memory, attention, action planning, perception, and so one. How they interact to give rise to behaviour, conscious experiences, memories, and so on, and what the precise organization of their constitutive components consist of (as well as how they are operationally organized) remain only partially answered questions. This deserves some comments. Firstly, the fact that neuroscience is currently an incomplete theory does not damage our account of epistemological reductionism, as it does not depend on discoveries of actual state of the art neuroscience but rather on an in principle reducibility established on metaphysical grounds. Ultimately motivated

254

From psychology to neuroscience

by the claim of causal completeness of physics, ontological reductionism assures any causally efficacious mental property can in principle be reductively explained; the current absence of such reductive explanations does not impair this claim. Even if it turns out that we have to acknowledge that such reductive explanations will possibly never be available (due to the complexity of the mechanisms involved), the difficulties we face here are practical rather than theoretical. After all, both the current state of the research in the neuroscience and the degree of complexity of the brain are contingent facts. As in any scientific investigation, we need some paradigmatic examples – as will be provided in this chapter – to establish the validity of reductionism. But it would be inappropriate to ask for a reduction to be carried out in each situation in order to accept epistemological reductionism. Secondly, since contemporary neuroscience formulates hypotheses rather than definitive results with respect to the mechanisms that are involved within the production of complex behavioural patterns and vulnerable, as well, to pathological failure, it can turn out that some of the empirical results we are going to discuss are disputable or even simply mistaken. However, again, this has no bearing on our account, since this one depends on the kind of explanations that are permitted to account for mental properties rather than to the particular explanation we mention here. As long as neuroscience produces reductive mechanistic explanations, our account will hold firm. To sum up, this chapter illustrates the application of our reductionist account. But the examples under discussion do not justify, by their empirical substance, our particular form of epistemological reduction under discussion. And thus, neither should weaknesses within the empirical accounts we are using be employed to delegitimate our approach. 7.1.1 Inter-hemispheric disconnection syndrome The “inter-hemispheric disconnection syndrome” picks out a set of behavioural abnormalities that appear, from the neurological point of view, after injury to or defect of the nervous fibres inter-connecting

Reduction of psychology to neuroscience: cases studies

255

both hemispheres through the corpus callosum and the anterior commissure (Kolb and Whishaw 2003: 428-30). The corpus callosum is a large medullary strip linking the neocortical areas of each hemisphere to their contralateral homologous areas, allowing thereby information exchange between the neocortex of each hemisphere. It is composed of circa 200 to 800 million commissural fibres that can be sorted into three main classes. Most of them are topographic in the sense that they connect nervous areas to their respective contralateral counterpart. A second group of projections goes to areas to which the homotopic area on the contralateral side projects; and, finally, a last group of connections has diffuse terminal distributions. The anterior commissure enables information exchanges at the level of the anterior temporal lobe, the amygdale and the paleocortex of the temporal lobe surrounding the amygdala. The syndrome of inter-hemispheric disconnection is usually a result of damages to the corpus callosum, which may be caused by tumors, congenital malformation or surgical ablation. One of the most common congenital brain malformations is the agenesis of the corpus callosum, observed in humans with a prevalence of 3-7 per 1000 individuals (Bedeschi, Bonaglia, Grasso, Pellegri, Garghentino, Battaglia, Panarisi, Di Rocco, Balottin, Bresolin, Bassi and Borgatti 2006). Patients suffering of this congenital malformation present however limited behavioural impairments, since their anterior commissure is significantly enlarged and connect also neocortical areas that are usually connected through the corpus callosum. The surgical removal of both the corpus callosum and the anterior commissure, the therapeutic commissurotomy, was effected in the early 1960s by the surgeons Philip Vogel and Joseph Bogen as an elective treatment for severe cases of epilepsy. The motivation for such a heavy and irreversible surgery lays in the interpretation of hard epileptic crises as an electro-magnetic storm spreading through all the brain from isolated sources of unrest (Purves, Augustine, Fitzpatrick, Katz, Mcmantia and Mcnamara 2002: 451) Removing the corpus callosum and the anterior commissure “helps to confine the epileptic seizure to one side and tends to preserve consciousness during an attack […] and enables the patient to take precautionary or control

256

From psychology to neuroscience

measures at onset of a seizure” (Sperry, Gazzaniga and Bogen 1969: 274). Behavioural issues of commissurotomy have been extensively studied in medicine as well as in neuroscience since its introduction. The most surprising result of this surgery is the apparent lack of change with respect to daily life of patients (Sperry, Gazzaniga and Bogen 1969: 275). One year is in general required for recovery from the surgery. Within two years, the so-called split-brain patients are able to return to school or go to work to such an extent that a conventional medical examination cannot reveal anything extraordinary in the behaviour of these patients (Kolb and Whishaw 2003: 433). Indeed, one needs specific cognitive tests to detect differences between commissurotomized patients and healthy subjects from the behavioural point of view. Although in the split-brain, each hemisphere processes separately one half of the information processed by the brain as a whole, in the behaviours of ordinary life, this can be hard to detect. Standard tests allowing discriminating between split-brain and normal patients aim to determine whether sensory information presented to only one hemisphere is at disposal of both or only of one hemisphere for motor action. For instance, a very simple test consists in asking a blindfolded individual to touch an object with the left hand and to find a similar object in a hidden collection with the right hand (Kolb and Whishaw 2003: 437). As the left hand is under the somesthesic and motor control of the lone right hemisphere and viceversa for the right hand, in the absence of inter-hemispheric communication, split-brain patients are unable to match the objects correctly. Although we are not interested in describing all the tests designed to capture behavioural specificities of split-brain patients, let us consider in more detail the case of visual perception and recognition of objects. The connectivity of the human visual system is crossed in a very specific manner. Roughly speaking, visual information of the left visual hemifield is computed by the right hemisphere and vice versa. Of course, since this crossing concerns the visual hemifield and not the eyes themselves, the structure of the central visual pathway is quite sophisticated. Optical signals coming from the left visual field are

Reduction of psychology to neuroscience: cases studies

257

received by the right half part of the retina of each eye. Both of them are connected to the right lateral geniculate nucleus through the optic nerves and the optic chiasm. From there, the information is transmitted by the optic radiations to the layer IV of the right visual cortex, namely the right striate cortex. However, in the case of normal agents, information stemming from a specific half visual field is available to both hemispheres by way of two mechanisms. First of all, commissural fibres connect homologous areas of the left and the right hemispheres. Secondly, eyes are perpetually scanning by making micro-movements. These unconscious micro-movements have the effect of distributing a large part of the information stemming from the centre of the visual field to both sides of the visual system, which would otherwise be confined to a specific hemisphere. The existence of these scanning micromovements explains why split-brain patients are able to live quite normally as regards their visual abilities in spite of the lack of communication between both hemispheres. Since information stemming from the central part of the visual field is sent to each side of the visual cortex, left sided visual information can for instance be employed to form a verbal report, although the main speech area is principally left-sided. The scanning micro-movements are a natural and very powerful mechanism to allow split-brain patients to partially compensate for their deficit, i.e. the lack of communication between hemispheres in the domain of visual perception. In the case of visual perception, the classical tests go through the following steps (Sperry, Gazzaniga and Bogen 1969: 275-77). The subject is seated at a table and faces an adjustable screen that prevents her from seeing her hands. Images are flashed on one or the other half of the screen whilst the subject is asked to fix a red dot in the central part of the screen. Images are flashed during circa 1/10 seconds to prevent the scanning micro-movements of the eyes from distributing visual information to both hemispheres. Visual material can in that manner be presented selectively to one specific or both hemispheres. Objects are placed on the table in the back of the screen in such a way that the patient can be asked to identify them just by touching them. By

258

From psychology to neuroscience

these means, the experimenter seeks to determine which visual information is at disposal of which hemisphere for motor action. In the execution of such test required tasks, commissurotomized patients appear to be essentially normal with respect to the right half of the visual field. They can describe, read and use visual material in the same manner as before the surgery. However, if an image is flashed on the left side of the screen in a randomized sequence of stimulations, verbal reports of patients tend to show that they do not perceive the stimulus. If the time duration of the presentation of the stimulus is increased, patients are able to report what is presented in the left visual hemifield, through the operation of the compensation mechanism. It should be emphasized that commissurotomized patients do not recover their pre-surgery abilities in carrying out such tasks. Such behavioural consequences of commissurotomy seem to be definitive. Looking at the permanent behavioural effects of the commissurotomy, these results could at first glance also be taken to indicate a defect in the right visual system of the patients. But further tests exclude this possibility: when non-verbal responses are employed to evaluate patients’ perceptions, we can be sure that they understand the task and that they perceive and identify the stimulus. For example, if an image of an object is flashed on the left visual hemi-field, in spite of the pathological incapacity of verbally reporting the stimulus, patients are nonetheless able to pick out by touch with the left hand a corresponding object in a collection of other items. Further tests demonstrate that the right-sided visual abilities of these patients are clearly unimpaired by the commissurotomy. A defect in verbal abilities is as well excluded since the subjects can produce complex verbal reports about right-sided stimulations. The cause of the disconnection syndrome appears therefore not to be a defect in primary sensory areas and the motor cortical area, but a defect within the information processing between both of them and more specifically to be the consequence of a lack of communication between both hemispheres. Let us sum this section up by emphasizing that a large variety of such visual tests confirm the fact that visual information stemming exclusively from the left or the right half of the visual field is, in the case of commissurotomized patients processed separately in each

Reduction of psychology to neuroscience: cases studies

259

hemisphere. Normal interaction between elements within each half field is preserved, but split-brain patients cannot integrate information of both halves of the visual field. However, basic override mechanisms allow these patients to lead a normal life, in spite of the behaviour differences picked out by specific cognitive tests. Using this case as example, the next section is intended to make clear how the sub-type strategy of reduction can be applied to the relation between psychology and neuropsychology. 7.1.2 Application of the sub-types strategy Let us introduce the following example to get the application of the strategy of reduction by means of functionally defined sub-types started. Take the case of two pairs of individuals watching TV and discussing the content of a very attractive film. Each individual can be described in psychological terms as perceiving the TV screen and, from that point of view, each of them exemplifies the same perceptual mental property. The first step of functional reduction consists in the functionalization of that mental property. In a very simplified way, the considered property can be functionalized as follows: (A) A subject S visually perceives an object x if, when she looks conspicuously at x and when she desires to give a verbal description of x, S verbally describes x in a successful manner. There are of course many other cases of perception, as common sense ascribes this type of mental state in a lot of other scenarios. For example, verbal report is not a necessary condition for perception. But our purpose here is not to provide an extensive functional definition of perception. What matters here is that the general form of this definition matches the requirements of common-sense functionalism, namely to specify the typical causes and behavioural effects of a mental property and how tokens of this type of mental property are usually related to other mental tokens of other types. As argued, a generic formulation of such functional descriptions can be given in the classical RamseyLewis shape of theoretical concepts (4.2.2 Common-sense functionalism pursued):

260

From psychology to neuroscience y has mental property Φ if and only if (∃x1)(∃x2)(∃x3)…(∃xn) [T((x1 is caused by environmental circumstances c1 ∨ c2 ∨ … ∨ cn) & (x1 stands in relation to x2, x3…xn such that x1R1x2 ∧ x1R2x3, ∧ … ∧ x1Rn-1xn) & (x1 causes behaviour b1 ∨ b2 ∨ … ∨ bn) & y has x1)]

The second step of functional reduction consists in the identification of physical truthmakers of (A). Let us now assume that one of the individuals in each pair is a split-brain patient, having undergone a surgical ablation of both the corpus callosum and the anterior commissure. Nonetheless, (A) applies to both individuals in each pair. As explained, split-brain patients, once having recovered of their surgery, behave in daily life like any normal subject, with that result that they are perfectly able to verify the preceding definition. Now taking into account the fact that one individual in each pair is characterised by a lack of inter-hemispheric nervous connections, we face here a clear case of multiple realization, since one and the same functional description applies to individuals that come under two distinct neurological descriptions. As we have seen in the previous chapter, the inter-theoretical difficulty raised by the multiple realization thesis is met, in our strategy of functional reduction, by introducing functionally defined sub-types to achieve co-extensionality between mental and lower-level types. These psychological sub-types are always constructed relative to the environmental circumstances in which different physical entities, here exemplified as different neurological types, fall under the scope of the same mental types and even as such are distinguishable from the psychological point of view. In order to do this, let us suppose a situation in which one of both pairs of individuals is watching a perfectly normal movie while the second pair is watching exactly the same movie, except that in their version, the filmic sequences contain very short left-sided apparitions of a given item. Against the background of our explanations with respect to the information processing by split-brain patients, the choice of the additional features characterizing the movie that is watched by the second pair of individuals of our example is obviously designed to

Reduction of psychology to neuroscience: cases studies

261

enable the construction of psychological sub-types, since it captures precisely the critical circumstances in which the behavioural specificities of these patients are observable from a psychological point of view. Observing the first pair, we note that both the normal and the split-brain person behave in exactly the same way. Both of them perceive and comment on the movie in exactly the same manner, thereby coming under the type (A). However, observing the second pair, we should note the behavioural capacities of both individuals differ markedly, even though they come under the type (A) as well. Within this pair, the normal subject perceives and comments the movie, being able to describe left-sided and short apparitions that characterize the version of the film they are watching. The split-brain patient of that pair, however, though she can perceive and comment most of the features of the movie, fails to be able to describe left-sided and short apparitions. Such behavioural differences illustrate our point, that functional specificities that may not appear in certain “normal” situations may appear instead when the grain of functional description of psychology is made finer by taking into account the way in which the different realizers of a given mental property behave in certain particular environmental circumstances. On that basis, it is therefore possible to construct functionally defined sub-types of the target psychological type by including these behavioural specificities under certain contexts to our original definition (A). We thus get to the formulation of two new functional descriptions: (A’) A subject S visually perceives a movie if, when she looks conspicuously at the movie and when she desires to give a verbal description of it, S verbally describes the movie in a successful manner including possibly the description of left-sided apparitions of less than 1/10 second. (A’’) A subject S visually perceives a movie if, when she looks conspicuously at the movie and when she desires to give a verbal description of it, S verbally describes the movie in a successful manner excluding systematically the description of left-sided apparitions of less than 1/10 second.

262

From psychology to neuroscience

The starting point was that the coarse-grained functional description (A) applies to all the individuals considered in our example. However, the more fine-grained description (A’) applies in the second pair only to the normal individual, since the split-brain patient is unable to verbally report the short and left-sided apparitions. Consequently, the description (A’’) by contrast to (A’) applies to her. At this point, the simple bifurcation presented in our example backs up the theoretical premises with which we have been working to allow us to obtain co-extensionality between descriptions of different theories, as required by epistemological reductionism in general. Adding functional specifications to the general description at the outset has had the consequence of reducing the extension of that description. Of course, description (A’’) does not apply just to split-brain patients watching TV. There could be several other cases and reasons in which an individual is unable to notice the brief and left-sided apparitions; nevertheless, we would always be able to introduce some further specifications of the functional description (A’’) in order to narrow down its extension to pick out the difference in sense processing exhibited by the split brained subject by creating even more finegrained distinctions. For example, having distinct memories and motor abilities (Kolb and Whishaw, 2003: 433), the split-brain patient of the second pair is able to draw the object of the mentioned apparitions or to pick out a similar object in a hidden collection with the left hand, because this hand is under control of the right hemisphere, which governs visual information processing relative to appearances. Sharpening in that manner the grain of our functional descriptions, we can capture precisely the behavioural particularities of a split-brain patient, namely e.g. the inability to successfully match sensory information originating solely from one side of the body with similar information originating from the other side. In other words, introducing functionally defined sub-types for both normal and split-brain patients enables us to reach the required condition for inter-theoretic reduction, namely co-extensionality, and thus rely on bi-conditional bridging principles. This enables the reduction in the Nagelian sense of the coextensional psychological sub-types (A’) and (A’’) to neurological types under which the individual in the second pair comes.

Reduction of psychology to neuroscience: cases studies

263

Thirdly, against the background of ontological reductionism, neuroscience is in principle able to provide a reductive mechanistic explanation accounting for why the individuals of our example present the particular behavioural dispositions they do, which is sufficient to secure the third step of functional reduction. Of course, we did not present such an explanation and it is not the case that neuroscience is currently able to provide a complete explanation of the production of the behaviour, taking into account the interactions obtaining between the fundamental components of the nervous system. However, neuroscience is already able to build sketches of explanations of that type and, against the precisions provided in the preceding sub-section, the behavioural specificities of split-brain patients, which we used here to build sub-types, are perfectly understood. Thereby, neuroscience is in principle able to explain why the split-brain patient of the second pair falls both under the descriptions (A’) and (A’’). Finally, this example enables us to illustrate the relation of abstraction as it links together psychological abstract types and their functional sub-types, which accounts, for the ability of psychology to provide homogeneous explanations of neurological entities that fail to be both compositionally and causally identical, to which neuroscience, in consequence, accords a heterogeneous explanation. As our example shows, functional description (A) is retained as a perfectly valid and well grounded higher level and abstract generalization by preserving functional specifications shared by both definitions (A’) and (A’’). As we have remarked in Chapter 6, the choice of the degree of focus of a description within a given theory is a matter of conceptual leeway (6.3.3 Relation between psychological types and sub-types). Psychology, for instance, can rationally take into account or ignore certain functional details in order to bring out commonalities shared by smaller or larger group of entities. In the case of our movie watching pairs, psychology is able to retrieve the original psychological abstract type, and thereby to focus on what is shared by the four individuals, by bracketing the contextual behavioural specificities distinguishing the split-brain subjects from the other ones. By ignoring the details relative to the brief and left-sided apparition, psychology is able to group together individuals that are prima facie different in homogeneous

264

From psychology to neuroscience

treatments of their respective behaviours. The definition (A) stands thus in a relation of abstraction with respect to definitions (A’) and (A’’). The work of abstraction, here, consists in pruning those functional specifications that distinguish between the many realizers of a given higher level property. When we introduce functional sub-types, we are really making room for a reductionist but conservative position with respect to the higher level generalizations of psychology, since the commonalities shared by the different realizers of a psychological property are completely characterized in a functional manner. To sum up, this example shows how it is possible in this case to straddle the seeming paradox presented by the multiple realization thesis: we need only to build sharpened functional sub-types to grasp the behavioural specificities of certain types of individuals to obtain the co-extensionality that is so important to epistemological reductionism. In the process, we equip ourselves to provide homogeneous reductive explanations of the mental properties of such disparately endowed individuals. Moreover, the possibility of varying the grain of functional description enables us to handle multiple realization as an intratheoretic issue that does not threaten the scientific legitimacy of psychological descriptions.

7.2 Finding critical conditions of manifestations We have now presented a first and rather simple example that shows the gain of functionally defining sub-types as we vary environmental factors to bring out behavioural differences in differently endowed individuals. The case of commissurotomized patients is ideal for the two following reasons. Firstly, the behavioural consequences of commissurotomy are simple to observe and well understood, relative to our knowledge of brain connectivity. Secondly, the neural specificities of split-brain patients are relatively large scale specificities and it seems from that point of view not that surprising that folk psychology is able to detect the corresponding behavioural specificities characterizing these patients. We now intend to present some general considerations with respect to the methodology of neuropsychology, suggesting that our proposal holds for any other comparable type of neurological impairment, after which we will take some illustrations

Reduction of psychology to neuroscience: cases studies

265

from some ingenious experimental designs presented within current research. These examples will show how, in cases in which the empirical data seem to be prima facie too diffuse, we can vary environments to pick out behavioural impairments and trace them back to their neurological causes. 7.2.1 Neuropsychology and experimental design Let us here start simply by having a look at the definition of neuropsychology provided by Kolb and Whishaw, namely the “study of the relation between human brain function and behaviour” whose “central focus is the development of a science of human behaviour based on the function of the human brain” (2003: 2). Accordingly, neuropsychology is nothing but the project of building a theory of the production of the human behaviour on the basis of the mechanism of the central nervous system. Obviously, the very idea of such a project only makes sense if we assume that there is a systematic relation between brain and behaviour, and, thereby, between neural specificities and behavioural specificities. From that point of view, neuropsychology perfectly fits within the overall project of neuroscience, which is to produce a mechanistic account of the production of human behaviour on the basis of the functioning of the brain. After all, neuropsychology deals with brain components that are functionally organized in order to cause human behaviour. These components are individuated by their causal contribution to behaviour, indicating that they are functionally individuated by the causal roles they fulfil within the brain architecture, which in turn causes behaviour as a whole. However, through its focus on behaviour, neuropsychology has a narrower field of investigation than neuroscience in general, since it always takes behaviour as its standard feedback for theorizing about cognitive mechanisms, while neuroscience is often concerned with many other levels of organic functioning. Neuroscience in general aims to uncover the organization and the activities of brain components by means of which behaviour is produced, which spans levels all the way down, theoretically, to microphysics (4.3.3.1 Mechanistic explanation as reductive explanations). The example of explanation of spatial

266

From psychology to neuroscience

memory highlights the iterative character of mechanistic explanations. The phenomena of spatial memory is explained by means of hippocampal structures, whose functioning is in turn explained by means of electrico-physiological properties of hippocampal neurons, and so on, potentially down to fundamental physical property tokens. Neuropsychology has a more restricted domain of investigation. It takes into account brain components mechanistically considered in relation to brain functioning only to the extent that their causal contribution to the behaviour is publicly observable in some environment – as we saw with our example of split brains. The criterion it depends on is that of mutual manipulability, as discussed in our fourth chapter (4.3.2.1 Mechanisms and mechanistic explanations). This obviously distinguishes neuropsychology from other domains of neuroscience such as neurobiology, which, as much as it is concerned with mechanisms, has different causal parameters – for instance, the interplay of the neuron components in the neuronal axon (their molecular cues) without direct reference to any kind of overt behaviour. This is why neuropsychology, in as much as it links behaviour and brain components directly by means of mechanistic explanations, is a field of first interest for our agenda. After all, the criterion for the constitutive relevance of the mechanistic framework imposes an order such that variations among the components of two occurrences of mechanisms, which count thereby as two occurrences of two different types of mechanisms, have an impact on the way in which the phenomena for which they are mechanisms is brought about. Note, again, that the causal interplay between the different constitutive components c1,…, c2 of a mechanism M for a phenomenon P, functionally defined as a causal role R, has to account for every feature of R. Provided that the constitutiveness of components is established by mutual manipulability of the components and the phenomenon P, it follows that if a component ci in a mechanism M is changed or altered, then it has to have an impact on R that is exhibited in some possible environment. If it is not the case, then ci is simply not a constitutive component of M for P. Reversely, engaging in a certain behavioural

Reduction of psychology to neuroscience: cases studies

267

task requiring the execution of the target causal role R must induce changes within the activity of the constitutive components. An example of the scope and limits of neuropsychological investigation is found in the vital issue of the localization of cognitive functions. In order to investigate the actual components that are involved in the production of particular behavioural tasks, neuropsychologists design protocols of inter-level experiments according to the four schemata discussed in Chapter 4 (4.3.2.1 Mechanisms and mechanistic explanations). These experimental setups are designed to maximize the observability of the relation of mutual dependency between the execution of the task and the activity of the brain components, i.e. to create a statistically significant contrast at the behavioural level between control subjects and individuals presenting neurological specificities precisely to highlight the causal contribution of what distinguishes them from the neurological point of view. In other words, they contrast a correlation, on the one hand, between the level of performance presented by control subjects in the execution of a given cognitive task and their intact neurological structures; and, on the other hand, a correlation between the production of behavioural particularities exhibited under experimental conditions by subjects executing the same task and their neurological particularities. The central of part of the neuropsychologists’ job consists therefore in two mutually dependent tasks: making hypotheses about the cognitive functions that together produce human behaviour and designing experimental set ups that enable us to test the validity of these hypotheses. If we consider the production of behaviour as the overall function of the mind, the first task of hypothesis amounts to making a space in which we can highlight choices between different possible models of cognition that are suggested by functional analysis. We already saw that functional analysis à la Cummins enables us to draw up different possible models of mechanisms for achieving cognition, the crucial issue being to know which of them is as close as possible to the reality of cognition. To make a good hypothesis, as we can tell from the history of neuroscience, is not an easy matter. That history shows countless

268

From psychology to neuroscience

examples of mistaken taxonomies of cognitive functions. Take, for example, phrenology, stemming from the hypotheses of Gall and Spurzheim, who, on the one hand, had a firm grasp of one idea that came to be integrated into modern neuroscience, viz, the localization of brain function, but who, on the other hand, produced a taxonomy that was grossly culturally skewed and had no physiological footing. They located no less than twenty-seven faculties under the surface of the skull at the beginning of the XIXth century by comparing skulls of individuals presenting salient traits of personality. Among these faculties were represented behavioural traits such as language, but also exotic items such as cautiousness, hope or veneration (Kolb and Whishaw 2003: 8-9). Although their taxonomy was gradually abandoned over the XIXth century, their insight into the division of cerebral labour, with functions allotted to different locations in the brain, served as a foundational attempt to correlate behavioural specificities with brain particularities from a less culturally biased, more physiologically informed point of view. The localization of brain functions could not have come about without experiment. Designing good experimental set ups is, evidently, the second step after hypothesis and aims to provide empirical evidence supporting these hypotheses. Experimental protocols have therefore to fulfil the following constraints. Firstly, they enable us to study cognitive abilities to the extent that experimental tasks executed by subjects in the experiment make conspicuous the target cognitive ability, limiting the influence and interference of other cognitive functions. Secondly, the correlation of behavioural specificities in the execution of a particular cognitive task with a neural specificity at the level of the hypothesised components of the mechanism responsible for the execution of the function under investigation is not sufficient to secure that a relevant cognitive function and its locus of execution have been isolated. Good experimental practice requires comparison with other sets of subjects differing neurologically with respect to the constitution of the investigated mechanisms. In this way, we are justified in our observation of the relation between behavioural results and neurological results, which are mutually dependent. Only the

Reduction of psychology to neuroscience: cases studies

269

condition of mutual dependency allows us to consider that a genuine cognitive function and its location have been discovered. The localization of cognitive functions has been pursued for a long time by comparing the behavioural abilities of healthy patients with pathological cases through a post-mortem investigation of the brain structures of the latter. The history of neuroscience displays several examples of this procedure, one of the most famous being the works by Broca and Wernicke on language that uncovered the standard double dissociation between the lexical function of Wernicke’s area and the syntactical function of Broca’s area. Damages to the Wernicke’s area cause impairment in language understanding and in the production of meaningful sequences of words. Damage to the Broca area tends to cause impairments with respect to the syntax and the fluency of speech and difficulties in the understanding of grammatically complex sentences. Such dissociations remain the basis for the localization of cognitive functions. Post mortem dissection has now been supplemented by the recent development of modern neuroimaging, which allows for localization in real time for living subjects. Neuroimaging enables scientists to observe directly the neural activity and, in that manner, to establish a correlation between a specific result in the execution of a certain behavioural task and a particular pattern of neural activation. This holds out the possibility of working with normal, rather than impaired, subjects, to understand the mutual dependence of neural mechanisms and higher level phenomena. At this point, it is clear that this two-step methodology perfectly fits the mechanistic framework developed in Chapter 4, since it proceeds by correlating bi-conditionally behavioural specificities with neurological specificities, which obviously matches the criteria of mutual manipulability between a phenomenon and the different components of the mechanism for that phenomenon. Unlike other domains of neuroscience, neuropsychology is focused on data generated by comparing observable human behaviour and the mechanisms of the central nervous system in order to develop a model in which the latter operates as the mechanism bringing about the former. The experimental environment in which the behaviour is exhibited is aimed at showing the mutual dependence of behaviour and

270

From psychology to neuroscience

neural mechanism. Now, given this strong link between the relevant brain’s components and behaviour, it is clear that neuropsychology is interested in neurological specificities that have observable consequences on the overt behaviour. The exhibition of behaviour is the variable upon which experiment design depends, for the whole point is to obtain data showing a stable pattern of behaviour depending on the activation of certain very precisely designated neural components. In the split-brain case, the behavioural effects of the commissurotomy were extremely well delimited using pretty simple experimental conditions highlighting the cognitive function of the inter-hemispheric projections. However, we can imagine an argument, at this stage, to the effect that other cases of neuropsychopathologies have behavioural effects that are too diffuse in order to be clearly graspable using neuropsychological experimental set-up. We aim therefore in the next section to examine an exemplary complex neuropsychopathology, schizophrenia, to show that under the appropriate environmental conditions, i.e. experimental set-ups, we can distinguish the normal and schizophrenic subjects from a purely behavioural point of view, in spite of the apparent lack of clarity of the ordinary behavioural data. The upshot here is that even in case of pathologies of which the behavioural effects are not as isolatable as in the case of the disconnection syndrome, it is nonetheless possible to trace individual behavioural differences back to physical differences even from a psychological point of view, which can registers certain otherwise diffuse behavioural specificities under precise environmental conditions. 7.2.2 Example of complex neuropsychopathology: Schneiderian schizophrenia Schneiderian schizophrenia, also known as type I schizophrenia, is certainly among the most fascinating of neuropsychopathologies. In the current state of neuroscience, both the neurological origins of

Reduction of psychology to neuroscience: cases studies

271

schizophrenia and its behavioural symptoms are under dispute.63 On the behavioural side, Schneiderian schizophrenia is supposed to be essentially characterized by symptoms such as delusions and hallucination (Squire, Zigmond, Bloom, Landis and Roberts 2003: 1192), but also thought insertions, feeling of other’s influence on the patient’s though, actions and emotions (Farrer, Franck, Frith, Decety, Georgieff, d'Amato and Jeannerod 2004: 31). On the neurological side, type I schizophrenia is hypothesized to results from a dopaminergic dysfunction by reason of the responsiveness of patients to neuroleptic drugs (Kolb and Whishaw 2003: 726). However, establishing a causeto-effect relationship here is still challenging and we still lack of why the neuroleptic drugs work, what causes the abnormal functioning of dopamine synapses and whether the latter alone, or even always, is the cause of the behavioural impairment characterizing these patients (Squire, Zigmond, Bloom, Landis and Roberts 2003: 1192). Moreover, schizophrenic patients often present other neurological abnormalities, making the correlate(s) of schizophrenia difficult to isolate (Squire, Zigmond, Bloom, Landis and Roberts 2003: 1191-94). Although the biochemical and structural abnormalities characterizing schizophrenic patients’ brain are not so well understood, recent developments in the identification of the neural mechanism of recognition of self-generated actions have suggested new hypotheses with respect to the neurological correlate of type I schizophrenia. From the cognitive point of view, the central monitoring hypothesis constitutes the current general framework for understanding the mechanism of recognition of self-generated actions. Roughly, the idea is that “each time the motor centres generate an outflow signal for producing a movement, a copy of this command (“the efference copy”) is retained. The reafferent inflow signals are compared with the copy. If a mismatch arises between the two types of signals, new commands are generated until the actual outcome of the movement corresponds to the desired movement” (Jeannerod 2003: 8). Self-generated actions 63

Kolb and Whishaw insist in the fact that “one universally accepted criterion for diagnosing schizophrenia is by eliminating the presence of other neurological disturbance or affective disorder – a definition by default” (2003: 726).

272

From psychology to neuroscience

recognition is based in this framework on the concordance between a desired action and its predicted sensory consequences. According to Farrer et al. (2004: 31), defectiveness of these mechanisms leads to the cognitive impairments characterizing Schniederian schizophrenia. For example, hallucination are explained within the framework of the central monitoring hypothesis as a pathological incapacity of the patients to recognise self-generated actions as their own (Franck, Farrer, Georgieff, Marie-Cardine, Daléry, D'Amato and Jeannerod 2001). In the case of auditory hallucinations, the motor areas are engaged offline in order to generate an inner linguistic representation (the inner speech), but the auditory cortex interprets this activity as if it were produced by an external cause (Jeannerod 2003: 14). From this perspective, schizophrenia would be a pathology affecting the communication and the comparison of information between motors and sensory area. In order to investigate the localization of the mechanisms underlying the phenomena of self-generated action recognition, many successive experimental designs have been set up (for a review, see Jeannerod 2003). Farrer et al. (2004) designed an experiment whose central idea was to develop a task requiring of the subject to mobilise so far as nothing but the execution of the investigated cognitive function, namely to compare a motor command and its intended result. In this experiment, healthy and pathologically diagnosed subjects watched images displayed on a screen and reported whether they were effects of their own manipulations of an out-of-eyesight joystick. The experimenter was able to vary the degree of discrepancy between patient’s inputs and visual output on the screen. Without entering in more details with respect to this experiment let us highlight two points. First of all, brain imagery study of healthy subjects enables neuroscientists to correlate a task execution with a delimited pattern of neural activity. In the case of the studies of Farrer et al., a positive correlation has been reported between the degree of control by the subjects of the perceived movement on the screen and the activation of the right angular gyrus (2003: 332; 2004: 37), particularly of Brodman area 39 and 40 on the right side (reported in Jeannerod 2003: 9-10). Maximal activation occurred when the movements shown to the

Reduction of psychology to neuroscience: cases studies

273

subject were unrelated to the subject’s own movements. The degree of control appeared also to be reversely correlated with the degree of activity of the right posterior insula (Farrer, Frank, Georgieff, Frith, Decety and Jeannerod 2003: 328-29). However, these correlations do not establish beyond doubt that the considered areas are causally responsible for the correct task execution. Only when we obtain results from healthy subjects and compare them with results of impaired patients do we have grounds to infer the relation between patterns of neural activity and behavioural patterns. In other words, we are looking for evidence of counterfactual dependency, evidence which makes it more plausible to consider the localized neural activity as a genuine cause of the observable behavioural particularity. Farrer’s research group obtained these results. Schizophrenic patients perform the experimental task very poorly, showing that they encounter difficulties in distinguishing selfgenerated movements from alien generated movements, while exhibiting an aberrant activity of the right angular gyrus and an absence of variation of the posterior insula activity in the very same experimental conditions (Farrer, Franck, Frith, Decety, Georgieff, d'Amato and Jeannerod 2004: 41-42). This result confirms the hypothesis that the right angular gyrus and the right posterior insula are the locus of the comparison between efferent copy and reafferent signal and thereby, in conjugation with the central monitoring hypothesis, they contribute to this explanation of the Schneiderian schizophrenia disease. Hallucinations and delusions are understood in terms of abnormal interactions between different cortical areas. This dysfunctional integration of information is explained at the cognitive level, in the framework of the central monitoring theory, as a failure to integrate perception and action; at the physiological level, it is considered to be a defect in connectivity. This abnormal connectivity “disrupts the modulation by frontal region of those more posterior brain areas involved in the processing of the sensory consequences of an action” (Farrer, Franck, Frith, Decety, Georgieff, d'Amato and Jeannerod 2004: 41), making it difficult to identify the source of the perception as internal or external. The incriminated absence of modulation then can be explained as the cause of the fact that the

274

From psychology to neuroscience

subject processes the sensory consequence of action as if it were the result of an external cause (Blakemore, Smith, Steel, Johnston and Frith 2000; Frith and Dolan 1996).64 Let’s return to the functionally defined sub-types strategy by formulating the following remarks with respect to our present example. First of all, a research programme like Farrer’s relies on the possibility of creating testing conditions in which a certain targeted behaviour is made conspicuous, thus providing an empirical feedback with respect to the hypothesis about the cognitive sub-functions that together compute the studied cognitive ability: in this case, the set of subfunctions taken for granted by the central monitoring hypothesis, which is supposed to account for the ability of recognising self-generated actions. As said above, this possibility is open ipso facto in cases of schizophrenia and other neuropsychopathologies, since neuropsychology focuses the explanation of the normal and impaired behaviour on the basis of the functioning of the brain normal or abnormal functioning. Second, the cognitive tests used by neuropsychologists are intended to establish a contrast between normal and pathological functioning at the physiological level that mirrors the results obtained

64

The debate is here significantly simplified, since our first interest is not the mechanisms involved in the recognition of self-generated actions but rather the role played by experimental set-ups in the study of complex neuropsychopathologies. Given our illustrative concerns and the fact that the alternative accounts are built on the same kind of explanatory schemes, our presentation should be sufficient. Note however that some authors (for instance Synofzik, Thier, Leube, Schlotterbeck and Lindner 2010) stress the role played by the imprecision manifested by schizophrenic patients in the prediction of the sensory consequences of their own actions. Similarly, Jeannerod and Pacherie (2004) emphasize the role of actions simulation in these mechanisms and their pathological functioning in case of schizophrenia. Proust (2006) argues that the comparator is under attentional influence, with the result that abnormalities in agency may result from a metacognitive impairment of the capacity to store intentions of action and to inform the comparator about them. Finally, other authors, for instance Synofzik et al. (2008), claim that the comparator model is inadequate to account for the sense of agency and ownership for thoughts.

Reduction of psychology to neuroscience: cases studies

275

at the behavioural level. As the case of schizophrenia shows, even in cases where both the behavioural impairments and their neural correlates are difficult to isolate, we can isolate and test behaviours by designing conditions in which they become conspicuous, while at the same time monitoring brain activity. Indeed, the schizophrenic patients selected for Farrer’s et al. study did not manifest permanently first rank symptoms of Schneiderian schizophrenia. Rather, they were selected for the presence of such symptoms during the weeks preceding the experiments (Farrer, Franck, Frith, Decety, Georgieff, d'Amato and Jeannerod 2004: 33). However, as the study shows, it is nonetheless possible to design an experiment that could reliably highlight the targeted behavioural impairment by putting these subjects under testing conditions designed to make their behavioural specificities conspicuous. These critical test conditions are precisely the kind required in order to construct functionally defined sub-types. After all, what is required from the behavioural point of view for the application of the strategy of reduction by means of functionally defined sub-types is the possibility of distinguishing individuals that fulfil causal roles which are sufficiently similar to fall under the same psychological description, using appropriate environmental circumstances. These circumstances are designed to highlight differences in the way these individuals fulfil this causal role; the differences that are thus made conspicuous must be salient from the psychological point of view. Farrer’s experiments, in which healthy and pathological subjects are clearly distinguished by means of the way they are executing the very same task, satisfies these requirements, giving rise to two distinct sets of results. Since the results of these tests are graspable from a behavioural point of view, they can be employed in order to make the mental functional definitions more precise, as explained in the case of split-brain patients. Of course, the detailed functional reports at issue in the case of schizophrenia will be much more complex. However, this is a practical, not a theoretical, problem. That we only dispose of a sketchy account of the means by which we are able to distinguish between self and alien generated action does not impair our account of reduction. For, thirdly, as we have argued,

276

From psychology to neuroscience

the point is to highlight the fact that the multiple realization thesis gives us no reason in principle to doubt the reducibility of psychological descriptions. The theoretical availability of a mechanistic and reductive account is backgrounded by ontological reductionism, while its empirical possibility is a contingent question depending on the state of the art in neuropsychological research at any one time. The trend of contingent discoveries points to the programmatic structure guiding our empirical research, and making true the claim that science is progressing in the quest to account for the neural causes of neuropsychopathologies such as schizophrenia. Thus, the facts that the sketch of account we provided here does neither explain the origins of the abnormal functioning of both the right angular gyrus and the posterior insula, nor does it make any link with the dopaminergic abnormalities characterizing schizophrenic brains, doesn’t pose any problem for our account, which is the central focus of our developments. Fourth, our considerations with respect to the object of neuropsychology associated with the observation that it is possible to construct psychological sub-types for complex cases of neuropsychopathologies such as schizophrenia should definitively convince the reader that there is no objection in principle to apply the procedure to any neuropsychological pathology. In accordance with our principle that differences of composition eventually come out in some environment, it is always possible in principle to overtly individuate pathological effects from the behavioural point of view. Hence, it is in principle possible to introduce a functionally defined sub-type of a given mental type for each type of neuropsychologically impaired patient, that is, for normal individuals in contrast to split-brain or schizophrenic patients, but also for patients suffering from unilateral neglect syndromes, blind-sighted patients, and so on. What is required is just to sharpen in the following way the functional reports, using the behavioural testing tools of neuropsychology: y has mental property Φ if and only if (∃x1)(∃x2)(∃x3)…(∃xn) [T((x1 is caused by environmental circumstances c1 ∨ c2 ∨ … ∨ cn) & (x1 stands in relation to x2, x3…xn such that x1R1x2 ∧ x1R2x3, ∧ … ∧ x1Rn-1xn) & (x1 causes behaviour b1 ∨ b2 ∨

Reduction of psychology to neuroscience: cases studies

277

… ∨ bn) & y has x1)] + further descriptions of actual causes + further descriptions of mental states implicated + further descriptions of actual behavioural effects of x. This sharpening is possible because all neuropsychological pathologies induce specific behavioural effects that neuropsychology can exploit in order to build a global theory of the production of behaviour by the brain. The requirements of the reduction strategy by means of functionally defined sub-types are thus met. The possibility of constructing sub-types exists as well for complex neuropathologies, with the result that one can obtain fine-grained psychological descriptions that are coextensive with neuropsychological descriptions. At this point, it should be clear that the requirements for the applicability of the sub-type strategy to the relation between psychology and neuropsychology are met. On the one hand, it is in principle possible to build sub-types of our ordinary mental types that are co-extensional with the physical descriptions of individuals falling under different types from the point of view of neuropsychology. On the other hand, neuropsychology makes progress by using the method of building reductive explanations of the production of the behaviour by the brain, presupposing ontological reductionism to assure the in principle availability of such explanations.

7.3 Down to neurobiology Up to now, we have only considered examples borrowed from neuropsychology, which focuses, in reasons of its link to the overt behaviour, on the role of relatively large-scale brain components. In this last section, we will show that there is no principled objection to apply our approach down to neurobiology, and will provide some illustrations of that claim. 7.3.1 Motivated in-principle applicability Our argument in favour of the claim that psychology is in principle reducible does not contain anything suggesting that it does not apply to the fundamental level of components by means of which the human behaviour is produced. This argument runs as follows. First, according

278

From psychology to neuroscience

to the multiple realization thesis, there are individuals differing from the physical point of view who fall nonetheless within the domain of the same psychological description, i.e. they satisfy the same functional description. Secondly, against the background of ontological reductionism, it is not the case that they fulfil the same causal role tout court because physical differences imply causal differences as well. Thirdly, for any difference in the way two physically distinct individuals fulfil the causal role characterizing a certain mental property type, we can vary environmental circumstances so that we can produce corresponding behavioural differences that can be grasped by folk psychology. Fourth, since each type of physical individual verifying a certain mental description manifests idiosyncratic behavioural specificities in appropriate environmental circumstances in a psychologically salient way, it is possible to construct functionally defined sub-types of each mental type such that co-extensionality between functionally defined psychological sub-types and physical types designating entities falling within the domain of the target mental type is obtainable. Thus, there is accordingly nothing special in our argument that would justify a bottom level constraint on the applicability of our reductionist strategy. The mechanistic framework developed in the second part of the fourth chapter implies that any differences, no matter how fundamental, among constitutive components transitively imply differences at the behavioural level. The argument is the following one. Neuroscience spans multiple levels of investigation by constructing mechanistic explanations accounting for higher-level phenomena on the basis of the organization and the activities of the constitutive components of the mechanism for the target phenomenon. A constitutive component C of a mechanism M for a phenomenon P is qualified as such by the criteria of mutual manipulability, according to which the alteration of C has an impact on P and vice-versa. Now, if an entity C is qualified as a component of M for P, then this entity and its causal role in M is also mechanistically explainable by reference to the organization and the activities of the components of a mechanism M-1 for C. Provided that the criteria for constitutiveness of components applies at every level, it follows that if one intervenes to modify the

Reduction of psychology to neuroscience: cases studies

279

activity of a components C-1 of the mechanism M-1 for C, then this modification will have an impact on the activity of C, which in turn will have an impact on the phenomenon P. Given that the integration of multiple levels of mechanistic explanation proceeds by taking the activity of a component as a phenomenon to be explained by a lower level mechanism, it is clear that the mutual manipulability between phenomena and components holds in theory through the whole hierarchy of mechanisms. Behaviour is the highest level phenomenon that falls within the ambit of neuroscience, and it should in theory be the case that the manipulation of fundamental constitutive components of the brain has an impact on the behaviour itself. Of course, the claim is a theoretical one, given that it might possibly never be the case that we are actually able to observe in practice all such links through the entire hierarchy of mechanisms, due to the sheer quantitative complexity of the system. Although each synapse is supposed to be a constitutive component of our brain, their causal contribution to the behaviour remains most of the time obscure and difficult to quantify. But this is a contingent question for empirical research, not a principled objection to our claims concerning the mutual dependency of mechanism and behaviour, brought out by mutual manipulability in experimental situations. The manipulability is mutual in as much as the very criteria for constitutiveness of components is bi-directional (4.3.2.1 Mechanisms and mechanistic explanations). This implies that changes at the level of the phenomena have to have an impact on the activities of the components. If this reasoning is reported one level lower, taking the activity of one of these components to be explained by a lower level mechanism, it is clear that the changes in question also impact the activities of the components of this lower level mechanism, and so on, down to the fundamental components of the brain. Therefore, engaging in certain behavioural tasks should in principle affect the state of fundamental components. Again, we are not currently able to observe such relations of dependency, but this simply doesn’t affect the theoretical possibility held out by our argument. Note that it is precisely these practical limitations that delimit the field of investigation of neuropsychology. As explained above (7.2.1

280

From psychology to neuroscience

Neuropsychology and experimental design), neuropsychology is a subdiscipline of neuroscience concerned with isolating relations of mutual dependencies between behaviour and brain components. Although this relation of mutual dependency holds in principle through the whole hierarchy of mechanisms, our current techniques of investigation drastically limit the inquiry about the contribution of lower-level components to the production of the behaviour. While these practical limitations currently and contingently define the field of neuropsychology, the latter can and will be enlarged as we make progress in understanding the precise effect of micro-level causes. To come back to the issue pursued here, if we affirm both ontological reductionism (3.7 The token-identity thesis as ontological reductionism) and the implication of the multiple realization thesis (6.2.2 Implications for the multiple realization argument), against the mechanistic framework, this implies that even tiny variations among the physical properties contributing actively to the exercise of a certain higher level causal role have an effect on the way in which this causal role is fulfilled, independently of our ability to construct scientific theories taking into account these variations using the standard mechanistic methodology of empirical research. If now we add to that observation the claim that for any such difference between two individuals verifying the same psychological description, there is a set of environmental circumstances in which distinguishing individual attributes will be expressed in terms of the descriptive resources of psychology, it follows that our strategy of epistemological reductionism is in principle applicable down to neurobiology. 7.3.2 From molecules to behaviour: some empirical evidence To illustrate this claim of salient differences on every level pertaining to neural functioning, we refer to a series of experiments due, among others, to Bourtchouladze et al. (1994) that correlated molecular to functional differences in mice. The mechanisms underlying the process of memory consolidation, i.e. the formation of stable and long lasting

Reduction of psychology to neuroscience: cases studies

281

memories,65 were the target of Bourtchouladze’s research. Before turning to that experiment, we need to introduce some basic pieces of knowledge with respect to memory and its hypothesised neurological substrata, long-term potentiation. We have previously presented LTP as the neurobiological mechanism supporting memory (4.3.1 The scope of neuroscience). As explained, we have strong evidence in favour of the critical role played by hippocampal neurons in spatial memory. Firstly, rats with bilateral hippocampal lesions exhibit important deficits in behavioural tasks requiring the formation of a spatial representation of the environment. Secondly, the implication of hippocampal neuron, the “place cells”, has been highlighted by the in vivo recording of the electrical activity of single hippocampal neurons showing that certain neurons fire preferentially according to the location at which the rat stands in its environment. If the rat is introduced in a new environment, these neurons adapt their responses in order to fire as a function of the position of the rat in this new environment. Thirdly, the mechanism by means of which the neurons adapt their response to a particular environment is the LPT process, by means of which synapses increase the propensity of the post-synaptic neuron to respond to the firing of the pre-synaptic cell. In order to understand how the experiments conducted by Bourtchouladze et al. contribute to the understanding of the consolidation of memory, it is crucial to distinguish between two different stages within the LTP process, namely early-LTP (ELTP) and late-LTP (LLTP). ELTP is induced by single high frequency stimulation of the pres-synaptic cell and lasts for one to three hours. This relatively short synaptic potentiation relies on the interplay of two ion channels on the dendrite of the post synaptic neuron. A stimulation 65

We shall borrow most of what follows to the excellent synthesis by John Bickle (2003: 88-102). However, we shall significantly simplify the empirical discussion, since our interest is in the main result, namely the fact that blocking the expression of certain genes disrupts the chemical processes leading to the late phase of LTP, which is the crucial step leading to the consolidation of short term memories into long lasting memories.

282

From psychology to neuroscience

of the pre-synaptic cell causes the release of glutamate, which binds to post-synaptic AMPA channels with the result that Na+ ions enter the post-synaptic cell, which in turn briefly and weakly depolarizes the membrane of the post-synaptic cell. If this stimulation is repeated at a high-frequency, the membrane of the post-synaptic cell is depolarized enough to cause the Mg2+ ions blocking the NMDA receptor to be popped out, with the result that Ca2+ ions enter the post-synaptic cell. At this point, a complex interplay of biochemical components leads to a temporary increase of the conductance capacity of AMPA channels for Na+ ions, with the result that the post-synaptic cell has temporarily a higher propensity to respond to the firing of the pre-synaptic cell. At this stage, the duration of the potentiation resulting from a brief tetanisation of the presynaptic cell is well correlated with the observed duration of short-term memory, strengthening the general hypothesis that LTP is the basic mechanism accounting for memory. The stabilisation of synaptic potential for periods from hours up to weeks, month or years requires multiple tetanisation of the presynaptic cell, accounting thereby for our need of repetitions in order to establish long lasting memories. It was only during the 1990s that molecular biology started to understand properly the mechanisms undertaking the transition from ELTP to LLTP and thereby explaining how short-term memory is supposed to be consolidated into long-term memory. Given the complexity of the molecular process involved, we shall skate over the empirical data. To be brief, a repeated tetanisation of the presynaptic cell causes a cascade of chemical processes in the postsynaptic cell leading to the conversion of ATP into cAMP. In turn, this second messenger (cAMP) activates catalytic PKA subunits, which modify in the nucleus of the postsynaptic cell the expression of certain genes by interacting with a transcription factor, namely CREB proteins. These changes in the expression of certain genes result in the formation of new AMPA receptors that will take place at the level of the synapse, increasing thereby the overall conductance of the synapse in Na+ ions in cases of the weak depolarization of the pre-synaptic cell. At this point, the synapse is potentiated. Moreover, the structural changes at the level of the dendrite resulting from that process accounts for the stability over time of LLTP. Thus, the consolidation of memory

Reduction of psychology to neuroscience: cases studies

283

appears to be based on the transition from ELTP to LLTP, which requires the modification of the expression of certain genes in order to induce long-lasting structural changes at the postsynaptic level, the CREB proteins playing a crucial role in this process. Let’s preface our outline of the Bourtchouladze experiments, with two remarks. Firstly, although most of the experiments having contributed to these results have been conducted in Aplysia, Drosophila, mice and rats, the explanatory scheme drawn above extends to humans as well. For one thing, it appears that subjects with hippocampal injuries present significant memory impairments, even though the precise contribution of hippocampus to memory in the human case is still relatively unclear (Kolb and Whishaw 2003: 467). For another thing, the importance of the CREB transcription factor for the consolidation of memory is suggested in humans as well, since certain mutations of the genes interacting with CREB results in the Rubisein-taybi Syndrome, which includes, among other symptoms, important learning deficits (Petrif, Giles, Dauwerse, Saris, Hennekam, Masuno, Tommerup, Ommen, Goodman, Peters and Breuning 1995). Secondly, LTP is much more than simply the mechanism for memory generation in the hippocampus and the molecular correspondent of spatial memory. It is rather a molecular explanatory scheme that accounts in large part for the phenomenon of synaptic plasticity in general and thereby hypothetically for most human learning capacities, including explicit memory (episodic and semantic) and implicit memory (skills and habits, priming, non-associative learning and classical conditioning). However, some cases of learning, notably in Aplysia, appear to be based on an enhancement of excitability of the sensory neurons, in addition synaptic plasticity. Thus, memory seems to be mainly, but not exclusively, based on LTP. (Squire, Zigmond, Bloom, Landis and Roberts 2003: 1291). With these two codicils in place, let us turn to the Bourtchouladze experiments (1994). The Bourtchouladze group studied mice in which the gene responsible for the production of CREB proteins was partially knocked-out. These CREB- mutant mice were tested along with control mice in two different tasks of memorization. In the first, they were induced to go through the classical Morris water

284

From psychology to neuroscience

maze, in which they have to memorize where a submerged platform enabling them to rest instead of swimming is located, using certain environmental cues. After fifteen days of training at the rate of one trial per day, knocked out mice exhibited significant memory deficits in different tests aiming to evaluate their ability to find the platform in comparison to the control group. However, given the temporal extension of this experiment, a differentiation between short and long term memory is difficult to isolate, given that mice in general need several days to memorize the location of the platform. This motivated a second experiment to observe more precisely differences in the consolidation of mice memories. In this second experiment, CREB- mice and controls were put in a box for two minutes. At the end of this time, a speaker emitted a loud sound for thirty second before an electrical shock was applied to the rodent through a grid laying on the floor of the box. Right after the shock, mutants and control mice displayed similar behaviours, namely freezing. Up to thirty minutes after the shock, both groups reacted in perfectly comparable ways to a repetition of the two initial stimuli, namely to their introduction in the box and to the tone. This demonstrates that short term learning was not affected by the lack of CREB transcription factor. However, mutants showed significant memory impairment later. Both at one and at twenty four hours after the initial training, they spent much less time freezing when introduced into the box again, as well as when hearing the tone, demonstrating that the consolidation of memory was impaired by the absence of CREB. If this is so, then the absence of a single type of protein is causally linked to deficits in the phenomenon of memory consolidation and its behavioural correlates, a result that has been confirmed by several subsequent studies (for an improvement of the same experiment, see notably Abel, Nguyen, Barad, Deuel, Kandel and Bourtchouladze 1997; for a general review of the function of CREB proteins in memory, see Silva, Kogan, Frankland and Kida 1998). What is the meaning of this for our functional reductionist program? First of all, this account of memory consolidation fits the mechanistic framework. After all, both these experiments and results just extend the sketch of the account of spatial memory we borrowed

Reduction of psychology to neuroscience: cases studies

285

earlier to Craver (4.3.1 The scope of neuroscience), bringing out the chemical components that are involved in the two stage of the LTP process. The knocking out of certain genes suppresses the synthesis of CREB proteins in the mutants, with the result that a crucial component is lacking in the mechanism underlying the late phase of LTP. We are thus looking at a perfect case of an inter-level bottom-up inhibitory experiment (4.3.2.1 Mechanisms and mechanistic explanations) aiming, in this case, to evaluate the role played by CREB protein with the organization of the different components accounting for LLTP and, consequently, for the higher level phenomenon of the consolidation of memory. The experiment relied, in the classical manner, on the manipulation of the component under investigation, the CREB protein, relative to behaviours induced by various designed environments. The top down manipulability from phenomena to component is gained by going to the cellular level for the observation of what happens during the whole LTP process (Frey et al. 1993). Secondly, this experiment illustrates perfectly the claim that we defended above, according to which the relation of manipulability runs across the whole hierarchy of mechanisms, so to speak from molecules to behaviour. In the present case, the absence of the synthesis of a single type of proteins at the cellular level disrupts the LLTP process, which in turn impairs the mouse’s ability to durably store information relative to its environment. The impact of the genetic specificities characterizing the mutant rodents is finally observable from the behavioural point of view through their freezing attitudes, given an appropriate stimulus in the experimental set-up. This brings up a third remark. Bourtchouladze et al. experiments nicely illustrate how the resort to appropriate environmental conditions can help to make salient at the functional level the causal impact of differences among the components of a given molecular mechanism. After all, the behavioural differences such as those presented by Bourtchouladze et al. could be observed only given certain circumstances. These researches outline perfectly how different experimental set-ups can bring out, more or less precisely, reliable behavioural data. We moved from the muddled data given by the results of the Morris water maze, which, in its design, couldn’t help us

286

From psychology to neuroscience

observe salient differences in long and short term memory. We then moved to the second set-up involving a more intrusive paired stimulus, which could be re-applied over a shorter time frame. This obviously hooks up with our claim about the principles premised in neuroscience experiments, which operate to maximize a behavioural contrast between different groups of individuals that mirror their respective neurological specificities with the result that the behavioural specificities are salient enough to be observable using the descriptive resources of psychology (7.2.1 Neuropsychology and experimental design). This obviously has an unignorable bearing on the controversies around the multiple realisation thesis in the philosophy of mind. In the second experimental task, twenty-four hours after the presentation of the initial stimulus, control rodents were spending four times more time freezing than mutants, an observable difference that can be traced back to microscopic changes at the molecular and genetic level. This motivates our fourth point, which is that the second experimental set up, with its goal of creating conditions to maximize the behavioural consequence of biochemical specificities, is exactly what we would expect given our proposed strategy of functional reduction, namely, that the observer wants to be in a position to distinguish differences from the folk psychology perspective using the resources of common-sense functionalism, thus affirming the fact that different lower-level types may fulfil the same causal role up until a point. At this point, we hope to have clearly illustrated how our strategy is supposed to applicable down to molecular neuroscience. On that basis, let us sketch out what would be the guidelines of an application of that strategy in the present case. Assume firstly a roughly functional definition of memory, as an ability to adapt future behaviour in accordance with a past event. For instance, to remember event q in situation r enables an individual to engage a behaviour b2 instead of the behaviour b1 that would have been initiated in the absence of the memorization of q. In the case of the mouse involved in the Bourtchouladze experiment, rodents are considered to remember the relation between the two first stimuli and the electrical shock to the

Reduction of psychology to neuroscience: cases studies

287

extent that they spent a significant time freezing instead of not freezing when one of both inititial stimulus is presented again. Let us capture that causal role by “R”. Both groups of rodents fullfill at some point this causal role, since both control and mutant freeze given the novel presentation of one of both initial stimulus up to thirty minutes after the initial exposition. In the next place, it is also clear that we are facing a non-trivial case of multiple realization, since the biochemical differences distinguishing both groups of rodents concern the components that are involved in the execution of the causal role R. As explained, the CREB transcripiton factor is a crucial component of the mechanism by means of which long-lasting structural changes on the level of dendrites are achieved, which accordingly enables the phenomenon of memory consolidation. Now, according to our interpretation of the multiple realization argument, these physical differences between control and mutant mice imply causal differences that can be grasped from the behavioural point of view if we can provide adequate environmental circumstances, leading us to observe functional differences as well. Let us make clear how these environmental circumstances and functional specificities intervene in the context of the experiment under consideration. If one compares the behaviour manifested by each group of rodents more than one hour after the initial presentation of the stimulus, both group differ; finally, they present an important behavioural difference in their respective freezing attitudes measured twenty-four hours later. Therefore in this context c, namely being exposed twentyhours later either to the box in which the shock was applied or to the tone, we can observe that control rodents fulfill the causal role R with the specificity S1, freezing for forty percent of the time, whereas the mutants fulfill the causal role with specificity S2, freezing for only ten pourcent of the time. The latter is a difference easily grasped by an observer who does not have available the conceptual tools of microbiology. This can obvoiusly be described as a difference in the way they remember the initial elctrical shock.

288

From psychology to neuroscience

Against this background, one can construct two sub-types R’ and R’’ of R, which are respectively defined as R + S1 and R + S2. This is of course oversimplified here. But as our general argument suggests, we should now be able to construct functionally defined sub-types of R grasping in detail the functional specificities of each group of rodents in order to obtain a co-extensionallity with the neurobiological description of each type of rodent fulfilling R. This would in turn allow us to make a conservative reduction of R. On the one hand, coextensionality (6.3.4 Relation between functionally defined sub-types and physical types) forms the basis for our homogeneous reductive explanations, premised on the classical reduction to sub-types. On the other hand, (6.3.3 Relation between psychological types and subtypes), the original type R can be gained by bracketing the behavioural specificitites S1 and S2 that each group of rodent manifests in the context c through an intra-theoretical process of abstraction of functional specificities. At this point, we hope to have illustrated how the strategy of epistemological reduction by means of functionally defined sub-types can be applied to neurobiology, showing that current empirical research already provides hints in the required direction. Of course, we do not claim here that reduction is achievable in practice. This claim would be unreasonably strong both because of the complexity of the mechanisms at play here and given the current state of the art in neurobiology. Nonetheless, against the background of our developments, it seems justified to assume that the way for a conservative reduction of psychological types and laws is in principle open and that this claim is supported by recent empirical results showing that the crucial premises of our argument, namely the connection between compositional differences and functional differences, is a well grounded one, even if the latter concerns only components from which we could at first glance doubt that their contribution to the overt behaviour can be observed using the conceptual tools of psychology.

7.4 Summary and transition In this chapter we wished to bring our general account of epistemological reductionism in contact with the current state of the art

Reduction of psychology to neuroscience: cases studies

289

in neuroscience. The question we pursued is: does the neuroscience research paradigm give us any reasons to think that the program of a conservative reduction of folk psychology to neuroscience could actually be operationalized? The case of the disconnection syndrome enabled us to provide an overview of the whole approach. Starting with the functional definition of a mental property type applying to both normal and commissurotomized subjects, we saw that given a set of favoring circumstances in our experimental set-up, we could manifest the behavioural specificities implied by the split-brain condition, and thus build fully functionally defined sub-types of our targeted mental property type, which would then, in turn, be reducible to the lowerlevel types by means of homogeneous reductive explanations. Thus, the multiple realization issue is turned into an intra-theoretical issue and the original psychological type can be retrieved by bracketing the functional contextual details taken into account in the sub-types. Secondly, we tried to substantiate the claim that any neurological specificity will imply behavioural specificities as well if one provides appropriate environmental circumstances, the purpose being to make the behaviour conspicuous enough to be grasped from the functional point of view. Neuropsychology is in that respect a case of primary interest, since it uses advanced experimental set-ups in order to highlight the mutual dependency between brain components and behaviour. From that point of view, neuropsychological experimental set-ups are designed specifically to provide the type of circumstances that are required in order to build functionally defined sub-types of mental property types. As argued, even in the case of complex neuropsychopathologies that have very diffuse behavioural effects, it is nonetheless possible to impose environmental conditions under which these behavioural effects are perfectly conspicuous, supporting the claim that the appraoach illustrated with respect to split-brain patients is reproducible with respect to other neuropsychopathologies. Finally, we illustrated the claim that the approach developped here is in principle applicable all the way down to the physical components of the brain, as suggested both by ontological reductionism and by the mechanistic framework in conjunction with the multiple realization thesis. The discussed experiments on the consolidation of

290

From psychology to neuroscience

memory show the extent to which it is possible to observe at the behavioural level the impact of specificities that could seem completely negligeable at first glance if, to repeat, we are able to design an adequate experimental set-up. At this point we hope to have shown that the crucial claim on which our approach is based, namely that relevant compositional specificities among entities verifying the same psychological description imply functional differences as well, is not only supported by conceptual analysis, but by empirical evidences as well.

Chapter 8 FINAL REMARKS 8.1 Complete reductionism We developed a complete reductionist position with respect to the relation between the mental and the physical, arguing in favour of both ontological and epistemological reductionism. Let us sum up the main stages of our argument. On the ontological side, the token-identity thesis is the only position that can vindicate the causal efficacy of mental property tokens. If one takes for granted the premise of the causal, nomological and explicative completeness of physics, it is necessarily the case that behavioural patterns, qua being temporally extended complex arrangements of physical property tokens, have complete physical causes, which are sufficient to bring them about and that explains why they occurred. This reduces the logical space of possible solutions to the problem of mental causation to only three options. Firstly, one can simply abandon the causal efficacy of mental property tokens. But this amounts to reject an important part of our anthropological conception of ourselves as rational agents. Secondly, one can accept the causal overdetermination of the behaviour by previous physical property tokens and additional mental property tokens. However, causal overdetermination should be rejected, since it conveys several philosophical difficulties, among which figure the facts firstly that it is unclear what could be the distinctive causal works done by mental property tokens if every behavioural pattern has already a complete

292

From psychology to neuroscience

physical cause and, secondly, that overdetermination is committed to a Humean metaphysical framework, which excludes agency and faces the charges of quidditism and humility. This motivates the endorsement of the third option, according to which the mental is “nothing over and above” certain complex physical processes whose building blocks are fundamental physical property tokens. However, the way to ontological reductionism is not so straightforward since, on the one hand, Kripke’s argument shows that scientific identifications are necessary statements while on the other hand, the multiple realization argument shows that it is not the case, and especially not necessarily, that the exemplification of a mental property of a given type entails the systematic exemplification of a unique corresponding type of physical property. This result ultimately grounds the endorsement of a weaker version of the claim that the mental is identical to the physical, namely that any mental property token is identical to a certain configuration of physical property tokens. Without subscribing to the identity of types, the token-identity thesis nonetheless secures the causal efficacy of mental properties, since, by definition of strict identity, mental property tokens are configurations of physical property tokens and vice-versa, and that they are thereby perfectly qualified to cause the occurrence of further physical property tokens without contradicting the causal closure of the physical world. The epistemological side of reduction is suggested by the tokenidentity thesis and ultimately motivated by the ontological lessons that have to be drawn from its hypothetical failure. On the one hand, tokenidentity implies that the same entities in the world make possibly true predicates belonging to different descriptive modalities. This obviously suggests that the true descriptions of a given entity stand in a relation of derivability, which is grasped by an epistemological interpretation of strong supervenience. To put it in a nutshell, a complete physical description of an entity grasps its entire causal dispositions with the result that, as long as mental property types are functionally individuated, the psychological description applying to this entity is determined. If we add to this the assumption that these functionally defined psychological predicates have a nomological value, then it is also expectable that law-like generalization of psychology are derivable

Final remarks

293

from the laws of fundamental physics. On the other hand, if it turned out that psychological law-like generalization are not reducible to physical laws, this would jeopardize the ontological status of mental property tokens. After all, if the causal behaviour of an entity falling under a certain psychological nomological predicate was not graspable by physical laws, this would signify that there are entities in the world that enter in causal relation that are not graspable by physics, leading hence to the failure of ontological reductionism. The key role is here again played by the multiple realization argument, which prevents the establishment of bi-conditional bridge laws between psychological and constructed physical descriptions with the result that classical Nagelian reductionism, which is based on the derivability of laws by means of such correspondence principles, fails. According to the multiple realization argument, it is not the case that there is a necessary and sufficient physical condition corresponding to each mental property type. Therefore, it is neither possible to derive each psychological law-like generalization from a single lower-level law nor to account in physical terms for the ability of psychology to provide homogeneous explanations of entities that are only heterogeneously explained by physics. This is ultimately what leads both the New wave account of inter-theoretic reduction and Kim’s account of functional reduction to the elimination of psychology. In both cases, intermediary descriptive resources have to be introduced in order to bridge the gap opened by multiple realization. In the case of NWR, the multiple realization argument implies that there are several lower-level reductive bases from which an image-theory of psychology can be inferred. Given that each of these image-theories is formulated in a theoretical background that is more complete, this suggests the replacement of psychology by these image-theories, with the result that the ontology of psychology is dissolved in favour of the ontologies of these image-theories. In the case of Kim’s functional reduction, the introduction of intermediary types does not lead to any better result. Provided that these intermediary types are individuated in a hybrid way, mixing psychological and physical criteria, the initial problem of multiple realization is only displaced to the relation between mental types and these kind-specific types, with the result that

294

From psychology to neuroscience

this position is unable to account for the ability of psychology to abstract from physical differences. At this point, if we observe that the entities that make true the same psychological description but different physical descriptions are ipso-facto compositionally and causally heterogeneous, the ability of psychology to grasp objective features of the world is jeopardized, since it is unclear in virtue of what entities that are compositionally and causally heterogeneous can be grouped together under the same psychological type. Against this background, the multiple realization argument appears to drive to the discussed uncomfortable dilemma. On the one hand, the multiple realization argument, associated with ontological reductionism, suggests that the identical causal relations that are supposed to make true psychological law-like generalizations do not exist, what suggests their elimination. On the other hand, if one wants to vindicate the claim that psychology does grasp identical causal relations between identical mental tokens, one has to subtract these causal relations and their relata from the physical world and thus contradicts ontological reductionism. In both cases, the requirement that mental descriptions apply to causally identical mental property token motivates the move towards either non-reductive physicalism or eliminativism. However, if this requirement of identity is weakened in favour of a criterion of functional similarity, it is possible to provide a conservative account of epistemological reductionism, which accounts for the ability of psychology to abstract from lower-level differences.

8.2 Conservative reductionism Our account of reduction is conservative on both ontological and epistemological sides. Firstly, the token-identity thesis defines a form of ontological monism, since it implies that our world is made of one single ontological layer, which constitutes the truth conditions for the application of several descriptive modalities. Among theses ones, physics occupies a privileged position in the following sense. It is the finest grained descriptive modality available with the result that the world appears from its point of view causally, nomologically and explicatively closed. However, this descriptive privilege of physics should not appeal the untimely claim that the world is only made of

Final remarks

295

physical property tokens. Of course, it implies that any entity making true a psychological description is composed of physical fundamental property tokens. But since strict identity is symmetrical, taking these entities to be physical only would be mistaken. Certain complex configurations of physical property tokens, qua being complex causal powers, are originally and genuinely mental property tokens. This secures from the ontological point of view the conservative character of our account. The issue of the ability of psychology to abstract from compositional and causal heterogeneity of certain entities in order to group them under the same mental type raises the question of the conservative character of our account on the epistemological side, since it is that question that drives both NWR and Kim’s account towards eliminativism. The starting point of our conservative epistemological reductionism is precisely the observation that the different types of entities verifying the same psychological description are de facto causally heterogeneous. This implies that they have different causal powers and that they have different ways to bring about their psychological effect. Such causal differences can be grasped from the functional point of view, to the extent that appropriate environmental circumstances are provided. As a result, it is possible, for any physical type of entity coming under a certain mental type, to construct a psychological sub-type that is nomologically co-extensive to the former and that is individuated by the causal role of the original type plus additional functional specificities, whose observability depends on the presence of adequate environmental circumstances. To express it in mechanistic terms, if different mechanisms underlie the exercise of a certain causal role, then they differ in the way they bring about the effect that is characteristic of this causal role and these differences are observable from the psychological point of view under particular circumstances, as for instance experimental set-ups. The possibility of constructing fully functionally defined subtypes has five decisive outcomes. Firstly, the co-extensionality between physical constructed types and functionally defined psychological subtypes enables to reduce the latter to the former by means of either classical reductionism or homogeneous reductive explanations.

296

From psychology to neuroscience

Thereby, the psychological sub-types and the detailed psychological laws-like generalizations in which they stand can be safely reduced. Secondly, multiple realization is turned into an intra-theoretical issue, designating the relation between abstract psychological types and their respective sub-types, with the result that it raises no longer any problem for inter-theoretic reduction. Thirdly, this intra-theoretic relation opens the way to an account of the ability of psychology to provide homogeneous explanations of entities that are only heterogeneously explained by lower-level descriptive modalities: it is by abstracting from the functional contextual specification distinguishing the different sub-types of a certain mental type that psychology comes to provide homogeneous explanations, since all of them contain the functional definition of the abstract type. From that point of view, our account highlights that the different types of entities can be functionally similar enough in order to verify the same psychological description, even if they fail to be perfectly similar. As a result, we meet here the challenge raised by the multiple realization argument, namely to account for the process of abstraction grounding our psychological taxonomy. Contra the orthodoxy, according to which it abstracts from physical differences, psychology abstracts from functional specificities. Fourth, our account vindicates both the scientific quality of psychology, since it shows that different types of entities are sufficiently similar to come under a given psychological description, and, fifth, its scientific indispensability, since it is only by using coarse-grained functional concepts that we are able to bring out what relatively large sets of entities have objectively in common. Consequently, there is here no trend towards the elimination of psychology. Let us finally highlight that this conservative character prevents the sub-type strategy of reduction of running into the dilemma of multiple realization. To put it briefly, it simply dissolves the dilemma because it makes the reductionist horn unproblematic, providing an account based on relations of functional similarities that secures the legitimacy of the psychological taxonomy in spite of multiple realization. After all, we precisely started from the consequences of the multiple realization argument in the context of ontological

Final remarks

297

reductionism, according to which two mental property tokens that are respectively identical to two configurations of physical property tokens of two different types are at best imperfectly similar rather than perfectly similar, in order to build in a second time a tailor-made account, which can conciliate the reducibility of psychology and the causal efficacy of mental properties. The price to be paid in exchange of that result is obviously that there are mental property tokens verifying the same mental description but that are not perfectly identical. It is therefore possible that Jack’s belief that p and John’s beliefs that p are not identical but only imperfectly similar. However, we don’t see any decisive argument that should commit us to reject this claim. After all, there is no good evidence supporting the idea that such mental tokens are in fact perfectly similar and it is common in language to apply the same description to things that are only significantly similar. Moreover, we argued and illustrated that these property tokens are similar enough from the functional point of view to count as tokens of the same abstract type, since they appear to differ only to the extent that very precise environmental circumstances are provided.

8.3 Back to the mind-body problem All of this brings us back to the generality of the mind-body problem. As pointed out in the introduction, the mind-body problem is offered in a range of four different dimensions. Along this dissertation, we addressed the issue of mental causation, searching for a position vindicating the causal efficacy of mental property tokens in a physical world, which is governed by complete laws of physics. At this point, we have provided an account meeting this objective, which is characterized by its reductive but nonetheless conservative character. Thereby, the strategy of reduction by means of functional sub-types makes it possible to receive positively the progress neurosciences is currently making in view of producing reductive explanations of mental processes, uncovering the mechanisms by means of which these ones are implemented. There are no reasons to fear either a sad competition between different explicative theories or the elimination of certain ones among them.

298

From psychology to neuroscience

This leaves us with the other three dimensions of the mind-body problem. After all, mental causation is one aspect of the mind-body problem among others, which has not been addressed here. However, to secure mental causation is of first importance, since we expect the other features of the mental to contribute somehow to the production of the behaviour. In other words, as long as mental causation is not vindicated, one can wonder what can be explained by reference to the content, the intentionality or the phenomenality of mental property tokens. This brings two last brief remarks. On the one hand, accounts addressed to these other characteristics of the mental should firstly make room for the constraint uncovered along our developments. As argued in detail, if mental property tokens are brought outside the physical world for some reason, then their causal efficacy will be in any case jeopardized and the problem of mental causation will occur again. From that point of view, such accounts should focus on the question to know how it is possible that some complex configurations of physical property tokens, which fulfil a causal role that is typical of a certain type of mental property, exhibit one or several of these three features. This calls therefore for a naturalization of content, intentionality and phenomenality in terms of conservative identity. On the other hand, if the outcome of our investigations imposes constraints on upcoming accounts of these characteristics of the mental, it also delivers new possible directions for further investigations that we simply mention here. Firstly, we have shown that the descriptive resources of folk psychology, and thereby of common-sense functionalism, are significantly more extensive than expected, since coextensionality with constructed physical types is in principle reachable. This could be of importance for causal accounts of content and intentionality but also for accounts of phenomenality such as representationalism, of which certain versions use purely functional resources. Secondly, we also argued that, metaphysically speaking, properties are modes of objects, which are at the same time qualities and powers to bring about further property tokens, and that the distinction between categorical and causal properties is one of

Final remarks

299

predicates. Thereby, any property token, qua being a certain quality, is a disposition to cause some effect. Such a metaphysical stance opens the way to the development of approaches of the phenomenality of conscious experiences that mirrors our discussion of the relation between physical-compositional descriptions and functional descriptions. Just as we argued that functional and compositional descriptions can be related systematically because of the fact that, ontologically speaking, functional property tokens are identical to configurations of physical property tokens, one could pursue an account of phenomenality claiming that qualitative and functional descriptions can be systematically related – obviously in a way to be discovered – precisely because the qualitative and the dispositional descriptions of properties refer metaphysically speaking to one and the same kind of things, namely modes of objects. Beside this claim lays obviously the idea that if dispositions and qualities cannot be prized apart, then zombies are metaphysically impossible with the result that the most powerful arguments against physicalism fall (see for instance Heil 2003: chapter 20) This is however for future investigations. At this point, we simply aim to stress that a solution to the problem of mental causation requires a complete and conservative account of the relation between the mental property and physical property tokens and that the multiple realization argument is not as a conclusive argument against such a complete and conservative reductive account. This conclusion should be of first interest for any account addressed to the mind-body problem since it constitutes the starting point for a complete account of the insertion of the mind in the physical world.

BIBLIOGRAPHY Abel, T., Nguyen, P. V., Barad, M., Deuel, T. A. S., Kandel, E. R. and Bourtchouladze, R. (1997): 'Genetic demonstration of a role for PKA in the late phase of LTP and in hippocampus-based longterm memory'. Cell, 88, pp. 615-26. Azaiwa, Ken and Gillet, Carl (2009): 'The (Multiple) Realization of Psychological and other Properties in the Sciences '. Mind & Langage, 24, pp. 181-208. Bechtel, William (2001): Connectionism and the mind. Parallel processing, dynamics and evolution in networks. Oxford: Blackwell. Bechtel, William and Mundale, Jennifer (1999): 'Multiple realizability revisited: linking cognitive and neural states'. Philosophy of Science, 66, pp. 175-207. Bedeschi, Maria Francesca, Bonaglia, Maria Clara, Grasso, Rita, Pellegri, Alda , Garghentino, Rosaria Rita , Battaglia, Maria Amalia , Panarisi, Anna Maria , Di Rocco, Maja , Balottin, Umberto, Bresolin, Nereo, Bassi, Maria Teresa and Borgatti, Renato (2006): 'Agenesis of the Corpus Callosum: Clinical and Genetic Study in 63 Young Patients'. Pediatric Neurology, 34, p. 186. Beebee, Helen (2006): 'Does anything hold the universe together?'. Synthese, 149, pp. 509-33. Bickle, John (1998): Psychoneural Reduction. The New Wave. Cambridge Massachusetts Institute of Technology: MIT Press. Bickle, John (2003): Philosophy and neuroscience. A ruthlessly reductive account. Dordrecht: Kluwer. Bird, Alexander (2005): 'Laws and essences'. RAtio (new series), 18, pp. 437-61. Black, Robert (2000): 'Against quidditism'. Australasian Journal of Philosophy, 78, pp. 87-104.

302

From psychology to neuroscience

Blakemore, S. J., Smith, J., Steel, R., Johnston, E. C. and Frith, C. D. (2000): 'The perception of self-produced sensory stimuli in patients with auditory hallucination and passivity experiences: evidence for a breakdown in self-monitoring'. Psychological Medicine, 30, p. 9. Block, Ned (1980): 'Troubles with functionalism'. In Readings in the Philosophy of Psychology. Block, Ned (ed) Cambridge: Harvard University Press pp. 268-305. Block, Ned (1990): 'Can the mind change the world?'. In Meaning and method. Essays in honor of Hilary Putnam. Boolos, George (ed) Cambridge: Cambridge University Press pp. 137–70. Block, Ned (2003): 'Do Causal Powers Drain Away?'. Philosophy and Phenomenological Research, 67, pp. 133-50. Bourtchouladze, R., Frenguelli, B., Blendy, J., Cioffi, D., Schutz, G. and Silva, A. (1994): 'Dificient long-term memory in mice with a targeted mutation of the cAMP-responsive element binding protein'. Cell, 79, pp. 59-68. Braddon-Mitchell, David and Jackson, Frank (1996): Philosophy of mind and cognition. Oxford: Blackwell Publishing Bregant, Janez, Stozer, Andraz and Cerkvenik, Marko (2010): 'Molecular reduction: reality or fiction?'. Synthese, pp. 437-50. Chalmers, David (1995): 'Facing Up to the Problem of Consciousness '. Journal of consciousness studies, 2, pp. 200-19. Chalmers, David J. (1996): The conscious mind. In search of a fundamental theory. New York: Oxford University Press. Chisolm, R. (1959): Perceiving. Ithaca: Cornell University Press. Chomsky, Noam (1959): 'Verbal Behavior'. Language, 35, pp. 26-58. Churchland, Patricia S. (1986): Neurophilosophy: Toward a Unified Science of the Mind/Brain. Cambridge, MA: MIT Press. Churchland, Paul (1979): Scientific Realism and the Plasticity of Mind. New York: Cambridge University Press.

Bibliography

303

Churchland, Paul (1981): 'Eliminative Materialism and the Propositional Attitudes '. Journal of Philosophy, 78, pp. 67-90. Churchland, Paul (1985): 'Reduction, qualia, and the direct intospection of brain states'. Journal of Philosophy, 82, pp. 1-22. Churchland, Paul (2007): Neurophilosophy at Work. Cambridge: Cambridge University Press. Craver, Carl (2005): 'Beyond reduction: mechanisms, multifield integration and the unity of neuroscience'. Stud. Hist. Phil. Biol. & Biomed. Sci, 36, p. 24. Craver, Carl (2007): Explaining the Brain. Merchanisms and the Mosaic Unity of Neuroscience. Oxford: Clarendon Press. Cummins, Robert (1975): 'Functional analysis'. The Journal of Philosophy, p. 20. Davidson, Donald (1967): 'Causal Relations'. The Journal of Philosophy, 64, pp. 691-703. Davidson, Donald (1980): 'The Material Mind'. In Actions and Events. Davidson, Donald (ed) Oxford: Clarendon Press. Davidson, Donald (2004): 'Mental Events'. In Philosophy of Mind. Heil, John (ed) New York: Oxford University Press pp. 685-99. Descartes, René (2008): Meditations on First Philosophy. Oxford: Oxford University Press. Eck, Dingmar van, Jong, Huib Looren De and Schouten, Maurice K. D. (2006): 'Evaluating the New Wave Reductionism: the Case of Vision'. British Journal for the Philosophy of Science, 57, pp. 167-96. Ellis, Brian (2001): Scientific essentialism. Cambridge: Cambridge University Press. Endicott, Ronald P. (1998): 'Collapse of the New Wave'. Journal of Philosophy, 95, pp. 53-72. Esfeld, Michael (2005): La philosophie de l'esprit. De la relation entre l'esprit et la nature. Paris: Armand Colin.

304

From psychology to neuroscience

Esfeld, Michael (2005): 'Mental causation and mental properties'. Dialectica, 59, pp. 5-18. Esfeld, Michael (2007): 'Mental causation and the metaphysic of causation'. Erkenntnis, 67, pp. 207-20. Esfeld, Michael and Sachse, Christian (2007): 'Theory reduction by means of functional sub-types'. International Studies in the Philosophy of Science, 21, pp. 1-17. Farrer, C., Frank, N., Georgieff, N., Frith, C. D., Decety, J. and Jeannerod, M. (2003): 'Modulating the experience of agency'. Neuroimage, 18, pp. 324-33. Farrer, Chloe, Franck, Nicolas, Frith, Chris D., Decety, Jean, Georgieff, Nicolas, d'Amato, Thierry and Jeannerod, Marc (2004): 'Neural correlates of action attribution in schizophrenia'. Psychiatry research: Neuroimaging, 131, pp. 31-44. Feigl, Herbert (1958): 'The 'Mental' and the 'Physical''. Minnesota studies in the Philosophy of Sciences, 2. Ferber, Rafael (2003): Philosophische Grundbegriffe 2. Münschen: Beck. Fodor, Jerry (1968): Psychological Explanation. New York: Random House. Fodor, Jerry (1974): 'Special sciences'. Synthese, 28, pp. 97-115. Fodor, Jerry (1975): The Language of Thought. New York: Thomas Crowell. Fodor, Jerry (1987): Psychosemantic. The Problem of Meaning in the Philosophy of Mind. Cambridge, Massachusetts: MIT Press. Fodor, Jerry (1989): 'Making Mind Matter More'. Philosophical Topics, 17, pp. 59-80. Fodor, Jerry (1991): 'A Modal Argument for Narrow Content'. Journal of Philosophy, 88, pp. 5-26. Franck, N., Farrer, C., Georgieff, N., Marie-Cardine, M., Daléry, M., D'Amato, T. and Jeannerod, M. (2001): 'Defective recognition of

Bibliography

305

one's own actions in patients with scizophrenia'. American Journal of Psychiatry, 158, pp. 454-59. Frankfurt, Harry (1971): 'Freedom of the Will and the Concept of a Person'. The Journal of Philosophy, 68, pp. 5-20. Frith, C.D. and Dolan, R. (1996): 'The role of the prefrontal cortex in higher cognitive functions'. Brain Research: Cognitive Brain Research, 7, pp. 172-61. Garber, Daniel (2001): Descartes Embodied. Cambridge: Cambridge University Press. Goodale, M. A. (2000): 'Perception and action in the human visual system'. In The New Cognitive Neurosciences. Gazzaniga, M. (ed) Cambridge, MA: MIT Press pp. 365-77. Harbecke, Jens (2008): Mental Causation: Investigating the Mind's Powers in a Natural World. Heusenstamm: Ontos Verlag. Hawthorne, John (2004): 'Why Humen Are Out of Their Mind'. Noûs, 38, pp. 351-58. Heil, John (2003): From an ontological point of view. Oxford: Oxford University Press. Heil, John (2004): Philosophy of mind. New York: Routledge. Hempel, Carl Gustav and Oppenheim, Paul (1965): 'Studies in the logic of explanation'. In Aspect of scientific explanation. Hempel, Carl Gustav (ed) New York: The Free Press. Hooker, Clifford A. (1981): 'Towards a general theory of reduction. Part I: Historical and scientific setting. Part II: Identity in reduction. Part III: Cross-categorical reduction'. Dialogue, 20, pp. 38-60; 201-36; 496-529. Horgan, Terence (1984): 'Jackson on Physical Information and Qualia'. The Philosophical Quartely, 34, pp. 147-51. Horgan, Terence (1989): 'Mental Perspectives, 3, pp. 47-76.

Quausation'.

Philosophical

306

From psychology to neuroscience

Horgan, Terence (1993): 'From Supervenience to Superdupervenience: Meeting the Demands of a Material World'. Mind, New Series, 102, pp. 555-86. Horgan, Terence (1993): 'Nonreductive Materialism and the Explanatory Autonomy of Psychology'. In Naturalism: A Critical Appraisal Wagner, S. and Warner, R. (eds) Notre Dame, IN: University of Notre Dame Press. Hume, David (1977): An Enquiry Concerning Human Understanding. Indianapolis: Hackett Publishing. Hume, David (1978): 'A Treatise of Human Nature'. In Selby-Bigge, L. A. and Nidditch, P.H. (eds) Oxford: Clarendon Press. Huxley, T. H. (1898): Method and Results: Collected Essays. London: Macmillan. Jackson, Frank (1982): 'Epiphenomenal Quarterly, 32, pp. 127-36.

qualia'.

Philosophical

Jackson, Frank (1986): 'What Mary didn't know'. Journal of Philosophy, 83. Jackson, Frank (1998): From metaphysics to ethics. A defence of conceptual analysis. Oxford: Oxford University Press. Jeannerod, Marc (2003): 'The mechanism of self-recognition in humans'. Behavioural Brain Research, 143, pp. 1-15. Jeannerod, Marc and Pacherie, Elisabeth (2004): 'Agency, Simulation and Self-identification'. Mind & Langage, 19, pp. 113-46. Johnson-Laird, Philip N. (1983): Mental models: towards a cognitive science of language, inference and consciousness. . Cambridge: Cambridge University Press. . Kim, Jaegwon (1976): 'Events as Property Exemplifications'. In Action Theory Brand, Myles (ed) Dordrecht: Reidel pp. 159-77. Kim, Jaegwon (1992): 'Multiple Realization and the Metaphysics of Reduction'. Philosophy and Phenomenological Research, 52, pp. 1-26.

Bibliography

307

Kim, Jaegwon (1993): 'Concepts of supervenience'. In Supervenience and Mind. Selected Philosophical Essays. Kim, Jaegwon (ed) Cambridge: Cambridge University Press. Kim, Jaegwon (1998): Mind in a physical world. An essay on the mindbody problem and mental causation. Cambridge (Massachusetts): MIT Press. Kim, Jaegwon (1999): 'Making sense of emergence'. Philosophical studies, 95, pp. 3-36. Kim, Jaegwon (2005): Physicalism, or Something Near Enough. Princeton: Princeton University Press. Kim,

Jeagwon (1993): 'Postscripts on supervenience '. In Supervenience and Mind. Selected Philosophical Essays. Kim, Jaegwon (ed) Cambridge: Cambridge University Press.

Kim,

Jeagwon (2002): 'The Layered Model: Metaphysical Considerations'. Philosophical Explorations, 5, pp. 2-20.

Kolb, Bryan and Whishaw, Ian Q. (2003): Fundamentals of human neuropsychology. New York: Worth. Kripke, Saul (1980): Naming and Necessity. Oxford: Blackwell. Kulstad, Mark and Carlin, Laurence (2008): 'Leibniz's Philosophy of Mind'. In Stanford Encyclopedia of Philosophy. Zalta, Edward N. (ed). Langton, Rea (1998): Kantian Humility: Our Ignorance of Things in Themselves. Oxford: Clarendon Press. Levin, Janet (2009): 'Functionalism'. In The Stanford Encyclopedia of Philosophy. Zalta, Edward N. (ed). Levine, Joseph (1993): 'On leaving out what it's like'. In Consciousness: Psychological and Philosophical Essays. Davies, M. and Humphreys, G. (eds) Oxford: Blackwell. Lewis, David (1966): 'Lewis, David (1966): “An argument for the identity theory”. Journal of Philosophy 63, pp. 17–25. Reprinted in David Lewis (1983): Philosophical papers. Volume 1. Oxford:

308

From psychology to neuroscience Oxford University Press. Pp. 99–107.'. The Journal of Philosophy, 63, pp. 17-25.

Lewis, David (1972): 'Psychophysical and theoretical identifications'. Australasian Journal of Philosophy, 50, pp. 249-58. Lewis, David (1980): 'Mad pain and Martian pain'. In Readings in the Philosophy of Psychology. Block, Ned (ed) Harvard: Harvard University Press pp. 216-22. Lewis, David (1986): Philosophical Papers. Volume II. Oxford: Oxford University Press. Lewis, David (1994): 'Reduction of Mind'. In A Companion to the Philosophy of Mind. Guttenplan, Samuel (ed) Oxford: Blackwell. Lewis, David (2004): 'Causation as influence'. In Causation and counterfactuals. Collins, J., Hall, N. and Paul, L. A. (eds) Cambridge (Massachusetts): MIT Press pp. 75-106. Lewis, David (2009): 'Ramseyan Humility'. In Conceptual Analysis and Philosophical Naturalism. Braddon-Mitchell, David and Nola, Robert (eds) Cambridge: MIT Press pp. 203-22. Loewer, Barry M. (2007): 'Mental causation, or something near enough'. In Contemporary Debates in Philosophy of Mind. McLaughlin, Brian P. and Cohen, Johnathan D. (eds) Blackwell. Lokhorst, Gert-Jan (2009): 'Descartes and the Pineal Gland'. In The Stanford Encyclopedia of Philosophy. Zalta, Edward N. (ed). Marras, A. (2002): 'Kim on Reduction '. Erkenntnis, 57, pp. 221-57. Marras, A. (2005): 'Consciousness and Reduction'. British Journal for the Philosophy of Science, 56. Martin, C. B. (1997): 'On the need for properties: the road to Pythagoreanism and back'. Synthese, 112, pp. 193-231. McGinn, Colin (1977): 'Charity, interpretation and belief'. Journal of Philosophy, 74, pp. 521-35. McGinn, Colin (1989): Mental Content. Oxford: Basil Blackwell.

Bibliography

309

McLaughlin, Brian and Bennett, Karen (2010): 'Supervenience '. In The Stanford Encyclopedia of Philosophy. Zalta, Edward N. (ed). Mellor, D. H. (1991): Matters of Metaphysics. Cambridge: Cambridge University Press. Morgan, C. Lloyd (1923): Emergent Evolution London: Williams and Norgate. Morris, Kevin (2009): 'Does Functional Reduction Need Bridge Laws? A Response to Marras'. British Journal for the Philosophy of Science, 60, pp. 647-57. Nagel, Ernest (1961): The structure of science: problems in the logic of scientific explanation. New York: Harcourt Brace and World cop. Nagel, Thomas (1974): 'What it is like to be a bat?'. The Philosophical Review, 83, pp. 435-50. O'Keefe, J. and Dsotrovsky, J. (1971): 'The hippocampus as a spatial map: Preliminary evidence from unit activity in the freelymoving rat'. Brain Research, 34, pp. 171-75. Oppenheim, Paul and Putnam, Hilary (1958): 'Unity of sciences as a working hypothesis'. Minnesota studies in the Philosophy of Sciences, 2, pp. 3-36. Papineau, David (2002): Thinking about consciousness. Oxford: Oxford University Press. Papineau, David (2004): 'The Rise of Physicalism'. In Physicalism and its Discontent. Gillet, Carl and Loewer, Barry (eds) Cambridge: Cambridge University Press pp. 3-26. Penfield, W. and Boldrey, E. (1958): 'Somatic motor and sensory representation in the cerebral cortex as studied by electrical stimulation'. Brain, 60, pp. 389-443. Penfield, W. and Jasper, H. (1954): Epilepsy and the Functional Anatomy of the Human Brain. Boston: Little Brown. Petrif, Fred, Giles, Rachel H., Dauwerse, Hans G., Saris, Jasper J., Hennekam, Raoul C. M., Masuno, Mitsuo, Tommerup, Niels,

310

From psychology to neuroscience Ommen, Gert-Jan B. van, Goodman, Richard H., Peters, Dorien J. M. and Breuning, Martijn H. (1995): ' Rubinstein-Taybi syndrome caused by mutations in the transcriptional co-activator CBP'. Nature, 376, pp. 348-51.

Place, Ullin T. (1956): 'Is consciousness a brain process?'. British Journal of Philosophy, 10, pp. 44-50. Polger, Thomas W. (2008): 'Evaluating the evidence for multiple realization'. Synthese, 167, pp. 457-72. Proust, Joëlle (2006): 'Agency in schizophrenia from a control theory viewpoint'. In Disorders of Volition. Prinz, Wolfgang and Sebanz, Natalie (eds) Cambridge, Massachusetts: MIT Press pp. 87-118. Purves, D., Augustine, G. J., Fitzpatrick, D., Katz, L. C., Mcmantia, A. S. and Mcnamara, J. O. (2002): Neurosciences Paris: Bibliothèque nationale. Putnam, Hilary (1960): 'Minds and Machines '. In Dimensions of Mind. Hook, Sidney (ed) New York: New York University Press pp. 148-80. Putnam, Hilary (1967): 'Psychological predicates'. In Art, Mind, and Religion. Capitan, W. H. and Merrill, D. D. (eds) Pittsburgh: University of Pittsburgh Press. Putnam, Hilary (1975): 'The meaning of 'meaning''. In Mind, Language and Realizty (Philosophical Papers). Putnam, Hilary (ed) Cambridge: Cambridge University Press pp. 215-71. Putnam, Hilary (1975): 'The Mental Life of Some Machines'. In Mind, Language and Reality. Philosophical papers. Putnam, Hilary (ed) Cambridge, Mass.: Cambridge University Press pp. 408-28. Putnam, Hilary (1975): 'The Nature of Mental States'. In Mind Language and Reality. Putnam, Hilary (ed) Cambridge: Cambridge University Press pp. 435-49. Putnam, Hilary (1979): 'The Nature of Mental States'. In Mind, Language and Reality (Philosophical papers). Putnam, Hilary (ed) Cambridge: Cambridge University Press pp. 429-40.

Bibliography

311

Richardson, Robert C. (1979): 'Functionalism and Reductionism'. Philosophy of Science, 46, pp. 533-58. Robinson, William (2009): 'Epiphenomenalism'. In The Stanford Encyclopedia of Philosophy. Zalta, Edward N. (ed). Roland, P. E. (1993): Brain Activation. New York: Wiley-Liss. Rozemond, Marleen (1998): Descartes's Dualism. Cambridge, MA: Harvard University Press. Sachse, Christian (2007): Reductionism in the Philosophy of science. Frankfurt (Main): Ontos-Verlag. Schaffner, Kenneth F. (1967): 'Approaches to reduction'. Philosophy of Science, 34, pp. 137-47. Shagrir, Oron (1998): 'Multiple realization, computation and the taxonomy of psychological states'. Synthese, 114, pp. 445-61. Shapiro, Lawrence A. (2000): 'Multiple realization '. The Journal of Philosophy, 97, pp. 635-54. Shoemaker, Sidney (1982): 'The Inverted Spectrum'. Journal of Philosophy, 79, pp. 357-81. Shoemaker, Sydney (1984): 'Causality and properties'. In Identity, Cause and mind. Philosophical essays. . Shoemaker, Sydney (ed) Cambridge: Cambridge University Press. Silva, Alcino J., Kogan, Jeffrey H., Frankland, Paul W. and Kida, Satoshi (1998): 'CREB and Memory'. Annual Review of Neuroscience, 21, pp. 127-48. Sipser, Michael (2005): Introduction to the Theory of Computation. Boston: PWS Publishing Company. Smart, John C. (1959): 'Sensations and brain processes'. Philosophical review, 68, pp. 141-56. Soom, Patrice, Sachse, Christian and Esfeld, Michael (2010): 'Psychoneural reduction through functional sub-types'. Journal of consciousness studies, 17, pp. 7-26.

312

From psychology to neuroscience

Sparber, Georg (2005): 'Counterfactual Overdetermination vs. the Causal Exclusion Problem'. History and philosophy of the life sciences, 27, pp. 479-90. Sperry, R. W., Gazzaniga, M. S. and Bogen, J. E. (1969): 'Interhemispheric relationship: the neocortical commissures; syndrome of hemisphere disconnection'. In Handbook of clinical neurology Vinken, P. J. and Bruyn, G. W. (eds) Amsterdam: North-Holland Publishing Company pp. 273-90. Squire, L., Zigmond, M., Bloom, F., Landis, S. and Roberts, J. (2003): Fundamental neuroscience. San Diego: Academic Press. Synofzik, Matthias, Thier, Peter, Leube, Dirk T., Schlotterbeck, Peter and Lindner, Axel (2010): 'Misattibutions of agency in schizophrenia are based on imprecise predictions about the sensory consequences of one's actions'. Brain, 133, pp. 262-71. Synofzik, Matthias, Vosgerau, Gottfried and Newen, Albert (2008): 'Beyond the comparator model: A multifactorial two-step account of agency'. Consciousness and Cognition, 17, pp. 219-39. Turing, Alan M. (1950): 'Computing machinery and intelligence'. Mind & Langage, 59, pp. 433-60. Weatherson, Brian (2007): 'Humean Aren't Out of Their Minds'. Noûs, 41, pp. 529-35. Williams, R. W. and Herrup, K. (1988): 'The control of neuron number'. Annual Review of Neuroscience, 11, pp. 423-53. Woolhouse, Roger S. (1993): Descartes, Spinoza, Leibniz: The Concept of Substance In Seventeenth Century Metaphysics New York: Routledge. Yablo, Stephen (1992): 'Mental Causation'. The Philosophical Review, 101, pp. 245-80.

INDEX A abstraction 14, 112, 210, 234, 236-238, 245, 263, 288, 296 anomalous monism.................18 autonomy ..............................112 B Bechtel, W. ...........................223 Bickle, J. 3, 183, 189-190, 192193, 197, 281 Bird, A. ........................ 133, 135 Block, N. 98, 100, 117, 121, 126, 162 Bogen, J. E............................255 Bourtchouladze, R. .... 280-281, 283, 285, 286 Bregant, J. .............................189 C causal role 57-59, 67, 69, 76, 103, 108, 110-112, 115,-117, 125-128, 130-131, 133, 140, 147, 148-149, 151-155, 157160, 167, 199, 201-203, 206207, 210, 218, 220-236, 238249, 252, 265-266, 275, 278, 280, 286, 287, 295, 298 causation 12, 17, 19, 22, 25-26, 31, 34, 38, 40, 45-47, 49, 51, 54, 76, 79, 81, 85, 88, 97, 99,

129, 130-132, 134, 140, 200, 298 ceteris paribus 177, 179, 227, 228 Chalmers, D. ............ 7, 154, 155 Chomsky, N. .................... 6, 109 Churchland, Patricia ............ 163 Churchland, Paul ..11, 191, 193194, 196, 199 commissurotomy 255-256, 258, 264, 270 completeness of physics ...8, 15, 25-26, 28, 30-32, 34, 40, 43, 48, 49, 51, 53, 55, 59, 61-62, 64, 67, 69, 79, 80-83, 86-90, 97, 101, 158-160, 180, 202, 248, 254, 291 component 104, 114, 122, 148, 149, 150, 152-154, 157, 159, 163-164, 166, 168, 222, 266, 278, 279, 285, 287 Craver, C. F. 3, 96, 140, 146, 150-151, 153, 159, 160-163, 230, 285 Cummins, R. ................ 151, 267 D Davidson, D. 17-18, 20, 66-67, 69

314

From psychology to neuroscience

deduction 172, 178, 180-181, 186, 193, 194-196, 205, 214, 218-219 Descartes, R. 5, 8, 27-28, 35-36, 38-39 dilemma 13, 39, 47, 88, 121, 125, 128, 130-131, 136-137, 171, 191, 212-213, 215, 219, 294, 296 disconnection 14, 82, 132, 251252, 254-255, 258, 270, 289 disposition 105, 137, 149, 151, 167-168, 220, 299 dualism 13, 19, 35, 37, 39-40, 42, 62-63, 78, 84, 214

71, 77, 86, 88, 98, 101, 126, 130, 160, 163, 172 Esfeld, M...................... 3, 25, 32 event Davidsonian event........ 17, 51 Kimian event.............. 16, 114 experiment 72, 75, 149, 150, 268, 270, 272, 275, 281, 284, 285-287 experiment, inter-level 148, 150, 152, 163, 267 experimental design ... 229, 265, 272, 280, 286 explanation heterogeneous .................. 263 homogeneous 116, 186-188, 192, 207-208, 210-213, 215, 219, 241-242, 248249, 263, 293, 296 mechanistic 104, 147, 151152, 154-157, 159, 161, 164-165, 166, 168, 230, 239, 253, 254, 263, 266, 267, 278, 279, 285 physical .............. 80, 157, 202 psychological ..... 82, 118, 202 reductive 13, 154-157, 168, 171, 185, 201, 203-208, 211-212, 217-219, 223, 230, 239-240, 242, 245, 248, 254, 264, 265, 277, 288-289, 295, 297

E Eck, D. van .......................... 195 eliminativism 11, 13, 47, 116, 128, 130, 176, 181, 195, 197198, 200, 209-215, 218, 234, 243-244, 247, 293-297 Ellis, B. ................................ 135 Endicott, R. .......................... 195 environment 9, 10, 17, 67, 73, 76, 142-144, 148, 156, 220221, 226, 229, 266, 269, 276, 281, 285 environmental conditions 14, 109, 121, 124, 139, 167, 217, 225-226, 228, 230231, 235, 238, 241, 249, 251-252, 260-261, 270, 275, 276, 278, 280, 285, 287, 289, 295, 297 epiphenomenalism 13, 18-19, 23, 32, 40, 43, 44-46, 51, 55,

F Farrer, C. .......................272-275 Feigl, H. ............................. 6, 54 Ferber, R. ............................... 42

Index

315

Fodor, J. 6, 22, 58, 112, 117, 155 Frankfurt, H. ...........................21 functionalism common-sense 13, 103, 112, 118, 120-121, 123, 125, 139, 167, 224, 234, 240, 259, 286, 298 empirical .. 112, 115-118, 121 machine state .... 112-113, 115 G Garber, D. ...............................37 Gillet, C. ...................... 189, 192 H Harbecke, J. ........................3, 25 Hawthorne, J.........................132 Heil, J............................ 135-137 Hooker, C. A................ 178, 193 Horgan, T..... 17, 18, 44, 77, 183 Hume, D......................... 21, 131 Huxley, T. H. ..........................44 I identification 10, 56, 58, 62, 67, 92, 101, 138, 260, 271 identity token-identity 10-13, 52, 54, 63, 77, 86, 90, 92-95, 9798, 100-101, 104, 126, 128, 132, 139, 162-163, 168, 177, 181, 202, 208-209, 211, 240, 280, 291-292, 294 type-identity 6, 12, 37, 51, 53-59, 61-62, 66-67, 78,

93-94, 101, 107-108, 111112, 116, 182, 192, 215 strict identity 91-92, 94, 163, 292, 295 incompleteness of psychology... 82 intervention .................. 150, 162 J Jackson, F....................... 44, 134 Jasper, H............................... 149 Jeannerod, M........................ 274 Jong, Huin L. D.................... 195 K Kim, J. 12, 16, 18, 19, 25, 30, 49, 51, 54, 63-64, 66, 74, 79, 83, 96, 97, 100, 131, 154, 159, 160, 162, 193, 197, 201205, 207-208, 210-212, 215, 218, 234, 237, 239-240, 242243, 245-246, 293, 295 Kolb, B................. 262, 265, 271 Kripke, S. ......37, 56-57, 63, 292 L Langton, R. .......................... 133 law law-like generalization.... 172, 175-176, 182, 188, 198, 202, 209, 213, 215, 223, 232-234, 236-239, 292 physical law 18, 28, 67, 7981, 167-168, 173-174, 177178, 180, 186-187, 204, 213, 233, 238, 242, 248, 293

316

From psychology to neuroscience

psychological law 11, 13, 42, 81, 83, 171, 173-174, 176177, 180, 182, 188, 197, 200, 204, 212-214, 219, 227, 232-233, 236, 238, 242, 293-294, 296 Leibniz, G. W. ......27-28, 40-42 level layered view of the world 97, 161 level of being 94, 97, 99, 100101, 158 level of description 96, 142148, 153, 155, 158, 161, 238 Levin, J. ....37, 57, 123, 154-155 Levine, J............ 37, 57, 154-155 Lewis, D. 60, 112, 120-121, 124, 126, 129, 133-134, 259 long-term potentiation (LTP) 144, 146, 148, 152, 156, 163, 281, 283, 285

118, 120, 132, 137, 140, 200, 291, 297, 298, 299 Morgan, C. L.................... 96, 99 multiple realization 6, 10-14, 25, 53, 58-59, 61-63, 66-67, 74, 78, 84, 87, 90, 92-93, 95, 101, 106-107, 112, 116, 118, 125, 127-128, 141, 157, 162, 171, 180, 182-186, 188-193, 195198, 200, 205-210, 212-215, 217-220, 222-227, 229-232, 234, 237, 240-241, 243, 245249, 252, 260, 264, 276, 278, 280, 287, 289-290, 292-293294, 296, 299 Mundale, J............................ 223

M manipulation 150, 230, 279, 285 McGinn, C. ............................ 72 mecanism 14, 147, 148-155, 158-159, 160-163, 165, 168, 195, 203, 219, 239, 257-258, 265-266, 268-269, 271, 278279, 281-283, 285, 287 Mellor, D. H......................... 135 mental causation 6-9, 12, 14-17, 19-22, 25, 31-34, 37-40, 4551, 53-54, 61-62, 67, 76, 77, 78, 84-85, 88, 97, 100, 109,

N Nagel, E.........176, 178-179, 193 neuropsychology 14, 142, 229, 251, 259, 264-266, 269, 274, 276-277, 279 neuroscience 6, 8, 11, 13-14, 96, 100-101, 103-104, 106, 140142, 145, 147, 151-152, 154, 156-158, 161, 163-169, 171172, 199-200, 249, 251, 253254, 256, 263, 265-267, 269270, 279-280, 281, 285-286, 289 O object 7, 16, 18-20, 24, 37, 51, 76, 114, 136-137, 141-142, 145, 251, 256, 258-259, 262, 276, 298 Oppenheim, P................... 96, 99

Index

317 P

Papineau D............ 25, 30-31, 69 pathology ............. 253, 272, 276 Penfield, W. ..........................149 Place, U...................................54 Polger, T. ..............................190 property properties as modes 10, 39, 137, 298 properties as powers 17, 21, 29, 43, 135-136, 158, 165, 173, 196, 197, 198, 221222, 244 property token (definition) .19 property type (definition) .106 Proust, J. .......................... 3, 274 psychology 6, 10-11, 13-14, 42, 47, 58, 63, 68, 76, 79, 81-83, 96, 98, 100-101, 103-108, 112, 116-117, 119-120, 123124, 126-128, 133, 139, 142143, 167, 171-176, 182, 184185, 187-189, 191-192, 195, 197-201, 203, 205-206, 208215, 217-219, 223-228, 231232, 234-239, 242-247, 249, 251, 259, 261, 263-264, 277, 280, 286, 288-289, 292-298 Putnam, H. 6, 58, 59, 72, 75, 96, 99, 112-115, 125 R realism ......................... 135, 192 realizer 111, 126-128, 130-131, 138, 140 reductionism

conservative reductionism 12, 13, 251, 288-289, 294, 299 epistemological reductionism 11, 13-14, 83, 106, 112, 142, 157, 162, 171-172, 176-177, 180-183, 185189, 191, 193, 201, 203204, 206, 208, 211-215, 217-219, 224, 228, 238, 240, 242-244, 247-249, 252-254, 262, 264,
280, 288, 291, 294, 295 ontological reductionism .. 5253, 90, 100-101, 106-107, 112, 162-163, 167, 171, 174-176, 180, 184, 187, 198, 200-202, 206-207, 209, 211, 213-215, 219, 223-224, 227, 234, 240, 248, 254, 263, 276-278, 280, 289, 292-294, 297 Rozemond, M......................... 35 S Sachse, C.................................. 3 Schaffner, K. F..................... 194 schizophrenia . 14, 225, 270-276 Schouten, Maurice K. D. ..... 195 scientific quality................... 296 Shagrir, O......................190-191 Shapiro, L. A. 191-192, 229, 230 Shoemaker, S. ...............134-136 similarity imperfect .......... 127, 227, 297 perfect 91-92, 95, 214, 222, 296, 297

318

From psychology to neuroscience

Sipser, M.............................. 113 Smart, J. ....................... 6, 54, 55 Soom, P................................ 224 Sparber, G. ............................. 48 spatial memory 142-144, 147, 153, 155, 156, 161, 163, 266, 281-284 split-brain patient 256-257, 259260, 262-264, 275, 289 supervenience local.................................... 59 strong 71-74, 77-78, 84-86, 89-90, 94, 101, 172-174, 292 weak .................. 70-71, 74, 78 T truthmaker...................... 18, 129 Turing, Alan. 111, 113, 115, 119 type abstract type 14, 234-239, 243, 247, 249, 263, 296, 297 physical type 14, 24, 58-59, 62, 94, 107, 111, 118, 179, 196, 206, 208, 210, 217,

219-222, 224, 228-234, 236, 238-249, 278, 288, 295, 298 psychological type ...14, 92, 107-108, 116, 137, 173, 186, 191, 208, 210, 213, 215, 217, 219-220, 222226, 228-229, 231-239, 241-249, 261, 263, 288, 289, 294, 296 sub-type …14, 208, 210-211, 213, 217-218, 224, 228, 231-249, 251-252, 259264, 274-275, 276-278, 288-289, 295-297 W Weatherson, B...................... 132 Whishaw, I. ...........262-265, 271 Y Yablo, S.................................. 85