Evolutionary Analogies: Is the Process of Scientific Change Analogous to the Organic Change? [Unabridged] 1443833541, 9781443833547

Citation preview

Evolutionary Analogies

Evolutionary Analogies: Is the Process of Scientific Change Analogous to the Organic Change?

By

Barbara Gabriella Renzi and Giulio Napolitano

Evolutionary Analogies: Is the Process of Scientific Change Analogous to the Organic Change?, by Barbara Gabriella Renzi and Giulio Napolitano This book first published 2011 Cambridge Scholars Publishing 12 Back Chapman Street, Newcastle upon Tyne, NE6 2XX, UK British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library Copyright © 2011 by Barbara Gabriella Renzi and Giulio Napolitano All rights for this book reserved. No part of this book may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without the prior permission of the copyright owner. ISBN (10): 1-4438-3354-1, ISBN (13): 978-1-4438-3354-7

A Sirio

TABLE OF CONTENTS List of Figures and Tables .......................................................................... ix Preface ........................................................................................................ xi Acknowledgments ..................................................................................... xv CHAPTER ONE .............................................................................................. 1 EVOLUTIONARY ANALOGY 1.1 Natural Selection............................................................................... 1 1.2 Evolutionary Epistemologies ............................................................ 4 1.3 Metaphor........................................................................................... 8 Some history Metaphor as a cognitive structure How do metaphors work? Type Hierarchies CHAPTER TWO ........................................................................................... 21 EVOLUTIONARY ANALOGIES 2.1 Kuhn ............................................................................................... 21 Scientific development Progress from: the evolutionary analogy in The Structure of Scientific Revolutions Local and global progress RSS: adaptation and niches Ecological niches, localities and adaptation in evolutionary biology Kuhn’s confusions 2.2 Campbell......................................................................................... 39 Knowledge gain and the blind-variation-and-selective-retention mechanism How blind is knowledge gain? Campbell’s intention 2.3 Toulmin........................................................................................... 52 Explanation or metaphor? Darwinian rationality Environment Ontological adequacy CHAPTER THREE ........................................................................................ 65

viii

Table of Contents

A TYPE HIERARCHY FOR SELECTION PROCESSES 3.1 Type Hierarchies............................................................................. 66 Top-down or bottom-up? 3.2 Existing generalisations .................................................................. 67 Hull Hull is not general enough Darden and Cain Darden and Cain are too general 3.3 Type hierarchy of selection processes ............................................ 72 A provisional hierarchy Mechanical Selection Evolution The complete Type Hierarchy Intentionality and Artificial Selection Cyber Selection Drift Immune system 3.4 The usual objections ....................................................................... 82 Variation Progress 3.5 The evaluative step ......................................................................... 86 A brief note on ontological and pragmatic approaches 3.6 Conclusion ...................................................................................... 91 CHAPTER FOUR .......................................................................................... 93 SCIENCE AS A SPECIAL CASE OF SOCIOCULTURAL ENTITY 4.1 Direction and convergence in science............................................. 93 Replication Interaction Identity by descent Convergent realism 4.2 Natural Selection and Sociocultural Selection .............................. 112 Science as a special case of sociocultural entity The Orthogenetic Hypothesis The Orthogenetic Analogy 4.3 Orthogenesis and scientific change............................................... 115 CONCLUSION ............................................................................................ 121 REFERENCES AND BIBLIOGRAPHY ............................................................ 123 INDEX ....................................................................................................... 135

LIST OF FIGURES AND TABLES

Figure 1 – A type hierarchy of harmonic oscillators ................................. 17 Figure 2 – A type hierarchy for atoms and the solar system...................... 18 Figure 3 – Evolutionary niche with two localities ..................................... 35 Figure 4 – Kuhn and the evolutionary analogy.......................................... 38 Figure 5 – Hull’s abstraction of selection.................................................. 67 Figure 6 – A provisional type hierarchy for selection processes ............... 73 Figure 7 – A type hierarchy for selection processes .................................. 78 Figure 8 – A type hierarchy for boats and trucks....................................... 79 Figure 9 – A graphical representation of convergent realism .................. 110 Figure 10 – A type hierarchy for hat selection ........................................ 115 Table 1 – Science course and content selection......................................... 51 Table 2 – Toulmin’s evolutionary analogy mapping ................................. 59 Table 3 – Darden and Cain’s abstractions ................................................. 72

PREFACE

Our aim in this volume is to analyse what we will refer to as ‘the evolutionary analogy’, a particular form of evolutionary epistemology which claims that scientific change is governed by the same mechanisms, or by mechanisms analogous to those at work in organic evolution, mainly natural selection. In terms of questions, the overall aim of this volume is to answer the following: “Is the process of scientific change analogous to or even the same as organic change?” Scientific change is and has been the subject of major philosophical debate. Science is now largely perceived by the layman as the enterprise capable of giving the best explanations of what reality is. The Philosophers, more subtly and depending on their orientation, discuss the truthfulness, the verisimilitude, the usefulness, the applicability of those explanations, or even whether there is a reality we can explain at all. If we exclude this latter case, all philosophers interested in science have addressed, in varying depth, the question of the change of scientific knowledge over time. The answers produced can be categorised, at first approximation, as normative or descriptive in varying degrees. Normative answers analyse the steps or general principles which scientists should follow when choosing or altering their theories, in order to produce new theories which would be better at doing what they are supposed to do. Their focus is on the objectives of scientific change and their character is primarily logical. Descriptive answers show how scientists or scientific communities actually behave. Their focus, thus, is mainly on the process of scientific change and their character is sometimes largely sociological. Most of the philosophers who proposed evolutionary theories of scientific change – evolutionary analogies – were able to overlap these two categories. By drawing analogies or even equating the mechanisms of organic and scientific evolution they described the process underlying the latter but also justified its value, implicitly or explicitly, by a simple analogy: better theories are those which survive old ones, as better species are those which survive previous ones. The results of these philosophers, however, have not been satisfactory. Most of them have embarked on only limited or sketchy analysis, while others, who were more persistent in their attempts, failed to provide persuasive answers. The reasons for these failures are manifold: the philosophers misunderstood or oversimplified the evolutionary

xii

Preface

biology concepts they employed; or forcefully introduced in their view of scientific change mechanisms which seem counterintuitive but that closely resemble biological ones; or employed an inadequate methodology – or none at all – for assessing the analogy. A novel approach, which does not incur the same problems, is thus needed. In this volume we are interested in a purely descriptive philosophy of science and our position will be based on the critique of the most recent and comprehensive attempt to defend the evolutionary analogy, made by David Hull. In order to answer the question we have formulated above, the following issues will be addressed: what is organic evolution; what is meant by ‘evolutionary analogy’; how analogy/identity can be evaluated; how evolutionary analogy/identity has been defended by philosophers; how these defences perform; what is the best defence available; whether it passes the evaluation and, if not, whether it can be improved; if all the positions fail, whether it is possible to conceive alternative analogies between other relevant processes which do not incur the same problems. By addressing these issues, we will be able to conclude that the process of scientific change is different from organic change and that only loose analogies can be defended. In the first chapter, the tools needed for the analysis are illustrated. In particular, a short account of what is meant today by ‘organic evolution’ and natural selection is given, together with a sketched survey of the main claims of most evolutionary analogies and common objections raised against them. Finally, current views on analogies and metaphors are discussed and some points of reference for the scrutiny of individual evolutionary analogies are provided. In the second chapter the most influential scholars who have endorsed an evolutionary analogy are discussed. In the analysis of their views, the tenability of their positions is evaluated in the framework of principles set out in the first chapter. The strengths and weaknesses of different positions are made clear and several detailed concepts and lines of argument are constructed. This constitutes a basis for the third chapter, where the approach we regard as the most coherent and fruitful is analysed and further developed. In this approach, the process that makes scientific concepts and theories persist and evolve is an instance of a general process of selection, and natural selection is another instance of it. Individual concepts and features of scientific and organic change are generalised and an evaluative framework for such generalisations is proposed. In the fourth chapter, after the roles of direction and convergence are further analysed in scientific and biological change, the same framework is used to evaluate alternative analogy views. One view compares scientific change to evolution by orthogenesis, the

Evolutionary Analogies

xiii

other compares more general cultural change to evolution by natural selection, with the purpose of investigating whether these analogies perform better than all the others. A brief conclusion summarises the most salient points of the volume.

ACKNOWLEDGMENTS

We wish to thank Springer Verlag and Chicago University Press for their kind permission to reproduce here, with some minor variations, some material that one of us (Barbara) had already published with them.

CHAPTER ONE EVOLUTIONARY ANALOGY

This chapter provides introductory answers to the following questions: What is the current status of the theories of evolutionary biology and natural selection? What is an ‘evolutionary epistemology’ and what is meant, in this volume, by the terms ‘evolutionary analogy’? What is an analogy and how can analogy/identity be evaluated? The answers to these questions will be further shaped and detailed in the following chapters, where required by the context.

1.1 Natural Selection We shall provide here a brief account of one of the existing views of natural selection and its relation to biological evolution. This is the view mostly accepted by leading biologists at present and in what follows we will refer to it exclusively. With some exceptions, advocates of evolutionary epistemologies – although they are grouped under this title – mainly focus on natural selection. Several other related concepts, though, will be encountered in individual sections of the book, for example ‘genetic drift’, ‘species’, ‘lineages’ and ‘ecological niches’ to name a few. They will be introduced and defined appropriately when first encountered. Natural selection1 is a mechanism, originally proposed by Darwin, for adaptive change in the biological realm. While at the time the occurrence of evolution was accepted by many biologists, the idea that natural selection was behind it was universally rejected. In the early 1930’s it was shown that Darwin’s natural selection could function within the Mendelian theory of genetics and with the kinds of variation observable in natural populations.2 This established what is known as Neo-Darwinian Theory or The Modern Synthesis.

1 2

Often abbreviated to NS hereafter. This was done by Fisher, Haldane and Wright (Futuyma 1998: 24).

2

Chapter One

Natural selection is now widely agreed to be the principal force that changes allele frequencies within large populations.3 In biological populations, a competition for survival usually occurs because more offspring than can survive is produced. Because of genotypic variety, greater capacity to survive and reproduce is usually possessed by some individuals, whose genotypes produce phenotypes better equipped for their environment. Succeeding generations will preferentially inherit their alleles, which will increase in frequencies in the population. Sewall Wright, François Jacob and other evolutionists have pointed out that selection, owing to its two-step nature, combines chance and necessity in a unique manner. At the first step, the production of genetic variation, chance predominates. At the next step, the survival of individuals of the next generation, chance plays only a minor role, and the survival depends largely on the adaptedness of the individuals exposed to selection. (Mayr 1991: 133)

It is important to stress the relevance of differences in genotypes; selection among genetically identical members of a clone, which only differ in phenotype, cannot have any evolutionary consequence. Even more importantly, though, it should be noted that the distinction between evolution and natural selection is often forgotten. They are two different phenomena: evolution can occur without natural selection, by genetic drift for instance;4 conversely, there can be natural selection without evolution, for instance when individual “genotypes differ in each generation in survival or fecundity, yet the proportions of genotypes and alleles stay the same from one generation to another” (Futuyma 1998: 365).5

3

The enormous importance of natural selection to the evolutionary process is not questioned by any serious biologists; nevertheless a controversy concerns the power of natural selection to overcome constraints on evolution. Doubts about its power have led some to think that other processes may have an importance comparable to that of natural selection in evolution, such as for instance the tendency to self-organisation, suggested by Kauffman (1993). This tendency, however, is not proposed as an alternative to natural selection. 4 This is an important mechanism of evolutionary change, which explains the differences with respect to traits with little or no influence for the survival of a species. When there is selectively neutral variability, it can still be the case that slight differences occur between the offspring and the parental generation. 5 An example is provided by the persistence of deleterious alleles in a population. Such alleles usually persist because recurrent mutation or gene flow from other populations (where the alleles are instead favoured by differences in the

Evolutionary Analogy

3

More formally, natural selection is defined by mainstream biologists as “the differential contribution to future generations (differential fitness) of different genotypes as manifested through different phenotypes and thus producing changes in frequencies of the genes giving those genotypes” (Ferguson 2002). In general, we refer to ‘advantageous’ or ‘deleterious’ alleles but we have to remember that selection, at least according to the definition just given, does not act directly on alleles. A more articulated, yet compact account of natural selection is provided by Ernst Mayr (Mayr 1993: Chapter 3).6 According to Mayr the concept of natural selection is formed by five observations and three logical inferences. The observations are the following: all species produce more offspring than can survive; usually populations do not increase exponentially; the limit of natural resources restricts the number of individuals that can survive; in the majority of cases no two individuals are exactly the same;7 and a great part of this variation has a genetic basis due to mutation and recombination which can be inherited by offspring. The inferences, meanwhile, are the following: the production of more individuals than the limited environmental resources can support causes competition for survival; survival is not totally random but depends in part on genetic make up; evolution (genetic change) is due to the ability to survive and reproduce, which in turn depends on our unique phenotype. Each individual has thousands of traits for which it could, under a given set of conditions, be selectively superior or inferior in comparison with the average of the population. The greater the number of superior traits, the greater the probability that it will survive and reproduce. But it is merely a probability, because even a ‘superior’ individual may fail to survive or reproduce. In order to grasp the concept of natural selection it is important to remember that it acts on populations, groups of individuals, with each individual in a given population different from the others. According to Mayr, without giving importance to the particular differences among individuals and without considering a species as an aggregate of populations,8 we would not have the concept of natural selection. environment) reintroduce them constantly. Eventually, equilibrium between their reintroduction and elimination by NS is reached and persists (Futuyma 1998: 381). 6 A clear analysis of the concept of selection was already present in the fourth chapter of Mayr (1981). 7 This is particularly true for sexually reproducing species, less for asexually reproducing species. 8 Mayr speaks of population and considers a species as consisting “of a group of populations which replace each other geographically or ecologically and of which the neighboring ones intergrade or interbreed wherever they are in contact or

4

Chapter One

We acknowledge that there is some controversy over the details. For instance, the majority of scholars consider sexual selection a kind of natural selection, while a minority treats it as distinct from natural selection. Some scholars classify as natural selection only selection at the level of genes, genotypes and individual organism, and exclude the level of groups, such as species or population. Others, such as Endler (1986) and Sober (1984),9 also extend natural selection to the level of groups.10 These variations on the mainstream view, however, do not affect what most evolutionary epistemologists regard as the characteristic features of natural selection.

1.2 Evolutionary Epistemologies Evolutionary epistemologies are epistemologies directly motivated by evolutionary considerations. The expression ‘Evolutionary Epistemology’ is used to group together different attempts of using biological concepts in philosophy and is traditionally associated with the names of Donald Campbell, Stephen Toulmin, Karl Popper.11 It is now widely accepted that the various attempts can be roughly divided into two main categories. These categories have been differently defined and named over time but which are potentially capable of doing so (with one or more of the populations) in those cases where contact is prevented by geographical or ecological barriers” (Mayr 1942: 120). 9 Further, there is a special case of group selection called ‘species selection’. ‘Group selection in which each group is a species has been called species selection by Steven Stanley’ (Futuyma 1998: 352, emphasis in the original). 10 It should also be remembered that, in 1962, the ecologist V.C. Wynne-Edwards had already suggested that social behaviours, such as the flocking behaviour of starlings, had evolved as mechanisms of population control. For instance, he maintained that populations are self-regulatory because of individual reproductive restraint, which evolves by group selection. Williams Edwards (1966) extended the examples given by Wynne-Edwards and included senescence and fixed life spans (Futuyma 1998: 350-2). 11 Mention should also be made of the Austro-German School of Evolutionary Epistemology, which has its origins in a paper published in 1941 by Konrad Lorenz, “Kant’s Doctrine of the A Priori in the Light of Contemporary Biology”. This is an attempt to interpret Kantian transcendental idealism along biological lines (similar points had also been made by philosophers and scientists such as Ernst Mach and Henri Poincare). Lorenz’s paper remained mostly unknown and his later book ‘Die Rückseite des Spiegels: Versuch einer Naturgeschichte menschlichen Erkennens’ (reprinted as ‘Behind the mirror’, 1977) raised little interest among philosophers, especially in English-speaking countries.

Evolutionary Analogy

5

their grounding has stayed the same. Kai Hahlweg (1986), for instance, differentiates between what he labels ‘[proper] evolutionary epistemology’ and ‘bioepistemology’.12 In his view, if the former mainly deals with the evolution of human knowledge, bioepistemologists are concerned with the evolution of the mechanism of cognition: “[bioepistemology] can be seen as an attempt to base epistemology on results derived from scientific investigations into the nature of knowledge acquisition” (Hahlweg 1986: 172). Analogously, John Losee (2000) draws a distinction between what he terms the “Evolutionary – Analogy View” and “The Evolutionary – Origins View”. The former supports a model for a descriptive philosophy of science, whose main feature is a process of competition leading to differential reproductive success. This process occurs both in organic evolution and science development. The Evolutionary – Origins View states, on the other hand, that the application of the epigenetic rules, encoded in homo sapiens in the course of evolutionary adaptation, directs the scientific inquiry. Michael Ruse (1995: 157-65 and 1986a: 29-66, 14968), for instance, has pointed out that there are several epigenetic rules informing human evolution and ascribes the growth of science to the use of evaluative standards which emerged from the struggle for survival. According to Ruse’s perspective, these standards are an extension of the perceptual and conceptual abilities that were valuable in the struggle to adapt to environmental pressure. The dichotomy, however, is now best known in the form and with the labels given to its components by Bradie in his well-known article ‘Assessing Evolutionary Epistemology’ (1986). Bradie divides evolutionary epistemology into two distinct programmes: the evolution of epistemological mechanisms (EEM) and the evolutionary epistemology of theories (EET).13 EEM tries to supply an evolutionary explanation of the development of cognitive structures, while EET attempts to give an account of epistemological norms and human knowledge. Advocates of EEM are, for instance, Lorenz (1977),14 Popper,15 Plotkin (1982) and Wuketits (1990). EEM and EET programmes, though distinct, are connected. Bradie points out:

12

This is the approach of the Austro-German School. We will usually use the expression ‘evolutionary analogy’ to refer, in the remaining chapters, to the EET stance. 14 Konrad Lorenz is important for his biological interpretation of Kantian a priori. He maintains that these forms of perception are a priori for the single individual, but they are philogenetically a posteriori 15 Popper is also famous for his arguments in defence of an EET programme. 13

6

Chapter One There is a slippery slope leading from a central part of the EEM program (the attempt to understand the philogenetic development of the biologically material cognitive apparatus) through the claim that all organisms and lineages have “built in” specific cognitive apparatus characteristic of their place in the philogenetic tree to the claim that each organism has its own characteristic “a priori” Kantian categories and finally to the central claim of the EET program, viz., that the content of knowledge as shaped, in part, by the “a priori” categories itself undergoes some form of evolutionary development. (Bradie 1986: 409)

David Sloan Wilson (1990) also suggests that evolutionary epistemologies draw a complex connection between adaptation and knowledge. He distinguishes between the two programmes using the following short formula: EET considers adaptation as knowledge, while EEM considers knowledge as adaptation. EET programme is inclined to consider knowledge as a result of evolution and, biological adaptation is sometimes considered a form of knowledge itself;16 EEM programme, on the other hand, has a propensity to tackle the ability to know and the knowledge generated as a biological adaptation. Past survival and reproduction are partly a result of these two factors: the capacity to know and the knowledge we have acquired. Briefly, evolutionary epistemology considers adaptation as a form of knowledge and the skill to obtain knowledge as an adaptation, which has evolved biologically. In this work we will consider exclusively evolutionary epistemologies belonging to the EET programme. These evolutionary epistemologies revolve around what we refer to as ‘the evolutionary analogy’, the claim that scientific change is governed by the same mechanisms, or by mechanisms analogous to those at work in organic evolution. As we already pointed out in the previous section, most evolutionary epistemologies focus on the development of selectionist models to explain the growth of human knowledge. However, an important distinction is needed. On the one hand, some scholars use metaphors of various complexities and, basically, argue that science changes by a mechanism analogous to natural selection. Roughly, they regard scientific change as the repetition of a two-step process. A pool of competing intellectual variants is produced in the first step, as a response to certain conditions, such as the presence of an unresolved research problem. At the second step a selection process establishes which variant(s) will ‘survive’. We will call the evolutionary epistemologies based on this view weak 16

By Campbell, for instance. He will later change his mind though (see section on Campbell in the next chapter).

Evolutionary Analogy

7

evolutionary epistemologies. On the other hand, others claim that science evolves by natural selection: The evolution of scientific knowledge is, in the main, the evolution of better theories. This is, again, a Darwinian process. The theories become better adapted through natural selection: they give us better and better information about reality (they get nearer and nearer to the truth). All organisms are problem solvers: problems arise together with life. (Popper 1984)

We will call those based on these claims strong evolutionary epistemologies. Hull’s work, in particular, is also oriented towards the identity of natural selection and ‘scientific selection’. He develops an analysis of evolution through natural selection which applies to biological, social and conceptual evolution (Hull 1982, 2006). Thus, weak evolutionary epistemologies consider natural selection as a metaphor which may be able to illuminate our understanding of the change of knowledge over time; strong evolutionary epistemologies argue that natural selection is an all-purpose ‘invisible hand’ theory, able to elucidate the emergence of co-ordination and design without calling for the presence of designers. The most common objections to the EET program are probably the following three (Bradie 1997: 400). Firstly, although an apparent design can be detected in the Universe, a possible mechanism for the development of intelligent life in the absence of evolutionary goal and direction was provided by Darwin’s theory and its more recent versions. Life evolution, however, is often pictured using the image of a branching tree and science could not make any use of this representation, since science is goal-directed and appears to proceed along completely different lines: “convergence and unification are often taken to be marks that we have the science right. This hoped-for and anticipated unity at the end of science is a vision inconsistent with a truly selectivist approach” (Bradie 1997: 401). Secondly, many theorists of EET consider the production of conjectures to be a ‘blind’ process, followed by a selective stage. It is often objected that this cannot be the case: the hypotheses generated by scientists are not random, since they have been formulated in order to solve particular problems. The third point concerns the progressive success of science, seen as a sort of ‘fit’ between knowledge and the world. In evolutionary biology, however, it seems impossible to identify any sort of general progression of the whole biological realm, because there is no fixed environment to fit even for a single species.

8

Chapter One

In this volume we will elaborate on these and other issues that arise from strong evolutionary analogies, claiming that science evolves by natural selection. This new elaboration is needed because, in the literature available, most of the attempts to criticise evolutionary analogies have erroneously ignored the aims and scope of the individual proponents. For instance, objecting that the production of novel theories is not random in the same way as mutations in the gene pools are is relevant only if the evolutionary analogy being criticised: - explicitly states this assumption or - is drawing an analogy which critically but implicitly depends on this assumption or - is claiming that the processes of genetic mutation and theory variation are the same. It is important, thus, to tailor analyses and criticisms on the scope of those epistemologies. More specifically, in each case the analysis will need to take into account whether the processes of biological and scientific development have been considered, by the proponents, the same process or analogous processes in a metaphorical discourse. Clearly, thus, what is needed at the next step is a theory of analogy and metaphor to be used for the evaluation of evolutionary epistemologies. In other words, a theory which would provide a means for measuring the soundness of the claims of such epistemologies on intersubjective grounds. In the next section we shall introduce some relevant concepts from contemporary theories of metaphor and produce the basis for an evaluative framework. In the third chapter we will expand what we believe is the most advanced theory of metaphor to solve the problem of deciding whether two processes can be regarded as the same process.

1.3 Metaphor Some history The first extended philosophical treatment of metaphor is given by Aristotle, who considers metaphors a potent tool to enable the understanding of hidden truths. In Poetics metaphors are described as giving the thing a name that belongs to something else; the transference being either from genus to species, or from species to genus, or from species to species, or on grounds of analogy. (Aristotle 1941a: 1475b)

Evolutionary Analogy

9

The main issues emerging from his works create points of reference for all future debates on metaphor. First, the metaphoric transfer is located by Aristotle at the level of individual words. Only in the twentieth century has the metaphoric transfer been situated at the level of sentences, because scholars realise that the semantic unit is larger than a word (Ricoeur 1977). Secondly, metaphor is seen as deviance from normal usage of language. Aristotle maintains that “diction becomes distinguished and non-prosaic by the use of unfamiliar terms, i.e. strange words, metaphors, lengthened forms, and everything that deviates from ordinary modes of speech” (Aristotle 1941a: 148a). Thirdly, metaphors are based on similarity relations: Aristotle writes: The greatest thing by far is to be a master of metaphor. It is the one thing that cannot be learnt from others; and it is also a sign of genius, since a good metaphor implies an intuitive perception of the similarity in dissimilars. (Aristotle 1941a: 1459s)

Furthermore, he maintains that there are two kinds of metaphors: proper ones, which show things in a new light and bad ones, which are obscure. Aristotle explains: Metaphors, moreover, give style clearness, charm, and distinction as nothing else can: and it is not a thing whose use can be taught by one man to another. Metaphors, like epithets, must be fitting, which means they must correspond to the thing signified: failing this, their inappropriateness will be conspicuous. (Aristotle 1941b: 1405a)

Lastly, he relates metaphor to simile: The Simile is also a metaphor; the difference is but slight. When the poet says of Achilles that he leapt on the foe as a lion, this is a simile; when he says of him ‘the lion leapt’, it is a metaphor – here since both are courageous, he has transferred to Achilles the name of ‘lion’. (Aristotle 1941b: 1406b)

Aristotle’s point of view has greatly influenced the way in which metaphors have been considered through history, and is the basis for the ever-recurring view that metaphor is an elliptical simile. This is the so called ‘Traditional View’ (Johnson 1981), in which metaphor is considered valuable for didactic purposes and for stylistic and rhetorical goals. In this view, metaphors can be translated into literal paraphrases without losing cognitive content.

10

Chapter One

Empiricist philosophers mistrusted metaphors, viewing them as confusing devices which mislead by making unclear the categorical distinctions between words. Metaphor is regarded as a matter of extraordinary language, a rhetorical flourish. Locke writes: “[a]ll the artificial and figurative applications of Words Eloquence hath invented are for nothing else but to insinuate wrong Ideas, move the passions and thereby mislead the Judgement” (Locke 1975: 508). Hobbes fears that the transfer of a name is to be expected to mislead those who consider the name as signifying just the original object. He writes: “metaphors, and senseless and ambiguous words, are like ignes fatui; and reasoning upon them is wandering amongst innumerable absurdities” (Hobbes 1962: pt I, chap. 5). He also points out that metaphors can generate an incorrect way of understanding. He considers the human conceptual system to be essentially literal and points out that ‘words proper’ are adequate to express meaning, as a metaphor’s meaning – when it has one – is its literal paraphrase. This stance and variations on it was dominant until quite recently, when the empiricist view was challenged by a series of novel approaches.

Metaphor as a cognitive structure Current research on the theory of metaphor approaches its subject as a linguistic or as a cognitive phenomenon. Here, our goal is an analysis of metaphor capable of providing evaluative tools for the assessment of evolutionary analogies. Thus, metaphor as a linguistics phenomenon does not offer any insights we could make use of. 17

17

The route of approaching metaphor by tackling it as a linguistic phenomenon is taken by Donald Davidson. Davidson claims that when we hear the statement ‘a man is an island’ it is naturally assumed that the statement is a metaphor, since it is really clear to everybody that a man is not an island. Davidson does not regard positively any perspective that regards metaphors as creators of new meaning; metaphors have just one meaning (the literal one). He states that “metaphors mean what the words, in their literal interpretation, mean, and nothing more” (Davidson 1978: 32.). Metaphors are utilised to get the hearer to understand or see something differently. They are a matter of pragmatism; they lead us to become aware of something that otherwise we would not have noticed: “I depend on the distinction between what words mean and what they are used to do. I think metaphor belongs exclusively to the domain of use. It is something brought off by imaginative employment of words and sentences and depends entirely on the ordinary meanings of those words and hence on the ordinary meanings of the sentences they comprise” (Ibid., p.33).

Evolutionary Analogy

11

The cognitive approach is largely adopted in linguistics and some scholars claim that we grasp the world only as mediated through our metaphors. Metaphor is undeniably omnipresent in our everyday life, because the nature of our ordinary conceptual system is fundamentally metaphorical: The locus of metaphor is not in language at all, but in the way we conceptualize one mental domain in terms of another. The general theory of metaphor is given by characterizing such cross-domain mappings. And in the process, everyday abstract concepts like time, states, causation and 18 purpose also turn to be metaphorical. (Lakoff 1993: 203)

George Lakoff and Mark Turner, among others, argue that metaphors are conceptual mappings: metaphor is “primarily a matter of thought and action and only derivatively a matter of language” (Martinich 1984). The presence of metaphors in our patterns of thought, thus, causes their appearance in our language. Nearly always, when we talk about abstract concepts, we choose language drawn from one or another concrete domain. A good example of this is our talk about the mind. Here we use the spatial model to talk about things that are clearly nonspatial in character. We have things “in” our mind, “on” our minds, “in the back corners of” our minds. We “put things out” of our minds, things “pass through” our minds, we “call things to mind”, and so on. It is quite possible that our primary method of understanding nonsensory concepts is through analogy with concrete experiential situations. (Rummelhart 1993: 71)

We come to new knowledge in two manners: either through direct physical experience with our environment or through metaphorical understanding built upon some initial direct physical experience. For instance, we have the direct experience of the concept ‘up’ and happiness is grasped metaphorically through the metaphor ‘happy is up’. This understanding is reflected in different expressions, such as ‘I’m feeling up today’ or ‘I’m high as a kite’. In Lakoff and Johnson’s perspective, metaphor is directional, since more abstract concepts are metaphorised and grasped in terms of less abstract ones. In the metaphorical pairing one concept is always better delineated and typically more concrete than the other.

18

For example on the topic see Lakoff and Johnson (1980 and 1999) For a different perspective from Lakoff and Johnson see G. L. Murphy ( 1996 and 1997).

12

Chapter One

These features of asymmetric reduction of abstract to concrete is mirrored, as we shall show, in the use of metaphors and models in philosophy or science to handle the unknown and the excessively complex through the known and relatively simple.

How do metaphors work? Currently, many theories of metaphor exist: emotive, tension, substitution, anomaly and interactive theories among the others, but only a few are relevant to this context and will be discussed. In particular, because the evolutionary analogies are introduced by their proponents to suggest a possible model for scientific change based on a range of analogies with biological evolution, the most appropriate theories to take into consideration are those addressing, directly or indirectly, the role of metaphor in science. This seems reasonable not because the subject of the investigation is an aspect of science, namely its change, but because the use of metaphors by philosophers in this field corresponds to the use of metaphors by scientists in their research. Evolutionary epistemologists are confronted with a complex phenomenon to explain and, like scientists, they seek help in the dynamic of another, more extensively – or at least, more successfully – studied phenomenon which, due to some of its features, resembles the one to be explained. This is the way models help scientists to tackle their problems.19 Eleonora Montuschi (2000) distinguishes two main views in the interpretation of the role of models in science: the comparison view and the interactive view. The comparison view is the approach of the logicist tradition, where models are an elliptical form of simile.20 Their value is purely illustrative and perhaps didactic and, as such, epistemically void and dispensable. In this view, for example, the sentence ‘science changes by natural selection’ should be rephrased as ‘scientific change is like change by natural selection’, meaning that the mechanism we know is at work in scientific change resembles that called ‘natural selection’. In this sense, the similarity is only intended as a means of clarifying – through a model that we assume is more familiar to the listener – a mechanism which is new or more difficult to grasp. The model, however, can be 19

We will not linger, however, upon the omnipresent question about models and metaphors in science, namely their disposability or necessity, an interesting question but quite separate from our main focus. 20 In a simile the similarities involved in the comparison are clearly defined and terms such as ‘like’, ‘as’ or ‘not unlike’ are present in the statement of the comparison.

Evolutionary Analogy

13

forgotten once the main subject has been well understood and formalised. This view is appropriate, in this context, when an evolutionary epistemology is proposed with no deep commitment. For instance, we shall show that this is the case with Kuhn’s analogy. In his case, the limited scope and the ephemeral nature of the metaphor does not require a criticism of the adequacy of the analogies as full correspondence of features between the two domains, because the analogies are meant to be only illustrative. Any evaluation thus, can be only based on the grounds of correctness, for instance on the right interpretation of the concepts in the two domains.21 The interactive view on metaphors was originated by Max Black in his famous article on metaphors (Black 1962). According to Black a metaphor is composed of two parts: the primary subject and the metaphoric secondary subject, which approximately corresponds to Richards’ tenor and vehicle22 and to what was formerly regarded as the ‘original idea’ and ‘the borrowed one’. The interactive view, going beyond the formalist and purely logical interpretation of scientific theories, incorporates the dynamism of language into the complexity of scientific change. From Max Black’s perspective (1962), analogies are dynamically created by the metaphors used by the speakers to put two subjects side by side. In developing this, Mary Hesse (1966) proposed a detailed analysis of the use of models in science, arguing that in developing or refining a theory scientists use a familiar system with a well consolidated and relatively unproblematic theory – the explanans – in order to model an unfamiliar system, the explanandum. They then draw comparisons and select relevant features from the familiar system to be transferred to the explanandum. The successful transferrals – that is, the properties of the model also displayed by the explanandum – will constitute the positive analogies between the system to be characterised and its model, while the unsuccessful ones will increase the number of negative analogies. However, some transferrals will be initially undecided, that is, it will be unclear or just unknown whether some properties of the model also belong to the explanandum. These undecided analogies constitute the set of 21

A different version of the comparison view claims that metaphors are based on a literal analogy of this form: A is to B as C is to D. Metaphors, thus, are elliptical analogies (isomorphisms between the entities of two systems) and not elliptical similes. 22 The word ‘tenor’ together with the term ‘vehicle’ were introduced by Richards (1936). Tenor is the underlying idea or the principal subject, while the vehicle is what is attributed metaphorically to the tenor. For instance consider the case ‘men are wolves’; ‘men’ is the tenor and ‘wolves’ the vehicle.

14

Chapter One

neutral analogies which provide scientists with further material for research, and it is their potentially promising content that constitutes the fruitfulness of a model, that is its provision of research interest.23 Analogies are employed in science to promote understanding of concepts. They do so by indicating similarities between these concepts and others that may be familiar or more readily grasped. They may also suggest how principles can be formulated and a theory extended: if we have noted similarities between two phenomena (for example, between electrostatic and gravitational phenomena), and if principles governing the one are known, then, depending on the extent of the similarity, it may be reasonable to propose that principles similar in certain ways govern the other as well. (Achinstein 1968: 208-9).

In the interactive view, models are an essential constituent of scientific analysis for two main reasons. The first one is the ‘interactive’ element of this position. Black pointed out that, rather than unveiling pre-existing analogies, metaphors create analogies due to the interaction between the two subjects of the analogy. Additionally, this mainly linguistic analysis was used by Hesse to claim that the relation between explanans and explanandum is also a dynamic one of mutual adaptation, so that both are modified by the use of metaphors. The second reason why models are essential in scientific discourse is the openness created by the neutral analogy, which instigates scientific investigation. This view is appropriate for the analysis of evolutionary epistemologies which claim to be explicative. In this case it will be legitimate to point out that allegedly positive analogies are in fact negative and that crucial neutral ones are also negative. As a whole, however, the epistemologies will be hard to evaluate due to a single difficulty, namely the issue of weighting positive and negative analogies: if the final judgment on the fitness of an analogical discourse is to be based on the predominance of positive analogies over negative ones, how do we measure this predominance? In order to do so, we need criteria to decide the relevance of analogies within positive and negative groupings, and then finally the whole set. This is because relevant negative analogies might be balanced by a higher number of relevant positive analogies, or by less numerous but much more relevant ones. For example, if it is agreed on the basis of investigation that intentionality and progress belong to the negative analogy between scientific change and natural selection while, say, fitness and replication belong to the positive one, it is arbitrary to 23

See Callanan (2008) for a clear analysis of the inductive power of analogical arguments in Kant.

Evolutionary Analogy

15

decide whether progress or intentionality is the worst disanalogy or whether they both are irrelevant when compared to the strong positive analogy. This arbitrariness, in turn, is due to the treatment of similarity as an unanalysed primitive (Way 1994: 151).

Type Hierarchies In the type-hierarchy approach (TH hereafter) the traditional analysis of similarity is reversed: “two things are similar in that they are [instances of] subtypes of the same supertype(s)” (Aronson et al. 1994: 105).24 The basic idea is that we do not define the type of the object as the class of all objects similar to it; rather we use types as primitives and define two things as similar with respect to a certain theory, represented by a TH, if they both belong to the same type. The TH approach is based on the general features of semantic networks. Although some authors are more stringent in their definition (Sowa 1984: 76), a semantic network is a set of nodes and arcs, sometimes labelled, connecting some of the nodes. The nodes usually represent objects, concepts or situations, the links represent their relations (Way 1994: 97). This general notation can be and has been used for many purposes, usually in linguistic analysis, artificial intelligence and knowledge representation. The particular type of semantic network used in TH is a hierarchical (i.e. directed, with the direction implicitly specified by the vertical or explicitly by using arrows instead of arcs), acyclic (no nonempty paths start and end in the same node) network of types or concepts, where the hierarchy models the level of generality or abstraction of the nodes.25 A type hierarchy is a complex network of types or concepts which are organised according to levels of generality, where the concepts get more abstract as one moves up the hierarchy. The nodes at the lowest levels of the hierarchy denote specific individuals or tokens, while nodes at the higher levels denote categories of individuals or types (Way 1994: 97).

And the types, in the approach of Aronson et al., are representations natural kinds, defined as “a set of entities which have a certain cluster of properties in common, that cluster being fixed by the natural laws

24

Our addition, in this quotation, reveals an instance of the sporadic lack of tightness in the language of Aronson et al. This, however, does not affect the merit of their arguments. 25 See also section 3.2 of (Sowa 1984) for some more formal definitions.

16

Chapter One

appropriate to the case”26 (Aronson et al. 1994: 39). Due to the difference in nature between nodes in the lowest levels with no subnodes (representing tokens, individuals, instances) and higher nodes representing types, the arcs connecting tokens to types are assigned an instance_of meaning, while the arcs connecting types are assigned an is_a or a_kind_of meaning. Embedded in the hierarchical arrangement is the property of inheritance: all properties and relations of any given type in the network are inherited by all of its subtypes, that is, the types at a lower level connected to it.27 It is worth noting, however, that types are intensional, they do not satisfy a principle of extensionality. In this they are opposed to sets, statements about which are verified by observing the extension.28 A physical example is the type hierarchy of harmonic oscillators as shown in Figure 1.29 Here different instances of harmonic oscillators inherit from different combinations of physical systems, each with its own set of specific laws. As a result, each oscillator ends with a specific law for its period and similarity is a result of this inheritance. For example, the oscillators that are also oscillating in a gravitational field have their period proportional to the inverse of the square root of the acceleration due to gravity (1/√g). Perhaps a more suitable example to explain how this approach is immune from the main difficulties that beset the comparison view would be a very simple type-hierarchy for two physical systems traditionally considered analogous, the solar system and the atom (Way 1994: 145). 26

The term ‘set’, here, has been used in the general sense of ‘group’ and not in the specific sense of set theory. 27 Insights into the subtype-supertype relation in a TH could be captured, according to Way (1994: 191-4), by the five conditions of Searle’s relation determinatedeterminable with the addition of a sixth condition of inheritance. 28 (Sowa 1984: 82). The distinction is explained by Sowa with the following example: “To say that the intersection of the set CAT with the set DOG is empty simply means that at the moment no individual happens to be both a dog and a cat. A biologist might examine litters of puppies and kittens looking for an exception. But to say that [the types DOG and CAT are disjoint] means that it is logically impossible for an entity to be both a dog and a cat. Any mutant that might arise could not falsify the statement; it would just force the biologist to invent a new type” (Idem). 29 (Aronson et al. 1994: 41). There is a problem in the way the nodes are labelled in these diagrams by Way. For example, it can be argued that ‘satellite’ is not an instance of ‘harmonic motion’, while ‘satellite motion’ would be. Analogously, ‘central force’ (in the next diagram) is not a kind of ‘complex physical object’, hence our slightly modified labelling.

Evolutionary Analogy

Figure 1 – A type hierarchy of harmonic oscillators

Figure 2 – A type hierarchy for atoms and the solar system

17

18

Chapter One

The example in Figure 230 shows a clear-cut, intersubjective way of comparing, say, the solar system to a group of cars racing around a house in elliptical tracks (Way 1994: 92). Although there are evident similarities between the two systems, it is obvious that these similarities are quite trivial when compared to the similarities between atoms and planetary systems. But how do we decide this? How do we decide that the absence of a central force for racing cars is a relevant disanalogy with planets? The TH approach solves this problem by explicitly appealing to what we think are the laws of nature, to the way we believe different systems obey these laws, and by identifying a system by its location in the semantic network (Way 1994: 198). Thus, the behaviour of both the atomic and the solar systems can be described by using the laws of an attractive central potential, proportional to 1/r, of momentum conservation, which are all properties that can be attached to the type Central force.31 In contrast, the system of racing cars is not described by those laws: in this case friction, thermodynamics and human skill are involved. The difference, ultimately, is then based on our understanding of the world32 and, as such, is based on the current status of science. Once the systems we are investigating are slotted into what we think are the correct places of a type-hierarchy – which could take some effort and may change over time, given the empirical nature of the task – the analogies “fall out as a result of the inheritance relation and the laws and properties of the supertypes” (Aronson et al. 1994: 109). Subsequently, there is no need to identify and ‘weight’ the negative analogies, because they are not represented at the level of the supertypes. Thus the question of what is similar to what and in which respects is meaningful only relative to some type-hierarchy. In any system some supertypes will be more important or salient than others. The similarity mapping is dependent upon the topology of the hierarchy, and the content and structure of the hierarchy in turn is dependent on an understanding of the world. (Aronson et al. 1994: 106)

Way (1994: 129) sees similarity as an abstraction of some of the properties found in the systems being compared. The similarity relation 30

‘Nucleus’, in this simplification, is considered to be as simple as ‘electron’. This font will be used to refer to labels of nodes in the diagrams. 32 And on the particular perspective adopted by the type hierarchy. For instance, this example shows that a system of racing cars and a system of planets are very different if we use a type hierarchy describing dynamic systems. If we adopted a purely cinematic point of view, however, the corresponding type hierarchy would bring the systems close together. 31

Evolutionary Analogy

19

then, is a derived relation because it depends on the TH that has been chosen to rank the entities we are dealing with, which is eventually an empirical matter because it depends on the way we view the world.33 As a result, we still need to analyse the TH we believe to be most suitable for describing the chunk of the world we are focussing on by, for example, labelling the type-nodes with the properties we believe they possess, and testing our hypothesis against what we know already. The TH approach is the best suited to analyse those evolutionary analogies that claim to be deeply explicative and those which, going beyond mere analogy, argue that the evolutionary processes at work in scientific and biological change are the same. In the third chapter we will analyse such strong evolutionary epistemologies and extend the TH approach to introduce a non-arbitrary method to evaluate their claims. In the next chapter we shall restrict ourselves to previous evolutionary epistemologies.

33

This may seem quite subjective. ‘We’, however, is in most cases referring to the scientific community or communities which have been exploring that chunk of the world. They carry, then, the same amount of subjectivity as can be attributed to scientific theories.

CHAPTER TWO EVOLUTIONARY ANALOGIES

In this chapter the following issues will be addressed: How have the most influential proponents of evolutionary analogies defended their positions? What criticisms can be expressed against their arguments? What points of strength can be identified in their positions and/or in their methodologies, which could be retained for the foundation of an improved and more defensible position? The absence of the most popular and popularised philosopher of science, Karl Popper, who was a persistent defender of evolutionary epistemology, will be noticed. This absence is not without a reason. Popper may have been persistent in his defence but he was neither profound nor consistent. His use of the terms ‘evolution’ and ‘natural selection’ were very general and kept oscillating between EET and EEM approaches. In our view he never provided very deep insights on this subject, at least when compared to the authors here presented and analysed. In brief, although his work has been occasionally referenced in this volume, full analysis of his position would add very little to the material we are presenting in this chapter, at least for our aim.

2.1 Kuhn Kuhn’s model of scientific development has been extremely influential and any scholar seriously interested in scientific change, either from a normative or descriptive perspective, usually feels the need to clarify her position in relation to Kuhn’s view. His Structure of Scientific Revolutions (SSR) is probably “the most widely cited work on the nature of science in the twentieth century” (Fuller 2001: 565) and Kuhn, in its final passages, shows interest in an evolutionary treatment of scientific development. He revises and refines his position in The Road Since Structure, where he gives indications of a more articulate position which was to be expounded upon in a later book. Kuhn did not live long enough to complete his evolutionary epistemology project. What he wrote, however, is of valuable stimulus to reflection on key issues discussed in the remainder of this

22

Chapter Two

volume, such as the role of progress in evolutionary change and the form of conceptual continuity which an evolutionary epistemology seems to imply. In contrast to other authors we will analyse, Kuhn did not make any attempt to build a generalisation of selective processes, or of any other process or concept found in evolutionary biology, with the aim of showing that such generalisations would also capture analogous features in scientific change. Instead, he preferred to focus on individual evolutionary concepts, such as ‘mutation’, ‘speciation’ or ‘ecological niche’, in order to show similarities in analogous concepts he claimed he identified in the development of science. The purpose of his focus was, in our view, to support controversial aspects of his thought by means of the evolutionary analogy. In this section we will show that he could support his analogy only because his interpretation of key evolutionary concepts is faulty.

Scientific development The general account of scientific development given by Kuhn is well known and has been extensively analysed and criticised. Kuhn himself applied several adjustments over time in reply to these criticisms. Only a very brief summary will be given here, in order to place the target of our criticism – his evolutionary analogy – in the wider context of his approach. SSR is generally known for detailing the distinction, in the development of scientific disciplines, between alternating phases of ‘normal’ science and ‘revolutionary’ episodes.1 During the normal periods of a scientific 1

The metaphor of revolution is, according to Shapin, now rooted in tradition; the expression ‘scientific revolution’ was not common until Alexandre Koyré used it in 1939. In 1954 it appeared in two books written by authors who belong to very different traditions: Rupert Hall’s The Scientific Revolution was influenced by Koyré and The Scientific and Industrial Revolution, written by J.D. Bernal, who was much influenced by the Marxist tradition. The word ‘revolution’ in that period took on a new meaning: if earlier the word ‘revolution’ was related to the idea of an uninterrupted and periodically recurrent cycle, from that point onwards the meaning changed completely and came to indicate an abrupt break. In Copernican astronomy planets complete revolutions around the sun (again we have the idea of an uninterrupted and recurrent cycle). Even the use of the term in politics, i.e. political revolution, was rooted in the idea that history is formed by ebbs and flows or cycles. The use of the term ‘revolution’ now does not indicate a cycle, but a break, the starting of a new state of things, something that had never occurred before historically. According to Shapin, this new meaning associated with the word ‘revolution’ and its use in scientific contexts has its origins in the French Enlightenment. He argues that it is possible that the notion of revolution as total,

Evolutionary Analogies

23

discipline, scientists are busy solving the puzzles arising from empirical results of the current paradigmatic ‘view of the world’ of that discipline, and from the need for local adjustments in the theoretical framework to account for minor anomalies. When the current paradigm begins failing to fulfil its purpose of puzzle solving, and the unresolved puzzles turn into a considerable number of unexplained anomalies, a revolutionary episode occurs after which the whole paradigm is substituted with a new one. In Kuhn’s view, some features of this change of paradigm are analogous to the features of both a Gestalt switch and a language change (SSR: 111-4) and as a consequence it is impossible to map all terms from the old one onto the new one or vice versa. In particular, when translating from one natural language into another, some terms are simply impossible to translate into corresponding terms, because a certain language sometimes lacks the term for a concept expressed by a term in a different language. An analogous phenomenon occurs in science between different paradigms. A new paradigm will also bring different parameters for the identification of puzzles to solve. Although some of the previous puzzles will be solved by the revolution, others will simply dissolve and will not be present in the new phase of normal science, not even as ‘solved’ puzzles. In one word, the difference between the theories belonging to different paradigms is called by Kuhn ‘incommensurability’. The gap implied by incommensurability, however, is not so wide as to imply the noncomparability of theories belonging to different paradigms. Theories can be compared in their capability to solve problems, and the increased power of successive theories in this task shows the progressive nature of scientific development. Kuhn is also convinced that this progression is enough to save him from the accusation of relativism, as he eventually specifies in the Postscript: [S]cientific development is, like biological, a unidirectional and irreversible process. Later scientific theories are better than earlier ones for solving puzzles in the often quite different environments to which they are applied. That is not a relativist's position, and it displays the sense in which I am a convinced believer in scientific progress. (SSR: 206)

epochal and irreversible change was first applied in the field of science and was only applied to political change at a later date. The first revolution was scientific, while the others (the French, American and Russian Revolutions) are its offspring. See the Introduction of Shapin (1996).

24

Chapter Two

Progress from: the evolutionary analogy in The Structure of Scientific Revolutions In the main text of The Structure of Scientific Revolutions, Kuhn maintains that there is much equivocation about the nature of progress in science. Scholars who argue in favour of scientific progress conceive of science as progressing toward something, commonly closer to the truth. His notion of scientific knowledge, however, does not allow for such a directed progression. There is no neutral point of view from which we could assess the ‘distance’ between our theories and the truth, and in Kuhnian terms this very statement does not even make much sense. This is reinforced in the Postscript: “the notion of a match between the ontology of a theory and its ‘real’ counterpart in nature now seems to me illusive in principle”. The paradigm is the world the scientific community lives in and the only interface between this and nature is the experimental apparatus used to test the proposed solutions to scientific puzzles. When the experiments or the observations confirm that the puzzle has been solved, there is no implication for the representation of nature implied by those theories, that is for the ontology they depict (SSR: 206).2 However, says Kuhn, there is a different kind of progress that science shows going through its revolutions and theory changes. It is the progress from something – that is from a certain set of tools to discover and solve puzzles to a better one. There is no direction towards anything as such, no goal, only (indefinite) improvement from the status quo. In our interpretation, Kuhn then introduces the evolutionary analogy with the intention of showing that this view of progress is defendable and tenable: it is possible and even plausible to regard science as progressive in the absence of a fixed goal because, he thinks, biological evolution is also progressive but non-directed. However, focussing on this defence does not provide a fruitful analogy. In SSR he observes that scientific change is comparable to a biological process: it is irreversible, it proceeds from a starting point and it has no teleological goal. Can we not account for both science’s existence and its success in terms of evolution from the community’s state of knowledge at any given time? 2

In his later work, Kuhn (2000) replaces his linguistic metaphor for ‘paradigm’ with a less problematic explanation in terms of taxonomies of natural kinds. These taxonomies radically change through revolutions, but yet progress is not measured by the taxonomies getting “closer and closer to matching the universe’s ‘own’ taxonomy” (Sharrock and Read 2002: 187).

Evolutionary Analogies

25

Does it really help to imagine that there is some one full, objective, true account of nature and that the proper measure of scientific achievement is the extent to which it brings us closer to that ultimate goal? If we learn to substitute evolution from-what-we-do-know for evolution-toward-whatwe-wish-to know, a number of vexing problems may vanish in the process. (SSR: 171)

Kuhn maintains that natural selection and revolutionary selection – that is the process which accounts for the ‘survival’ of only the new paradigm between competing ones – produce analogous results, since both choose the most viable of all possible alternatives, that which is most ‘fit’ for a particular, historical situation: the resolution of revolutions is the selection by conflict within the scientific community of the fittest way to practice future science. The net result of a sequence of such revolutionary selections, separated by periods of normal research, is the wonderfully adapted set of instruments we call modern scientific knowledge. (SSR: 172)

When a period of crisis occurs in a scientific community, he adds, the emergence of alternative theories might be seen as the occurrence of ‘mutations’, among which the most promising will survive.3 Not much more seems to be implied by this analogy, at least in SSR, where Kuhn actually warns against easy generalisations. The following is a quotation of the whole passage: The analogy that relates the evolution of organisms to the evolution of scientific ideas can easily be pushed too far. But with respect to the issues of this closing section it is very nearly perfect. The process described in Section XII as the resolution of revolutions is the selection by conflict within the scientific community of the fittest way to practice future science. The net result of a sequence of such revolutionary selections, separated by periods of normal research, is the wonderfully adapted set of instruments we call modern scientific knowledge. Successive stages in that developmental process are marked by an increase in articulation and specialization. And the entire process may have occurred, as we now suppose biological evolution did, without benefit of a set goal, a permanent fixed scientific truth, of which each stage in the development of scientific knowledge is a better exemplar. (Kuhn 1970: 172-3)

Here, Kuhn does not elaborate further on the term ‘selection’. He refers to a process analogous to NS but he does not go into any detail and 3

This view is quite close to Popper’s evolutionary epistemology (Popper 1978).

26

Chapter Two

we will not comment on it: Kuhn’s passage only shows that, in science, a mechanism analogous to biological evolution is in place which accounts for the adaptation of our knowledge and practices to the problem-solving activity. The necessary outcome of this process is, for Kuhn, the specialisation of scientific disciplines, an idea that will be further developed in The Road Since Structure (RSS, see below). However, Kuhn’s analogy already suffers from several defects at this stage. Firstly, it is not clear what exactly the terms of the analogy are. He opens the passage by drawing an analogy between organisms and scientific ideas, mapping ‘mutation’ into ‘theory change’. It is not clear, though, what evolutionary lineages of organisms should be mapped onto, given the fact that revolutionary episodes would suddenly create novel ideas, concepts and problems, many of which would be incommensurable with their predecessors. If we allowed the hypothesis that new ideas emerge slowly, as the continuous adjustment of previous ones, we would acknowledge the existence of chains of concepts which would destroy Kuhn’s main standpoint of the incommensurability over revolutions. The main feature of scientific revolutions is the introduction of a discontinuity in the way the world is viewed and in the rules that dictate what a scientific problem is and is not. The emergence of these new view and rules would be analogous, in terms of Kuhn’s sketched analogy, to the emergence of a new form of life from basic organic constituents. The problem would not arise, however, if we accepted the emergence of life from scratch as the biological event analogous to the emergence of these novelties. Nevertheless, even if we managed to overcome the difficulty posed by the observation that all living species seem to descend from the same ancestors, suggesting that the emergence of brand new life has been a unique (at least historically) event on Earth, there would be another major obstacle. In the analogy, it would remain unexplained how new ideas would rise in their whole complexity, while new life would obviously start from the simplest forms. Kuhn also adds that selection operates on scientific practices, and here it is even more difficult to follow the analogy and relate scientific practices to anything on the biological side, at least with the few details that Kuhn provides. His mention, in this context, of a sequence of revolutionary selections separated by periods of normal research could remind us of the concept of punctuated equilibrium,4 while the increased specialisation in 4

“The hypothesis of [punctuated equilibrium] holds (among other things) that most evolutionary changes in morphology, although perhaps continuous in the sense of passing through many intermediate stages, have been so rapid that the

Evolutionary Analogies

27

separate scientific disciplines will be developed by Kuhn into his analogy with speciation in RSS.5 The lack of further detail, however, prevents any fruitful analysis along this line of reasoning. At the end of this almost casual list of analogies between biological and scientific development, Kuhn states the main point he is trying to make, i.e. the deceptiveness of conceiving a “permanent fixed scientific truth” as the goal of science. Alexander Bird (2000) analyses this argument by comparing Kuhnian scientific progress to two alternative models: single-species and twospecies evolving systems. In the single-species model, one species evolves in a fixed environment, e.g. giraffes in an environment of fixed height trees. In this case there exists an optimal height (or range of heights) for the giraffes to evolve to under selective pressure. In this (oversimplified) case it makes sense to say that the evolution of giraffes is getting them closer and closer to a certain goal, towards an optimal height for feeding. However, in a two-species model, which is still an oversimplification but closer to the actual situation in nature, trees evolve as well.6 In this case the changes in one species modify the selective pressure on the other and vice versa and mutual evolution is in place. For example, trees would be advantaged if on average they were too high to have their leaves eaten by giraffes. Higher trees, however, would increase selective pressure on giraffes to grow even further and so on. Bird also states that in an evolutionary analogy, the selective pressure on the giraffes would map into the selective pressure of experimental tests on theories. As the results of experimental tests do not change (experimental replicability is a ‘requisite’ for scientific activity, due to the stability of the reality we are testing against), then the single-species model is the most accurate analogy for scientific development. In Bird’s opinion, this is enough to show that Kuhn is incorrect: a properly designed evolutionary analogy cannot avoid progression towards truth. If a hypothesis passes a well-designed and well-executed experiment we are usually confident that it will continue to pass tests of that kind if repeated, and that any extension of the theory will thus also pass the test. In broad terms, the single-species model captures the idea that in science our theories may change but the features of the world that they respond to are what they are independently of our theories, and are by and large constant fossil record presents the appearance of a discontinuous change.” (Futuyma 1998: 677). 5 Because of these resemblances, Bird (2000: 212) finds the evolutionary analogy “certainly fruitful when considering the development of science”. No further argument is provided by Bird, though. 6 Bird’s two-species actual example is about cheetahs and gazelles.

28

Chapter Two over time. Recall that in the single-species model we could say that the giraffe is evolving towards an ideal and that there would be no pressure on the species to evolve once that ideal had been reached. Similarly we can say of an area of scientific research that it will evolve towards an ideal beyond which it cannot go. In the limit the giraffe reaches an optimal "fit" with its (fixed) environment; a theory can reach a optimal fit with the world, and this would be a true representation of it, since only true theories cannot be falsified. (Bird 2000: 213)

We find Bird’s argument over-simplified in its treatment of both sides of the analogy. On the science side, for instance, it is not generally true that a certain hypothesis will pass a certain kind of test forever. An improved technique might be developed to perform the same kind of test, for example by increasing the accuracy of measurements and causing the results to change.7 Newton’s laws passed all tests of celestial mechanics until some irreconcilable anomalies of Mercury’s orbit were found. The focus, of course, is on what is meant by “well” in the phrase “a welldesigned and well-executed experiment”. In Kuhnian terms the goodness of an experiment is relative to the current paradigm, although crossparadigm tests are possible (and are decisive for the conversion of scientists to the new paradigm). we would add, however, that only after a certain amount of time is it possible to judge the quality of a test, usually in relative terms when compared to new versions of the test which have been available for some time. And even the newly available tests may be called “well-designed” in a provisional fashion only. we agree with “the idea that in science our theories may change but the features of the world that they respond to are what they are independently of our theories”.8 So, our desires cannot change what we read on a galvanometer, to use Campbell’s example (Campbell, 1974). This, however, brings no guarantee to the claim that a good test will be always regarded as such across any

7

It could be objected that this should be regarded as a new test. However, Bird is rightly careful to generalise and state that tests of the same kind will never fail. Refined techniques to perform a certain test do not change the nature of it. Otherwise Bird’s statement would be reduced to the trivial observation that when a theory passes a test it will always pass it – if kept the same all the times – for ever, which is clearly not Bird’s view. 8 (Bird 2000: 213). And so would Kuhn: he argued that the world “is entirely solid: not in the least respectful of an observer’s wishes and desires; quite capable of providing decisive evidence against invented hypotheses which fail to match its behaviour” (RSS: 10).

Evolutionary Analogies

29

number of successive theories extending the original one.9 And this claim is the main underpinning of Bird’s view of theories evolving towards an optimal state. We also find Bird’s biological models so simple as to be misleading. For example, to add some reality to the two-species scenario, it would suffice to note that trees and giraffes could not sustain their growth indefinitely. A number of additional effects and selective pressures would emerge: structural stability of both tree trunks and giraffe necks; exposure of a larger surface to the external environment; higher visibility to predators, just to mention a few.10 It does not seem to be a good idea to find simplistic scenarios for evolutionary biology in order to show that the least unrealistic ones do not fit the analogy with science development, because these attempts fail as soon as it is objected that the simplification itself causes the whole argument to lose its rigour. In particular, a better, more complex biological model would provide Kuhn’s view with stronger arguments. For instance in the more realistic scenario it is still possible to think of some sort of asymptotic stability for the two species plus their environment. In other words, for some time the giraffes’ and the trees’ length might reach an equilibrium within the given environment. This could resemble Bird’s optimal fit, but only temporarily: changes in the given conditions might result in the trees to evolve in a way that would affect the giraffes’ evolution as well, until a new equilibrium is reached. This, of course, fits much better Kuhn’s account of periods of normal science intercalated by revolutionary episodes than the 2-species scenario.11 Although Bird’s general perspective is understandable – we believe that science does follow a direction in a way that cannot be explained or modelled as an evolutionary dynamic – we prefer to follow a different route to show the weaknesses of Kuhn’s evolutionary epistemology. In our view, Kuhn’s account of a ‘Darwinian’ progress in the development of science can be better attacked on two fronts. On the first front it can be argued that Kuhn’s notion of progress in evolutionary biology is oversimplified and does not coincide with the notion currently accepted by evolutionary biologists. As a result, the recovery of progress in his account of scientific change is based on mostly rejected assumptions. On the second front it can be shown that, even if we ignore this detail about progress, the more articulate account of scientific change exposed in 9

Moreover, Kuhn talks about selection of practices, not theories (RSS: 6). It is in the evolution of practices that progress is visible. 10 Analogous objections could be sustained for any other 2-species scenario. 11 Again, this is related to the hypothesis of punctuated equilibrium in evolutionary biology – see footnote 4 above.

30

Chapter Two

RSS is based again on a more general misunderstanding of the biological side of Kuhn’s analogy, namely the notion of ecological niche and related concepts. We will begin by discussing the problems in Kuhn’s notion of progress, leaving the second front for later, once the ecological niche has been clearly discussed.

Local and global progress In the passage from SSR quoted above, Kuhn refers to what we will call ‘global’ scientific progress. This progress, which coincides with an increased degree of puzzle-solving power, is measured across successive paradigms and revolutions. This global reach allows Kuhn to regard himself as a non-relativist, at least in the sense that there exists a general and permanent criterion – puzzle-solving power – which measures the quality of a paradigm in comparison with the previous one. The revolutionary events also coincide with a higher degree of articulation and specialisation of scientific disciplines. In RSS he refines this point and its analogy with biological evolution: in the living realm, over time new species spin off previous ones as a result of geographical isolation of populations, which eventually become reproductively isolated from the rest of the original species. A new species will be adapted to its own biological niche, which becomes narrower and narrower (RSS: 11).12 Analogously, after a revolution the new scientific disciplines will become more and more isolated from the rest of the scientific community (Ibidem, p. 7), until cross-communication becomes impossible. The specialisation, says Kuhn, is accompanied by greater articulation; in other words the new disciplines focus on smaller and smaller portions of knowledge which become more and more complex and detailed. The main problem, here, is that Kuhn’s account of progress in evolutionary biology is different from the account that most biologists would provide, which we accept. In SSR (see passage quoted above), he connects together the two notions of adaptation (“wonderfully adapted set of instruments we call modern scientific knowledge”) and complexification (“increase in articulation and specialisation”), but the two notions are sharply distinct in evolutionary biology. Radick (2000) points out that “’progress’ names a conviction that most if not all evolutionists share: that life began with extremely simple forms, and that increasingly more 12

The concept of biological (commonly called ecological) niche will be illustrated below. Here it can be understood as ‘local environment’, which we will show is the wrong meaning that Kuhn gives to the term.

Evolutionary Analogies

31

complex forms have evolved at intervals thereafter” (p.476).13 We would also call this kind of evolutionary progress ‘global’, because it relates to the whole history of biological life on Earth and is measured across successive species. However, “[s]o far as ‘progress’ suggests getting better at something, adaptation no doubt has a stronger claim to the term. Species more tightly fitted to conditions are ipso facto better at keeping alive and leaving descendants. Species more complexly organised are not necessarily better at anything” (Idem). We would call this kind of adaptive progress ‘local’, because it refers to the development of individual species in the context of given conditions. Clearly, Kuhn blurs this distinction between ‘local’ and ‘global’ progress and associates the articulation of disciplines (complexification) to their increased local success (adaptation). Kuhn (2000) points out that, in his view, “specialisation and the narrowing of the range of expertise [is] the necessary price of increasingly powerful cognitive tools”. We find this remark reasonable and we are ready to share it. If we represent our disciplinary scientific knowledge as a taxonomy of concepts, higher success in problem-solving only seems to be achieved when the taxonomy gets more complex, populated by new concepts and in new relations with each other, until the complexity of some portions of it makes the taxonomy unmanageable by a single group (better: by individuals belonging to a single group) and new disciplines arise, bringing further complexification. It is almost commonplace that powerful scientific practices are associated with high level of specialisation and complexity. The analogue of this association does not exist in biology, however; adaptation is uncoupled from specialisation and species can lose complexity over time. Radick quotes an example of a series of four insects: a springtail, a dragonfly, a beetle with wings, and a beetle without wings: Compared with the wingless springtail, the winged dragonfly is the more highly organized creature. Compared with the dragonfly, which has two more-or-less identical wing pairs, the winged beetle is more highly organized because one of the wing pairs has become specialized for protection. [The] beetle that has lost its wings altogether […] show[s] that evolution can reverse “the trend of greater differentiation and specialization” (Radick 2000: 476)

13

We will not embark in the discussion of how we can actually estimate or judge complexity in living beings. We will leave this notion as primitive and intuitively accessible.

32

Chapter Two

In particular, Radick (2000) also quotes empirical studies “which found descendant species as likely to be less complex or more complex than their immediate ancestors” (p.478).14 So, if we interpret evolutionary progress as greater complexity, evolution by natural selection does not explain it. If evolutionary progress is interpreted as adaptation to local conditions, then natural selection does explain it. On the scientific change side of the analogy, this means that a mechanism for scientific development analogous to natural selection would explain only local progress for individual disciplines. In the long term, however, specialisation and higher articulation would remain unexplained. However, local progress as a necessary result of natural selection should not be taken for granted either. Lewontin’s argument against adaptation (and natural selection) as an improvement process (Lewontin 1978: 159) is particularly convincing. He mentions Leigh Van Valen’s “Red Queen” hypothesis, so called after the character in Through the Looking Glass who had to keep running just to stay in the same place. Van Valen’s theory is that the environment is constantly decaying with respect to existing organisms, so that natural selection operates essentially to enable the organisms to maintain their state of adaptation rather than to improve it.

Apart from the evidence that seems to support the Red Queen hypothesis,15 we find it very reasonable because it only involves a simple principle of reaction or adjustment of the species to changes in its living conditions, instead of unexplained tendencies towards “improvement” or complexification. At first this may, in the context of Kuhn’s evolutionary analogy, seem the confirmation that science is “not pulled forward but pushed from the back”. Scientific change may not be guided by the tendency to find the truth but rather stimulated by the accumulation of unresolved anomalies. This is a standpoint which might be reasonable, but only if we do not require this stimulus to account for the adaptation and

14

This, of course, does not mean that we cannot find many species, such as our own, which are now highly complex or even others, such as most bacteria, which have never increased their complexity. 15 For instance, extinction rates of species appear to be constant within the group the species belongs to, whether the species is short-lived or has been in existence for a long time. If natural selection improves fitness, however, we would expect the former to be more likely to go extinct than the latter (Lewontin 1978: 159).

Evolutionary Analogies

33

complexification of scientific tools in analogy with evolutionary biology, because in evolutionary biology this does not seem to happen.16

RSS: adaptation and niches Bird’s aforementioned criticisms target an account of scientific change belonging to an ‘intermediate’ Kuhn, and then become obsolete if we take into account the more sophisticated picture of the evolutionary analogy provided in RSS. There it is clear that Kuhn is well aware of the simple distinction between one-species and two-species scenarios and goes far beyond it. He questions the stability of the ‘environment’ against which theories would be tested in a naïve evolutionary analogy, by including some more details from evolutionary biology, mainly Lewontin’s (1978) account of the adaptationist program and his views on the concept of ecological niche. In evolutionary biology, notes the philosopher, the adaptation of a species does not occur simply towards a given environment in which the species lives, but the environment itself evolves with the species: every species lives in a biological niche which changes with the species adapted to it. More specifically, the biological niche cannot even be said to have an independent existence from the adapted population, so Kuhn’s distinction is between “creatures within the niche together, on the one hand, and their ‘external’ environment, on the other” (RSS: 11). The biological idea of niche and this latter distinction are transported by Kuhn into the conceptual realm, to give substance to his idea of a phenomenal world which changes with interaction with scientists, the “external world” (the reality) staying the same: [T]he world is not invented or constructed. The creatures to whom this responsibility is imputed, in fact, find the world already in place, its rudiments at their birth and its increasingly full actuality during their educational socialization […]. That world, furthermore, has been experientially given, in part to the new inhabitants directly, and in part indirectly, by inheritance, embodying the experience of their forebears. As such, it is entirely solid: not in the least respectful of an observer's wishes and desires; quite capable of providing decisive evidence against invented hypotheses which fail to match its behavior. […] [W]hat people can effect or invent is not the world but changes in some aspects of it, the balance remaining as before. In both cases, too, the changes that can be made are 16

It is worth noting that Kuhn, while referring to Lewontin (1978) for his account of ecological niche, seems to ignore his arguments against adaptation (and natural selection) as an improvement process in the same article. We acknowledge, however, that some details of Lewontin’s views are still controversial.

34

Chapter Two not introduced at will. Most proposals for change are rejected on the evidence; the nature of those that remain can rarely be foreseen; and the consequences of accepting one or another of them often prove to be undesired. […] Niches may not seem to be worlds, but the difference is one of viewpoint. Niches are where other creatures live. We see them from outside and thus in physical interaction with their inhabitants. But the inhabitants of a niche see it from inside and their interactions with it are, to them, intentionally mediated through something like a mental representation. Biologically, that is, a niche is the world of the group which inhabits it, thus constituting it a niche. Conceptually, the world is our representation of our niche, the residence of the particular human community with whose members we are currently interacting. (RSS: 1011)

As Sharrock and Read (2002: 189) very clearly put it, in Kuhn’s conceptual view we can distinguish between an ‘external world’ of nature and a plurality of ‘phenomenal worlds’ that scientists inhabit. These phenomenal worlds are like ‘evolutionary niches’ and their diversity is no more surprising than the “now routinely accepted idea of ‘plurality of evolutionary niches’”.17 We are now ready to discuss the second criticism of Kuhn’s evolutionary analogy. After a brief account of what we believe is the correct interpretation of Kuhn’s own source for the concept of ecological niche (Lewontin 1978), we shall show this is not the interpretation given or implied by the philosopher in RSS and argue that his evolutionary analogy can be defended only on flawed evolutionary premises.

Ecological niches, localities and adaptation in evolutionary biology Lewontin (1978) points out that a clear-cut definition of ‘ecological niche’ is difficult. In general terms, an ecological niche is a multidimensional description of the total environment and way of life of an organism. Its description includes physical factors, such as temperature and moisture; biological factors, such as the nature and quantity of food sources and of predators, and factors of the behavior of the organism itself, 17

Sharrock and Read (2002: 189-191) see in the niche’s analogy an attempt by Kuhn to avoid charges of idealism, by blurring the distinction between subjective and objective. In this analogy the world is not in the individual mind but retains a collective reality. “It is group and group-practices that constitute worlds (and are constituted by them)” (RSS: 11). This analysis is not relevant to our aim, though.

Evolutionary Analogies

35

such as its social organization, its pattern of movement and its daily and seasonal activity cycles (Lewontin 1978: 158).

Focussing on measurable factors and closer to Evelyn Hutchinson’s original definition, Futuyma (1998: 68) represents a species’ ecological niche as the portion of a multidimensional space, where each dimension represents a factor relevant to the species’ survival, in which the organisms of that species can persist. If we take the example of an easy-torepresent bidimensional case of only two factors affecting the species’ life, such as the sizes of the prey the individuals can eat and the environmental temperatures they can tolerate, the niche would be represented by the clear rectangle in Fig.3. The solid rectangles represent the conditions in different localities, and the species can persist in a locality (n.1, in this case) only if this overlaps its niche. Over time, the locality will change and the species’ niche will follow those changes or go extinct. As the changes in the locality conditions are also caused by the presence of the species itself, it is correct to talk about the coevolution of niche – a population characteristic – and locality and not of the adaptation of a species to a given environment.

Figure 3 – Evolutionary niche with two localities18

18

Adapted from Futuyma (1998: 68).

36

Chapter Two

The concept of ecological niche, when connected to a new and more dynamic concept of adaptation, may seem very well-suited to accommodate all of Kuhn's requirements. The traditional adaptationist view is well rendered by Lewontin's words, which resemble Popper – and Kuhn’s earlier – positions: The modern view of adaptation is that the external world sets certain “problems” that organisms need to “solve,” and that evolution by means of natural selection is the mechanism for creating these solutions. Adaptation is the process of evolutionary change by which the organism provides a better and better “solution” to the “problem,” and the end result is the state of being adapted (Lewontin 1978: 157).

This simplistic19 view is surpassed by the more complex dynamicity of the coevolution of the binomial niche-environment and the interactive dimension implied by the ecological niches is used by Kuhn to support his idea of a changing world which is evolving together with the group which inhabits it. This is the main point of the passages from RSS quoted above, but we shall soon argue that the interactive dimension is misplaced in Kuhn’s interpretation of the concept of niche.

Kuhn’s confusions The first confusion that Kuhn shows in his argument, however, is between EEM and EET.20 At the beginning of the passage quoted, he argues that natural selection is partially responsible for our knowledge of the world, in the same way theorised by EEM.21 The “solid world” experienced by the “creatures born into it” is the (external) world of nature, the physical one we already find in place, and from his passage it is clear it is this world which causes the newborns to internalise such knowledge as 'fire is painful' or 'falling hurts'. Some of this knowledge, Kuhn notes, is already internalised by inheritance; for instance, in our example, by giving an evolutionary advantage to the creatures which do 19

Lewontin notes that there is no end to adaptation (p. 157). “There is a constant interplay of the organism and the environment, so that although natural selection may be adapting the organism to a particular set of environmental circumstances, the evolution of the organism itself changes those circumstances” (p.159). Clearly, by ‘environment’ Lewontin is referring to what Futuyma calls ‘locality’, that is the representation of the natural world local to the species in the multidimensional space in which its niche is also represented. 20 The distinction was introduced in section 1.2 Evolutionary Epistemologies. 21 A point of view which is not difficult to share.

Evolutionary Analogies

37

not feel comfortable when getting too close to a fire or which can count correctly and infer from seeing two tigers entering a cave and only one coming out, that it is not safe to approach the cave.22 Immediately beneath this, however, Kuhn switches to EET and changes the meaning of the term “world” without any warning. Namely, he mentions the changes to the world we can effect or invent: trivially, it is the phenomenal world we can alter, we cannot change the 'external world'.23 In particular, the changes we can make, which are not rejected on the evidence,24 belong to a conceptual level: “the world is our representation of our niche, the residence of the particular human community with whose members we are currently interacting" (RSS: 11). This confusion of meanings greatly obscures the structure of Kuhn’s argument. To contribute to the obscurity, Kuhn finds the analogue of this representational level also in the biological side of the analogy, in what he calls "something like a mental representation" (Idem). At this point, additionally, a break in the analogy synchronism occurs. As we have seen, he equates the changes we produce in our representations (scientific change) to the changes in the coevolving couple niche/group of organisms, not to the ‘mental representations’ of the organisms. Indeed, we would argue that he could not do otherwise, due to the radically different nature of the two 'representations', the scientific and the organismic one. The scientific representation is distinct and autonomous from what the representation is meant to represent, in the sense that it can be changed at will by the scientists until a temporarily satisfactory fit is found. The organism's ‘mental representation’ instead follows the niche/group coevolution and is not intentionally and autonomously produced. Properly, it cannot be distinguished from the niche itself. The second and more fundamental confusion arises from what we believe is Kuhn’s misunderstanding of the distinction between niche and external environment in evolutionary biology. This misunderstanding spreads to the analogous distinction he specifies in scientific change and makes the whole analogy faulty at its root. 22

This is an example from Ruse (1986a: 162). Consider an ancestor who sees two tigers enter a cave she is using for sleeping and only one coming out. If she innately prefers ‘2-1=1’ to ‘2-1=0’ she would be at a selective advantage over other less discriminating individuals. 23 Of course we can change the ‘external world’ in the sense that we can extinguish a fire or build a barrier to prevent falls, but not in the sense that we can change the natural regularities we believe are behind the dynamics (and the effects on humans) of fire and of falling objects. 24 “Most proposals for change are rejected on the evidence” (RSS: 10).

38

Chapter Two

Kuhn, by reading Lewontin, extrapolates the possibility of drawing “a line between the creatures within the niche, on the one hand, and their ‘external’ environment, on the other” (Idem). If we read this in the context of Kuhn’s argument, in which he is arguing against species adaptation to a fixed environment,25 and arguing that “[w]hat actually evolves […] is creatures and niches together”, it is clear that he is contrasting coevolving couples species/niche to an immutable ‘external environment’. Kuhn interprets Lewontin's remarks as if it was not meaningful to define a niche without the community that inhabits it, in an interactive fashion of coevolution. Properly though, an ecological niche is not ‘the world of the group which inhabits it’, which instead is the locality overlapping the niche. The actual coevolution does not involve species and niche, because species do not interact with the niche, they define their own niche (and the evolution of this) while evolving. The actual coevolution, meaning coupled evolution of interacting realities, is between niche (indirectly, through its population) and locality, not between species and niche.26 Our argument will be made clearer, we hope, by the following diagram.

Figure 4 – Kuhn and the evolutionary analogy

In the diagram we represent what we believe is the correct representation of the relations between a species, its ecological niche, the locality the species happens to live in and the background of the natural reality. In Kuhn’s interpretation of this system, what he calls ‘niche’ (and we mark as

25

“Is it the creatures who adapt to the world or does the world adapt to the creatures?” (RSS: 10-11). 26 Sharrock and Read (2002: 192) analyse the same passage for their own purposes but they do not detect Kuhn’s mistake. Actually, they seem to replicate it, by using ‘niche’ and ‘environment’ interchangeably.

Evolutionary Analogies

39

‘K-niche’) is actually the ecological locality.27 It is from this incorrect interpretation that he derives his evolutionary analogy. The net effect of Kuhn’s erroneous interpretation is that, on the scientific side of the analogy, the scientific niche becomes the only world ‘visible’ to the group, the only reality the group can interact with. Each scientific group determines its own niche (whose representation constitutes the group’s world) which coevolves with the group. The “external world” still puts constraints on the results of our tests, and reveals the degree of adaptation of the representation of the groups’ niche – their world – to the niche. However, those tests do not carry any information about the correspondence of the group’s world to the external world and no progress towards a better correspondence can be conceived. This allows Kuhn to save his whole view of science and scientific change.

2.2 Campbell In 1974 Donald T. Campbell coined the term ‘evolutionary epistemology’ with his contribution to Schilpp’s volume The Philosophy of Karl R. Popper. In his work it is possible to see, from a different perspective, most of the EET arguments and their weaknesses, from their excessive generality to the emphasis on the social pressure over scientific choices.28 Campbell, however, also pointed out further issues, although in a minor key, which we regard as equally important. Namely, we refer to the role of the ‘real world’ in the selection of scientific concepts and theories and the need to move away from the biological metaphor in the generalisation of selective processes.

Knowledge gain and the blind-variation-and-selective-retention mechanism Originally, Campbell’s main objective was to explain, via a general theory of selection processes, the mechanisms of knowledge gain. Like Popper, he used to believe that between “a modern experimental physicist and some virus-type ancestor there has been a tremendous gain in knowledge about the environment” (Campbell 1987: 91-2). The basis is a ‘blind-variation-and-selective-retention’ (BVSR) mechanism for gaining 27

Which, in the biological world, evolves while interacting with the niche and causes the selective pressure on the species to keep the niche and the locality overlapping. 28 Additional criticisms to Campbell can be found in Richards (1977).

40

Chapter Two

new knowledge. New knowledge is the result of a trial-and-error process, an explorative activity, in which the successful explorations were in origin as blind as those which failed. The difference between the successful and unsuccessful was due to the nature of the environment encountered, representing discovered wisdom about that environment (Ibidem, p. 92). Campbell claims that this inductive process, responsible for all knowledge gains, is analogous to natural selection in biological evolution, although conceptual variation, instead of ‘random’, is qualified as ‘blind’ because it is ‘independent of the environmental conditions’ and ‘uncorrelated with the solution’ (Ibidem, p. 93). Variation, in Campbell’s system, is not ‘random’, because all variations do not need to be equally probable nor statistically independent of one another. Campbell specifies that the adjective ‘blind’ also derives from the assumption that an incorrect trial could never cause the generation of a ‘correcting’ variation, of an adjustment of direction. From a more general evolutionary point of view, Campbell elaborates on the amoeba-to-Einstein continuity (Popper 1978: 261), arguing that in more complex organisms we observe a localisation of the trial-and-error mechanisms, from whole organisms or gene pools to the single organism, in which: Such [trial-and-error] processes are numerous, each being not only a device for obtaining knowledge, but also representing general wisdom about environmental contingencies already achieved through organic evolution, making possible more efficient achievement of detailed local knowledge. (Campbell 1987: 93)

Examples of such processes are the exploratory locomotion of protozoa, echo-location and vision. In the mechanism of vision as an aid for movement, for example, each single receptor of an eye blindly receives information about potential directions of locomotion and the retina collectively ‘explores’ the possibilities of locomotion in a wider area. “These possibilities have been made ‘blindly’ available without prescience or insight”, in Campbell’s opinion, because the single receptors are ‘blindly’ receiving information from the direction they are pointing to. At a higher level of knowledge processes, Campbell claims that even creative thought produces knowledge by the same mechanism, because also in this case “the emitting of thought trials one by one is blind, lacking prescience or foresight”. Also in this case, knowledge gain does not look overtly blind because of the available general wisdom about the environment in which the trials are performed. This wisdom, however, has been previously acquired by a BVSR mechanism at a different level and

Evolutionary Analogies

41

real gain of new knowledge will be only achieved by a new BVSR exploration on that basis. Although knowledge is always acquired by blind variation and selective retention, it is possible that prior knowledge “may restrict, though not abrogate, thought trials”. This prior knowledge, however, “must itself be a product of previous blind variation and selective retention” (Ibidem, p. 498). As a result, the whole process leads to intelligent behaviour and, as a whole, it provides ‘foresight’, although the source of the final knowledge is always blind.29 A complete hierarchy of levels at which the BVSR is responsible for knowledge gain can be found in Schilpp (1974: 422-434), where cultural cumulation and its special case, science, lie at the top. Our aim here, however, is not an analysis of the details of how this hierarchy can be defended or attacked. The main point of interest for our argument, at this stage, is Campbell’s insistence on the blindness of the knowledge gain process. This blindness is a unifying feature across different levels of the knowledge acquisition hierarchy and is, after all, the heritage of the biological character of the original idea. If we wish to subsume both biological and intellectual evolution under the same general selective process, we have to account for the difference in character of variation between them. On the one hand we can admit that variation is intentional and directed in intellectual evolution, while random or blind in the biological one, and at the same time downplay this difference as nonsubstantial.30 Campbell, on the other hand, chooses to defend the blindness of the process in all cases, in order to construct as general a process as possible. We will now focus on the questionable nature of this defence and, then, on the excessive generality of Campbell’s BVSR.

How blind is knowledge gain? Space of solutions It is possible to object that, if we recognise the enormity of the potential solutions domain for any given problem, it would be humanly impossible to find the right answer by a blind search as described in BVSR. This objection is judged by Campbell (1974: 105) to be equivalent to the similar one against natural selection in biological evolution and his position is thus defended analogously: not all problems are solved; 29

Later, Campbell (1997) will reject the broad meaning he gave to the term ‘knowledge’ and exclude from it the adaptive organic form. The bulk of his view is unaffected though. 30 As David Hull does, see next chapter.

42

Chapter Two

solutions are found very slowly, and are accompanied by a huge amount of discarded thought trials and achievements which, although published, are never used; not all possible variations are explored, because selection occurs at every step and “variations on nonadaptive variations of the previous generation are never tested – even though many wonderful combinations may be missed therefore” (Ibidem, p. 106). Here a number of issues can be identified, mainly deriving from a basic flaw in Campbell’s analogy: in biological evolution, the term ‘all possible variations’ does not make much sense. In principle, if there is a goal and a system with a finite number of possible states is trying to achieve it,31 then it might make sense. This, however, is not the biological case as we find it reasonable to think of the space of biological variation as a continuum, in other words as uncountably infinite. This is an important point but its discussion will be delayed to the next, more synthetic chapter. Here, a different point is more relevant to our present argument. Campbell seems to feel the need to unify all levels of conceptual production under the same mechanism, from the individual creation of thought trials by the mind of a scientist to the publication of results in journals. There is no ground, however, for pursuing such unification, apart from the trivial observation that all levels of conceptual production embed some sort of filtering mechanism. The tremendous number of nonproductive thought trials on the part of the total intellectual community must not be underestimated. Think of what a small proportion of thought becomes conscious, and of conscious thought what a small proportion gets uttered, what a still smaller fragment gets published, and what a small proportion of what is published is used by the next intellectual generation. There is a tremendous wastefulness, slowness, and rarity of achievement. (Campbell 1974: 105)

Campbell seeks to abstract different processes under the same general evolutionary explanation, but this does not mean we need to find analogies or correspondences between all features of the processes. An abstraction is necessarily a simplification of those processes, in which only a core of features necessary for the characterisation of both is required. Blindness is not in this core and is not needed at all levels of concept change. For example, individual creativity is widely open to different explanations and it could be conceded that the emission of blind trials plays a partial role in the knowledge discovery process. At a higher level, however, where ‘trials’ are interpreted as experiments actually performed and research 31

Or within some constraints that make those states at least countable.

Evolutionary Analogies

43

really pursued, the blindness of the trial creation is difficult to defend. Let us focus on the level of journal publications. What is usually published is the set of results and conclusions arising from the testing of a certain hypothesis or from the study of a set of phenomena. The choice of the hypothesis or phenomena is hardly the result of blind trials: individual research is usually embedded in wider research projects and programmes, for example programmes to find a cure for cervical cancer or to produce usable energy from confined nuclear fusion. The individual bits of published research, thus, are not blind attempts to provide the solution to the main problem. They originate from and find their justification in hints and suggestions provided by previous works, usually acknowledged and discussed in their introductions. In the case of nuclear fusion, for instance, the study of laser-driven shock-wave propagations in certain materials is embedded in the general research on nuclear fusion ignition and confinement. At the end of the Introduction to a paper presenting results related to such research it is stated: For the sake of illustration, we concentrated on targets of CH2 and aluminum as these are materials used in previous experiments. Our selection of materials was also influenced by our previous experience with such composite targets and the availability of a complete set of atomic data for them. The 1.06-μm wavelength laser in an intensity range of 3×10133×1014 W/cm2 and a modified Gaussian pulse shape with widths of 300, 600, and 1000 psec were chosen, to make the calculations close to the experimental range recently proposed in the literature. (Salzmann et al. 1983: 1739)

Clearly, the choice of the experimental parameters is shown to have been dictated by the status of the research. If we take a step back, we also find that the choice of the particular phenomenon to investigate was not blind. We read at beginning of the same Introduction: It has been known for many years that laser-plasma interaction can generate a strong shock wave propagating into the cold material. The pressure of these shocks may attain values up to a few terapascals, far beyond the pressures achievable in laboratory experiments and comparable to the pressures measured in nuclear-explosion-driven shock waves. (Idem)

Of course it is possible to follow a back-chain of choices caused by previous ones, until we reach the mind of the authors of this paper or, alternatively, of those of any of the authors that can be tracked back recursively from the works they reference. Eventually, then, if we switch from the group to the individual and if we accept Campbell’s account of

44

Chapter Two

individual blind thought trials, then we could find a blindness at the root of conceptual change. However, if we restrict ourselves to conceptual change as a collective phenomenon, Campbell’s blindness is not to be found. Genius For the most part, Campbell is able to defend his standpoint on the blindness of thought trials by possibility arguments. In other words, he shows that his view is possibly correct and does not contradict the evidence, rather than being the most plausible among the possible ones. For example, one objection to his apparatus comes from the ‘mystique of the creative genius and the creative act’: we are reluctant to attribute to the proponent of a great and innovative idea just extraordinary luck, as opposed to talent and to a history of brilliant antecedents. To Campbell, however, “as we do not impute special ‘foresight’ to a successful mutant allele over an unsuccessful one, so in many cases of discovery, we should not expect marvellous consequents to have had equally marvellous antecedents” (Campbell 1975: 103). In Campbell’s model of trial-anderror, thinkers can only differ in the accuracy and detail of their representation of the external world and in the number and range of trials performed. These latter factors, in particular, seem central to his account. Because the generation of trials is essentially blind, the more numerous and varied they are, the greater the chance of success. Fanaticism, extreme dedication, age, exposure to different cultures, work environment, they all affect the number of hypotheses a thinker can generate in a certain time, but also their variety, which Campbell links to the combinatorial ability of assembling simpler elements, “as in going through the alphabet in finding rhymes or puzzle words” (Ibidem, p. 104). In the subsequent stage of the selection of hypotheses, different abilities gain importance such as for instance the ability to retain and transmit solutions. As a consequence, Campbell argues that while all those individual differences explain why new ideas are more likely to come from some thinkers, they also “make it likely that many important contributions will come from the relatively untalented, undiligent, and uneducated”, given a large population. Although Campbell will later claim to have moved away from the biological metaphor, here he is trying to accommodate the disanalogy between blind biological evolution and (what we regard as) directed conceptual change, by negating the directionality of the latter. As a result, the usual explanation for the amazing adaptational success in the biological realm, (i.e. the enormous number of variations), is translated into the generation of a great number of blind trials to achieve success in conceptual change. However, it is often noted that innovative ideas and discoveries in a given field are more likely to arise from scientists coming

Evolutionary Analogies

45

from a background different from that field. We cannot see how this would be due to an increased rate of hypothesis production. It is more likely that the scientists moving to a different area of research bring with them the capacity of exploring different options and connections or the possibility of applying different heuristics or search strategies. These abilities would be characteristic of the original disciplinary fields of those scientists and not of the new field they are now involved in, and their introduction in the new area might explain the achievements. Therefore, we see the role of chance/blindness here on a different level: it is possible that in the conceiving of new ideas the element of chance lies in the availability of the right elements to fit together, and not in finding the answer by an obsessive blind search of a space of solutions. Examples come from the anecdotal accounts of discoveries which struck the author in the most trivial situations. One is the invention in 1983 of Polymerase Chain Reaction, a method of amplifying DNA regions created by Kary Mullis. He was in his car going home after work, trying to continue his line of thought to devise a new method for the analysis of DNA mutations, which was his current problem at work (Mullis 1998). Instead, he came up with a very simple method for exponential amplification of DNA molecules, which is now ubiquitous in genetic techniques and for which he gained a Nobel Prize in Chemistry. Although, in this case, the discoverer was not even searching for the solution he eventually found, his mental activity is perfectly compatible with the mental ‘composition’ of available genetic tools and techniques for finding a suitable one. This composition is very likely to be guided by previous experience, the knowledge of the features of existing tools and techniques and so on. To use a visual analogy, we can try and build, say, a model airplane from a components box. Imagine we are in a dark room, the pieces are scattered all over the place and we have never seen the aircraft we are going to build nor the instructions provided in the box. If we just blindly reach for different components trying to assemble them, we are approximately performing Campbell’s BVSR. If we switch the light on, however, our activity will better resemble the majority of scientific research activities. We still keep a degree of blindness, because we do not know what exactly we are building. We might even ignore the function of most of the components, which will be clear only once they are fitted in their position. This blindness, however, is radically different from what is implied in the continuous production of all possible combinations until the right one is found. Finally, even if our argument is not conclusive at all, it shows at least that Campbell’s explanation is just one of many other possibilities.

46

Chapter Two

Human versus machine In the previous two subsections we have argued that the sort of blindness Campbell attributes to conceptual change, if any, is only likely to be found in the psychological layer of the scientific enterprise, i.e. in the activity of the individual scientists’ minds searching for solutions to a given problem. We have also argued that Campbell does not show that alternative explanations are less plausible than his. Here we will provide a short example indicating that Campbell’s explanation is actually less plausible than alternative ones. We believe a strong counterexample to Campbell’s hypothesis about the blind variation mechanism operating in our mind searching for solutions comes from chess playing. After decades of attempts it was not until 1996 that a purpose-built computer, Deep Blue, was able to defeat the world champion, Garry Kasparov, by using a ‘brute force’ strategy (M. Campbell et al. 2002). This strategy, by simply calculating all possible future positions within a certain number of moves32 and evaluating their strength, is equivalent to a blind-variation of possible future positions and selective-retention of the move leading to the most favourable one. We can then assume, if Campbell is right, that both the machine and the human player are using the same mechanism to decide the next move. If we think of the chess game as a competition between two players acquiring knowledge about the current position on the board, in terms of possible outcomes from that position, and we agree that the search speed of alternative solutions is much lower in human brains than even in simple computers, then this example makes Campbell’s hypothesis implausible. Otherwise, a machine ‘searching’ for moves the same way as the human does, but faster, would easily win any match. Some could object, though, that in this case the human player is not in the same situation as the scientist: the scientist is evaluating possible solutions to a problem against a stable external reality, while a chess player is confronted by an ever changing configuration. However, we find the two conditions the same if we think of each move individually and there is no reason why, in this specific activity, the solution-finding process should differ from all other comparable processes. It is remarkable that Campbell himself found automated chess playing to be an example of how the processes likely to happen in a player’s mind can be emulated (Campbell 1987: 107). This happened in the late ’50s, long before it became clear that even extremely powerful machines implementing 32

The system was able to perform a full-width search of the game tree to the average depth of 12.2 in three minutes (M. Campbell et al. 2002: 61).

Evolutionary Analogies

47

Campbell’s strategy to problem solving struggle by comparison with exceptional human players.33 Simultaneous discoveries A further difficulty can be identified in Campbell’s treatment of simultaneous discoveries in science. A major empirical achievement of the sociology of science is the evidence of the ubiquity of simultaneous invention. If many scientists are trying variations on the same corpus of current scientific knowledge, and if their trials are being edited by the same stable external reality, then the selected variants are apt to be similar, the same discovery encountered independently by numerous workers. This process is no more mysterious than that all of a set of blind rats, each starting with quite different patterns of initial responses, learn the same maze pattern, under the maze's common editorship of the varied response repertoires. Their learning is actually their independent invention or discovery of the same response pattern. (Campbell 1974: 435)

The difficulty arises from Campbell’s confusion between simultaneity and agreement. What needs to be explained about simultaneous discoveries of the same solution to a given problem is the fact that blind variations in two or more distinct scientists’ minds happened to produce the same result at around the same time. In contrast, Campbell offers a plausibility explanation which only applies to the identity in content of the discoveries, not to their simultaneity. The analogy with the rats only shows that the same constraints force the animals to find the only solution available, not that it will be found in the same time across different animals. Given the properties that Campbell attributes to the blind variations and the assumption that an external reality participates in the selection process, it is plausible that eventually but at different stages different scientists would produce the same discovery. We doubt, however, that genuinely blind processes happening in scientists’ minds in the way Campbell suggests would ever be capable of producing even just a pair of simultaneous independent findings. And Campbell’s own requirements about the vastness of the space of solutions for any given problem does not seem to strengthen his position.

33

The designers of Deep Blue seem to be aware of this: “The search should be highly non-uniform. It is well-known that strong human players are able to calculate well beyond the depth reachable by a uniform searcher of any conceivable speed” (M. Campbell et al. 2002: 60, emphasis in the original).

48

Chapter Two

Campbell’s intention Until now, we have pointed out the main shortcomings in Campbell’s BVSR and the need to confine the role of ‘blindness’ in it if we wanted to use the resulting process as a general explanatory tool for knowledge gain at all levels. The problem is, however, that Campbell’s account of a general process of selection is already far too general, at least in our view. In our view the core of his standpoint, with its sophisticated account of a blind trial-and-error process, does not emerge as very informative. This is revealed by its applicability to almost any phenomenon where change is involved, with blindness to be found wherever there is no mind to give a direction. Even the evolution of gravitational systems like Saturn’s rings or the asteroid belt can be explained in the view of Campbell’s analysis:34 some particles with initially random positions and trajectories (blind variation) are eventually grouped in a distinguishable structure, while others are lost in casual orbits (selective retention). Occasionally, it is pointed out that this process is so fundamental that it can be found from the well known example of the growth of crystals (Campbell 1974: 420) to, allegedly, the change of scientific knowledge. We regard this as a weakness though, not as a strength of the way the process is characterised: when an explanation becomes so general it also becomes uninformative regarding the peculiarities of the individual phenomena it is supposed to explain. One of the first refinements that we claim is needed in this theory, for instance, is the differentiation of the ‘selective retention’ into ‘selection’ and ‘replication’. We regard these two processes as distinct, and different combinations of them produce (are found in) different phenomena. On a more positive note, we agree with Campbell’s intention to move away from the biological metaphor: Most of those who have elaborated on [EET] have employed too close an analogy between science and biological evolution, carrying over many details from the latter that are inappropriate […]. This essay approaches an epistemology for science from the perspective of a much more general ‘selection theory’ in which biological evolution is just one nested cluster of exemplars. This biological cluster is, of course, to be mined for useful insights and analogies, but is not to be taken as a compulsory model to be followed in every detail. (Campbell 1997)

34

This example will be viewed in more detail and from a different perspective in the next chapter.

Evolutionary Analogies

49

His was only an intention though. If we keep forcing all sorts of analogies between biological evolution and all other processes explaining the change over time of different systems, we will be obliged to find features (e.g. blindness) where they are not to be found or to dilute our explanation to a trivial one. The way forward In our view, the correct generalisation of two distinct processes – scientific and biological change – which are thought to have some features in common goes through the decomposition of the processes into simpler elements. These elements will display features that can be derived from more general elements, and some of these more general elements will display features that are in common with the two processes at stake. The process resulting from this group of shared, core elements and features characterises the correct generalisation of the two processes. Although Campbell seems to point vaguely in this direction in the passage just quoted, we do not believe he pursued it. He devised a very general process which, as we pointed out, is universally applicable, based on blind variation, selection and retention and then went on to characterise in great detail the properties of the first two components, though mainly focusing on blind variation. The evolutionary analogy, then, was expounded in a series of mappings of variation or selection properties between biological and scientific change: adaptation, fitness, group selection and so on. He did not, however, feel the need to find an analogue to biological replication in conceptual change, as David Hull did (see next chapter), because he did not identify it as a feature of his ‘retention’. In this respect, Hull was much more thorough in his generalisation of selection processes, by identifying abstractions for their components and trying to analyse how they would be instantiated in specific cases, such as in the evolution of living beings or concepts. However, in our opinion, he was guilty of staying too close to the biological metaphor, so his generalisation is also inadequate. On the other hand, Campbell finds Hull’s analysis of scientific change inadequate on a different basis. He claims that Hull’s characterisation of scientific enterprise is not based on the biological analogy (Campbell 1997). Maynard Smith (1988) finds Hull too generic, because he fails to correctly recognise the specificity of scientific change as opposed to the evolution of any other social system. Campbell (1997) takes the same position and he points out that science is the product of a special kind of self-perpetuating belief community, one which is not locating the truth “in a long-past revelation” but in the future, promoting innovation and “decrying tradition as a burden and source of error”. This, however, would only happen in the idealisation of science in the mind of

50

Chapter Two

young scientists or in popular accounts of its aims and rules. Campbell insists that this myth about scientists looking for the truth, away from the ‘contaminations’ of their values by those of the rest of society, such as monetary interests or prestige, should be kept alive, though. From Campbell’s perspective, “the ideology and norms of science are not clearly distinguished from ‘scientific method’”, so the ideology should be kept alive as a means to induce scientists to adhere to scientific method as much as possible. Although we agree and we will show that Hull does not do enough to distinguish scientific enterprise from other social enterprises, we think it should be allowed that his standpoint is not too far from Campbell’s own view. Hull (1988c: 394 and 437) attributes the consistent application of the scientific method in scientific enterprise to the coincidence of its global (group) rules and goals with local (individual) ones. He explains this, however, in terms of rewards and punishments: outright data fabrication, for example, rarely happens because, when discovered, it is punished most severely, much more than plagiarism. This is because data fabrication is much more damaging to the scientific community, where each individual bases their own research on the basis of previous research by others, on the assumption that other scientists have been honest in their activity. Campbell follows the same line of reasoning and the only layer he adds to Hull’s view is the participation of nature in the selection process: “A selectionist model for a scientific belief ‘justifies’ such a belief to the extent that it is plausible that ‘the way the world is’ has participated as one of the systematic selectors of that belief from among the historically developed rival beliefs.” (Campbell 1997). This point, however, is far from being central in his work. We would go further than that. We will argue, in Chapter 4, that the referential content of scientific concepts prevents the selective process acting on them from being equated to natural selection. Research course and results Evolutionary epistemologists, usually by playing down or ignoring the role of both intentionality and reality in scientific change, have failed to recognise the relevance of the distinction between two fuzzily separated levels at which selective processes operate in science: we will call these levels the level of the course and the level of the content.35 What we will 35

Here we use the term ‘course’ instead of ‘direction’ as we mean something different from what will be meant in Chapter 4. There, the directionality of scientific enterprise will refer to its content, that is to the preferred directions which conceptual lineages follow due to their relation with a fixed external reality.

Evolutionary Analogies

51

call purely externalist36 selection operates at the level of research course, i.e. it can strongly influence what scientific research should look at. This selection is intentionally exercised by the pressures of laboratory directors, by the requests from wider society or by the policies of funding organisations. These players, however, do not dictate what is to be regarded as a theory rejection or not, they only influence the path followed by scientific research. These ‘social factors’ label specific research areas or routes as ‘interesting’ or ‘uninteresting’ in a wide sense and ‘decide’ whether they should be engaged. They cannot label research outcomes as ‘valid’ or ‘invalid’ once the research is pursued, though. This latter sort of decision occurs at the level we have labelled as ‘content’. At this second level, the ‘real world’ participates in the content selection, that is the retention or refutation of hypothesis and theories, through the experimental interface. This second stage, however, can be further divided into what we call an inner externalist layer and an internalist one. In other words, the outcomes of the content selection are, in their turn, the result of an inner layer of social phenomena, accounting for the human component of scientific experiments, and the outcomes of the interaction between the experimental interface and reality. These distinctions can be schematically represented in the following Table 1. Selective activity: What is affected:

Outer externalist Course

Inner Internalist externalist Content

Table 1 – Science course and content selection Kuhnian theory-ladenness37 of scientific observations and the degree of discretion in scientists’ judgements belong to the inner externalist layer; the actual results of the observations, such as the numbers we read on our instruments, do not. We agree with Campbell that once the experiment or observation has been set up “the results of a single galvanometer reading are out of the control of one’s own hopes and wishes” (Campbell 1985: 40). Thus, although we cannot regard content selection as guided by ‘facts’ or ‘reality’ alone, because of the inevitable dependence of any observation on accepted theories and judgments, outer social pressure can 36

We will use here the terms ‘externalism’ and ‘internalism’ in a sense close to Hull’s wide sense (Hull 1988c: 1): internalism explains scientist’s choice in terms of “weight of evidence and cogency of arguments”; externalism links those choices to social forces or personal motives. 37 If, of course, we accept this notion which we acknowledge is controversial.

52

Chapter Two

do nothing against a series of experiments showing that a certain hypothesis should be rejected or not.38 For instance, research on the dynamic of confined nuclear fusion has recently been pushed strongly after years of slow progress, because of a network of social reasons: concerns about global warming, uncertainty regarding oil reserves and profit perspectives. The individual results of individual pieces of research in that field, though, will not be influenced by what is pushing it and the conclusion will be generated by the interaction between the currently held view of the world and the experimental results. It could still be argued, however, that the evaluation context in which scientific hypotheses are tested partially depends on the status of the relevant research in general. In other words, even if we confine ourselves to the content selection, the decision between opposite conclusions could be dependent on whether or not specific results in related fields are available, which may be determined by direction selection. So, indirectly, content selection is still partially dependent, even in its internal mechanism, on directive selection. We would stress, however, that this dependence is only historical and not absolute. Like the memory game of matching cards, the order in which the cards are turned and matched to one another can greatly change the course of the game; in the long term, however, only the rate of its progression can be affected, not its outcome. Even in the idealisations of scientific enterprise, truth is regarded as only an asymptotic achievement, and society is among the delaying factors. Additionally, the relation between (outer/inner) externalism and internalism is asymmetric: while wishes, hopes and social pressure cannot change instrument readings, observational and experimental outcomes can still influence the predisposition of scientists and wider society. If it is true that no logically crucial experiment exists for any hypothesis, it is also true that a series of negative experiments are usually enough to induce scientists to review their understanding of that chunk of reality they believe is involved in those experiments.39

2.3 Toulmin Stephen Toulmin is regularly quoted and referred to by scholars dealing with evolutionary epistemology (EE), and he is often regarded as 38

This is not excluding the possibility, say for a rejected hypothesis, of being retained in an extra-scientific environment. 39 Sometimes this happens very quickly, as in the case of the rejection of parity conservation principle in weak interactions, which was abandoned within a few months in 1957 (Campbell 1997: 42).

Evolutionary Analogies

53

one of the precursors of the present approaches to it. A brief account of his position will be useful on three counts: to discuss some general objections that can be raised to EE; to introduce many distinctions and arguments that Hull’s work will develop and refine; to provide some basis for our own analysis.

Explanation or metaphor? In Foresight and Understanding, Toulmin suggests a ‘parallelism’ between the evolution of scientific ideas and the evolution of species (Toulmin 1961: 110). Two suggested analogies are between the merit of ideas and the survival-value of characters and between scientific disciplines and biological environments, but the argument there is very short and not very detailed. Toulmin himself stresses that his argument is only about a ‘parallel’ (Ibidem, p. 113) between the two forms of evolution, which “suggests new questions and possibilities.” (Ibidem, p. 111) A few years later, however, he points out that the he has “not been employing a mere façon de parler, or analogy or metaphor.” The idea about the evolutionary pattern of change of scientific thought is not “merely suggestive, but explanatory.” (Toulmin 1967: 470) Biological evolution is thought of as a model for scientific change, in the same way as the notions of “ideal gas” or “rigid body” (Ibidem, p. 467) are in physics. As such, because of the idealisation implied by all models, “[w]e are not obliged to demonstrate that all scientific changes whatever conform to this ideal, any more than we need demonstrate that all material bodies are ‘perfectly rigid,’ or all actual gases ‘ideal’.” (Ibidem, p 467, emphasis in the original). In Human Understanding he further qualifies his standpoint: [I]t will not be necessary to assume […] that intellectual evolution has something ‘biological’ about it, or even that the process of conceptual change in the sciences displays any substantial resemblance to the process of organic change. We shall be committed only to a more modest hypothesis, namely, that Darwin’s populational theory of ‘variation and natural selection’ is one illustration of a more general form of historical explanation; and that this same pattern is applicable also, on appropriate conditions, to historical entities and populations of other kinds. (Toulmin 1972: 135)

54

Chapter Two

This differs from the more recent and articulated positions of Darden, Cain or Hull,40 who seek a generalisation of selection processes in order to show that the same process is operating in biological and in scientific change. On the other hand, in a manner similar to the more refined approach undertaken by Hull, Toulmin points out the need to distinguish between two interrelated phenomena: the development of conceptual systems through time and the activities of the people who produce this permanent process of change. It is thus possible to identify two realms in their analyses, one conceptual (rational) and one sociological (professional). These two sides of “the same historical changes, as seen from different directions” (Toulmin 1972: 142) cannot be successfully analysed if taken independently and their mutual interaction not recognised and taken into account.41 Problems arise, however, when this interdependence is explained in terms of evolutionary processes. For instance, in The Evolutionary Development of Natural Science (1967: 466) Toulmin argues that variation is highly sensitive to ‘external’ factors, such as the social context in science or cosmic rays in genetic material. In contrast, selection is highly confined to ‘local’ interaction, such as plausibility arguments supported by scientists or the local habitat of living species. Toulmin’s synthesis is realised in the only question he believes should be asked: “Do the new forms meet the detailed demands of the situation significantly better than their predecessors?” (Ibidem: 466). This question links selection and variation, it gives meaning to the variation stage in the perspective of the successive selection. The way it is asked, however, is biased by the scientific side of the analogy, because it implies a global tendency to progress. Therefore, Toulmin’s perspective reveals a major problem we have already encountered and which affects most accounts of the evolutionary analogy: it implies global progress, which is relatively uncontroversial in science while almost unanimously rejected in biological evolution. In Toulmin’s case, progress seems to be simply ‘imposed’ on the evolution of living beings because he is more focused on the scientific side of the analogy. In other cases, like Hull’s, where the biological side is overemphasised it is the progression of the scientific enterprise which is questioned.

40

Which will be discussed in the next Chapter. Hull will argue that the personal goals of the scientific profession are modelled on, or coincide with, the aims of science.

41

Evolutionary Analogies

55

Darwinian rationality Toulmin claims that, in both the evolution of life and science, change is the result of the “selective perpetuation of variants” (Toulmin 1961: 110). He emphasises the importance of identifying both “the genealogical sequence of intellectual problems” (Toulmin 1967: 466) and the entities involved in the evolutionary process. To investigate the reasons why scientific ideas have success in the specific context in which they develop we must, according to Toulmin, concentrate on the evolutionary process of ideas, methods and aims of science. Each scientist’s generation brings new ‘variants’ to the tradition. The philosopher must, essentially, analyse the criteria used to evaluate and select good variations. This procedure can also be found in Darwinian biology, where variation is the result of the reproduction of selected variants. The question ‘what lets scientific ideas survive?’ leads to a richer perspective when answered in evolutionary terms: analogous to variations in biological species, new theoretical explanations and procedures can be good or bad depending on the context (environment) that they are applied to. Survival is not due to the fulfilment of a single requirement, but on the contrary it is the result of meeting the various constraints of the environment. Although this approach entails the dissolution of the dream of absolute (formal) criteria for comparing scientific theories, the parallel with the theory of biological evolution, in Toulmin’s opinion, is very strong and valuable. A populational approach42 to intellectual history, according to Toulmin (1972: 128), would provide us with ‘live’ historical entities which, while changing all the time, would still keep the observed unity and continuity: the rationality of scientific theories would then lie in the procedures of conceptual change.43 The study of rationality in scientific disciplines is the search for a full and 42

The approach of population thinking stresses the uniqueness of everything in the organic world: only the individuals of which populations are composed have reality, averages are merely statistical abstractions. Population thinking sees every species as a varying population of interbreeding individuals. It rejects the idea that each species has a natural type, as the earlier essentialist view had assumed. ‘What is true for the human species, that no two individuals are alike, is equally true for all other species of animals and plants… All organisms and organic phenomena are composed of unique features and can be described collectively only in statistical terms. Individuals, or any kind of organic entities, form populations of which we can determine the arithmetic mean and the statistics of variation […] only the individuals of which the populations are composed have reality’ Mayr (1970: 4-5). 43 “[a]n entire science comprises an ‘historical population’ of logically independent concepts and theories, each with its own separate history, structure, and implications” (Toulmin 1972: 130)

56

Chapter Two

detailed understanding of the features that make alternative strategies and procedural innovations adaptive to particular ecologically recognisable problem-situations (Cohen 1973: 45). This forms the basis, and the possibility, of rational and objective judgment of any concept and belief: “The burden of ‘rationality’ […] consists in the fundamental obligation to continue reappraising our strategies in the light of fresh experience” (Toulmin 1972: 503). In Human Understanding, Toulmin insists that the rationality of science can be explained in terms of patterns in the historical evolution of intellectual disciplines, as opposed to logical explanations of relations between propositions (Toulmin 1972: Conclusion). Several objections have been raised to Toulmin’s arguments. One such objection is that his evolutionary model for the history of intellectual disciplines is not Darwinian (Cohen 1973). Toulmin admits that his evolutionary framework differs from the Darwinian: there are, firstly, no “specifically biological details” and, secondly, conceptual variation and intellectual selection are coupled.44 Nevertheless, he plays those differences down as non-crucial. In contrast, Cohen underlines how the existence of a connection between the selective factors and the originators of variation, that is, the fact that “conceptual variants are for the most part purposively thought up in order to solve the intellectual problems that beset a discipline” (Cohen 1973: 47), is actually enough to change the meaning of the term ‘neo-Darwinian evolution’ completely. The coupling of variation and selection and its cause, intentionality, constitute a source of major problems for evolutionary treatments of scientific change. Campbell (previous section) tries to avoid it by appealing to an alleged ‘blindness’ of the intellectual variation process, but Hull (see next Chapter) will have to recognise it as the strongest disanalogy between natural and intellectual selection.

Environment The environment is particularly important in Toulmin’s parallel between biological and scientific evolution. While a specific variation in a certain species in a certain environment can prove advantageous, he writes (Toulmin 1961: 111), that the same variation may have totally different consequences in another species or environment. Analogously, in science an argument or a theoretical pattern may vary extensively in its fruitfulness across different disciplines. In Human Understanding he will equate, in the analogy, biological environmental pressure to criticism and 44

There are preferential directions of variations, something that would lead to Lamarckian interpretations of biological evolution.

Evolutionary Analogies

57

competition, thus biological environment to intellectual environment. This is a trend that will be followed by the other proponents of evolutionary analogies: the benchmark of intellectual fitness is commonly found in the intellectual milieu, in the professional setup of the scientists involved in the given research, in their shared values. Although Toulmin’s treatment is still embryonic and open to a wide range of criticisms of his sketchy analogies and parallels, he is already forced to downplay the role of the empirical benchmark or, in general, any reference to an external reality or regularities against which (however indirectly) the intellectual constructions of science are tested.45 After all, Toulmin writes, “the principles of intellectual selection for sifting the conceptual variants available at every stage [are] ‘the daughters of time’” (Toulmin 1972: 253). He knows that a relevant part of the ‘environment’ which is responsible for the selection of scientific ideas is reality itself. But this role is severely down-played. We shall argue later that it is exactly this focus on the sociological side of science, i.e. its being a social product, which allows for some arguments in favour of an evolutionary theory of its change. This, however, would be at the cost of losing specificity: the same arguments could be easily applied to all social activities, even more so than science where the activity does not involve the production of ‘testable’ entities.. In other words, entities which need a form of ‘adequacy’ test against non-temporal entities – natural regularities – they are supposed to refer to.

Ontological adequacy A proper analysis of the role of the environment in biological evolution, for Toulmin, also brings clarity to the tendency of formal philosophical argumentation about science to lose sight of the real practice of working scientists. For example, it does not usually make much sense to discuss at an abstract level how advantageous a particular trait might be for a species. Adaptive advantage is meaningful only in the context of other traits possessed by that species and the environment it lives in. Similar arguments may be used to play down abstract speculation on the problem of explaining a finite number of empirical observations with an infinite number of hypotheses. The basis for this remark is the simple observation that through any finite set of points an infinite number of mathematical curves can be constructed. 45

He claims that the difference between ‘regulative’ ideals (such as in law) and ‘representational’ ideals (such as in science) is only a matter of degree (Toulmin 1972: 260).

58

Chapter Two […] In fact, the scientist’s problem is very different: in an intellectual situation which presents a variety of demands, his task is – typically – to accommodate some new discovery to his inherited ideas, without needlessly jeopardizing the intellectual gains of his predecessors. This kind of problem has an order of complexity quite different from that of simple curve-fitting: far from his having an infinite number of possibilities to choose between, it may be a stroke of genius for him to imagine even a single one (Toulmin 1961: 112-3).46

Therefore, the scientific environment of existing theories and intellectual gains constrains the production of new theories.47 It cannot get away with the simple production of correct pre/retrodictions. This argument is particularly relevant to our aims. Toulmin’s argument provides a good example of a common shortcoming in the arguments held by the proponents of the evolutionary analogy. In many cases, as already pointed out in the discussion of Campbell, the proposed analogies between biological and scientific evolution are so abstract that they could be successfully extended to almost any phenomenon. Toulmin’s analogy is explicitly stated in the following passage: To begin with, we know from biology how a variation which confers an advantage on one species in one environment may have no merits at all for another species, or even for the same species in a different environment. So, in science, the same theoretical move can have merit in dealing with one group of problems, and yet prove an obstacle to progress in another field or situation. (Toulmin 1961: 111)

Table 2 shows the mapping that Toulmin is suggesting between the biological realm and science. Our objection is that Toulmin’s analogy is in reality a very general statement about the relativity of values. Apart from the more specific term ‘species’, mapped onto ‘field of research’, the entities in the first column of the table can be interpreted in a very abstract manner and, as a consequence, instantiated in almost anything. What Toulmin is telling us in the cited passage is no more than ‘what is good in a given situation can

46

This argument against “timeless logical ideals” in the evaluation of scientific ideas is also developed elsewhere (Toulmin 1972: p.229-230). 47 The remark that science usually only welcomes theories which fit the already well established knowledge environment is close to Kuhn’s view of ‘normal science’ periods, where he claims most scientists’ activity is devoted to fit the solutions to scientific puzzles into the framework of the accepted paradigm.

Evolutionary Analogies

Life variation advantage absence of merit species environment

59

Science theoretical move merit in dealing with problems obstacle to progress field of research situation

Table 2 – Toulmin’s evolutionary analogy mapping be bad in another’. This is a highly vague observation that can be applied to all situations where evaluating rules are context-dependent. It does not involve all the characteristic properties of evolutionary processes. Instead, it only involves a very limited subset of general properties, which causes the analogy to be very abstract. This vagueness in the evolutionary analogy can be avoided by generalising all terms at stake in the analogy and following with the analysis of their instantiation in the actual cases of biological evolution and scientific change. This is the current trend in this research area and it will be followed in our analysis in the third chapter. Furthermore, if we limit ourselves to his oversimplified mapping, Toulmin’s analogy actually turns into a disanalogy. Toulmin’s idea of the conceptual fitness of some new discovery, we have just read, also implies that the novel intellectual situation should not “needlessly jeopardiz[e] the intellectual gains of [scientists’] predecessors”. On the biological side of the analogy, however, this would map onto an absurd concept of fitness in biological evolution which confers advantage to species that do not endanger other species. The reason for this is the interconnections between different disciplines, which have in common the goal of explaining reality. The conceptual lineages in one discipline cannot go astray, regardless of the results of the rest of the scientific community, while one species may well happen to fiercely compete with other species in the same habitat, and eliminate them if needed. In Foresight and Understanding Toulmin emphasises the distinction between forecasting-power and understanding. In Chapter 2 of the book he mentions the example of the Babylonians and the Greeks, who had quite opposite approaches to celestial bodies: pragmatic and uninterested in ‘explanations’ in the former case, theoretical and unable to make any prediction in the latter case. “Nowadays, of course, we expect a scientist to combine merits of both kinds” (Toulmin 1961: 30). We expect science to go on by both increasing its capability to provide new and more accurate pre/retro-dictions, in addition to those already achieved, and consistently expanding the explanations which support those pre/retro-dictions.

60

Chapter Two

Toulmin claims that it is this latter explanatory activity which is central in scientific activity and qualifies it as ‘scientific’: classification, diagnosis, prediction can all be called ‘scientific’ only by their connection to the underlying explanation (Ibidem, p. 38). All of this has several interesting consequences for our arguments later. Firstly, Toulmin’s distinction between experimental and ‘environmental’ fitness of theories, i.e. their need to be good both at pre/retrodiction and at fitting into the existing ‘view of the world’ shared among scientists, will be further discussed and refined in Chapter 3. In that chapter, the distinction between empirical and ontological adequacy and their concurrence will be crucial for defending a form of convergent realism. Secondly, Toulmin notes that some biological sciences do not have explanations as their aim. (Ibidem, p. 21) For instance, he mentions taxonomy as an essential part of botany and zoology which does not contemplate explanations among its goals. This would classify taxonomy as an ‘accessory’ science, or at least as an atypical one. This remark will be relevant below, when we analyse Hull’s account of the evolutionary character of scientific change that makes use of taxonomy as its main working example of evolving scientific discipline. The distinction at stake here is related to the differences between physics and natural history analysed by Toulmin in The Philosophy of Science. Natural historians are mainly involved in classificatory tasks. For example, they investigate whether a particular species has evolved from another or whether or not the individuals of a certain population belong to (are part of) a given species. This activity, for Toulmin, is constrained by a certain rigidity. The classification principles are not likely to be modified in the light of new findings. For example, he writes, we do not cease to apply ‘mouse’ to the individuals of a population of mice just because we happen to discover that they have different habits from all other known mice and do not breed with them (Toulmin 1965: 51). On the other hand, physicists reclassify the objects of their investigations all the time. For example, they switch the explanation of a certain phenomenon from electric to magnetic in the light of new evidence. We agree with the general distinction between the conservatism of mainly classificatory sciences and the dynamism of mainly explanatory ones. However, we disagree with part of the basis that supports this distinction in Toulmin’s view. He claims that the rigidity of natural historians is mainly due to their commitment to everyday popular classification. In other words, they would not be free to reclassify herbivorous mice under a different name because the public would not accept the new terminology. Physicists would, however, be immune to

Evolutionary Analogies

61

such concerns because their objects and explanations are far away from everyday use by the public. We do not believe the general public has this sort of influence on scientists’ activity. In particular, to follow Toulmin’s own example, the scientific taxonomy of mice has become far more complex than the everyday use of the term, which actually refers to many different species. Scientific classifications are usually distinct from popular uses of the classificatory terms. There will always be people who consider dolphins as fish and spiders as insects. Conservatism in taxonomic sciences is due to a different reason, namely to the historical nature of their objects. By labelling some individual living beings with the term ‘mouse’, we are identifying an historical entity, namely the present population of animals we perceive belonging to a homogenous group. Nothing prevents us from further refining our classification, if we find that what we thought was a single species (or higher taxon) is actually the union of two or more. The common name for the general population, however, will stay. When we discover that a swimming animal is actually a mammal we cease to call it ‘fish’. On the other hand, the name of the species so reclassified will stay the same, and here is where the rigidity manifests itself. The rigidity derives from the immutability of ‘proper name’ designation. As Hull notices (1988c: 499), we would be forced to change the proper name of a population only if we can prove genuine mistakes, for example if we can prove that the population had been previously named differently. However, physical sciences do not reserve proper names for historical entities48 and try to identify regularities in objects that, eventually, may be given a common name: “Natural historians […] look for regularities of given forms; but physicists seek the form of given regularities.” (Toulmin 1965: 53) There is an interconnection, in physics, between the terms designating entities and the theories in which they occur. This, for Toulmin, does not happen in natural history. In natural history, one can distinguish sharply between two stages in any piece of research: the initial step of identifying an animal – unnecessary, of course, if it was bred in the laboratory – and the subsequent process of studying its habits. In the physical sciences, there is no such sharp division: the things that come to light as one goes along will frequently lead one to relabel the system being studied. […] [O]nly where one can ask separately, first, “What are these?” (Answer: As), and then, “What common properties have they?” (Answer: being Bs), is “All As are Bs” the natural form in which to couch one's conclusions. One can make this 48

Exceptions are names like ‘Sun’, which designate specific members of a kind of objects, in this case stars.

62

Chapter Two separation in natural history; but in the physical sciences the two questions are interdependent, and in consequence the simple generalization is out of place. (Toulmin 1965: 52)

Again, we agree with Toulmin’s general remarks. The rigid temporal sequence of identification followed by description can normally occur, for example, at the early stages of the study of unknown species inhabiting a recently discovered island or recess. However the two levels of research, in Toulmin’s account of the distinction, are different. The first step, the identification, would be carried out by techniques which are based on the observation of certain properties of the individuals. These properties are those which characterise the individual as belonging to (being part of) a certain taxon, for example a species. This will be dictated, for example, by exterior resemblance, potentiality to interbreed or genetic properties in common with the taxon. These observations will result in the researcher recording the observed individual under the name of a known taxon or creating a new name for the one just discovered. This first stage of research, then, relies on certain rules of classification which derive from what we believe are the natural demarcations of the current taxa. If we confine ourselves to the example of species, we do not regard individual species as immutable natural kinds any more and, although wide consensus is far from being attained, we do not think of the entities involved in the criteria used to identify species as historical entities. The criteria, as human products, are historical entities but they refer to what we think are real demarcation in nature. For example, genetic criteria refer to genetic differences. In other words, they rely on the description of the common instantiation of general properties. The second stage of research, the description of the habits of the individuals, occurs however at a different level. This level is based on historical entities, the habits of the individuals and there is not much interdependence with the identification (classification) level. Toulmin’s mapping of levels, then, is not correct. In physical sciences, identification and description are interconnected but it is their conjunction that corresponds to the identification in natural history. In physical sciences, the historical level is missing and the historical description of natural sciences simply corresponds to nothing. This observation will play an important role in the critique of Hull’s general account of evolutionary processes. In particular, we will argue that Hull’s account, when applied to scientific evolution, is better suited to taxonomy and natural history in general, from which most of his examples are drawn. In Chapter 3 we will also argue that the referential role of scientific concepts poses serious difficulties for Hull’s account of those concepts as historical entities. This, we will maintain, is mainly due to the

Evolutionary Analogies

63

reference of scientific concepts to what we regard as non-historical regularities of nature. Here it is possible to claim that, by confining ourselves to those sciences mainly concerned with the classification of historical entities into historical groupings, our argument based on the referential role of concepts would be weaker. For example, it could be argued that Hull’s account of scientific change applies to Toulmin’s description level of natural history. This, however, will be clearer in Chapter 3, after our analysis, critique and further development of Hull’s view has been completed.

CHAPTER THREE A TYPE HIERARCHY FOR SELECTION PROCESSES

In contrast to the usual answers offered by philosophers who saw in organic evolution a precursor or an analogue of scientific change, David Hull made a radical departure from these in his book Science as a Process, and his introductory paper A Mechanism and Its Metaphysics: An Evolutionary Account of the Social and Conceptual Development of Science, which was the Target Article of Biology and Philosophy in 1988. In his attempt to generalise the analysis of selection mechanisms in order to incorporate at a higher level both natural and cultural selection, Hull appeared to have developed a convincing rebuttal to the objections raised by the numerous opponents of evolutionary explanations in scientific change. Indeed, he is able to offer quite specific answers to most of the problems that had plagued previous attempts of the same sort, from the problem of intentionality to the alleged Lamarckism of cultural evolution. Although we believe Hull’s approach can be very fruitful and that the problems remaining to be solved are few, we argue here that his is only a partial attempt at making a true generalisation of selection processes, and that, in order to achieve full generalisation, a richer theory of analogy is required. More specifically, we argue that the recent work by Eileen Cornell Way on type hierarchies, introduced in the first chapter, can offer the kind of generalisation lacking in Hull’s approach and contribute to his work in three ways: to achieve its goal, to solve the main problems that still affect Hull’s theory and to show that the evolutionary analogy is, after all, just a particular way of grouping phenomena together. Hull’s shift from the assessment of individual analogies between Natural Selection (NS) and Science Selection (SS) to the proposal of a more general theory of selection – of which both NS and SS would be instances – is, in our view, a shift in the right direction towards a less arbitrary evaluation of evolutionary analogies. It is, however, only a partial attempt because his underlying approach to conceptual models retains the character of the older approach he is trying to depart from. Hull’s main objective is a unified theory of selection, which supports the idea that

66

Chapter Three

science selection and natural selection obey the same laws. We argue that the type hierarchy (TH) approach to models – which we believe to be more conducive to the aim – can be adapted to show that this objective is unsustainable as it stands, and is in need of further development. In this chapter we provide an introduction to Hull’s main points and difficulties. Then, we try and construct a TH for a general abstraction of selection processes. This will provide a framework for the assessment of both strong and weak evolutionary epistemologies. Finally we introduce the main criticisms that Hull’s work has faced from philosophers and scientists, and show how they compare with our proposed method of evaluation of evolutionary analogies.

3.1 Type Hierarchies In the first chapter we discussed the concept of type-hierarchy, which has been developed by Way (1994) and introduced as the best alternative to common accounts of models and analogies in science. Common approaches are based on the listing of properties, features and general laws that we think apply to the two objects we are trying to compare and showing whether or not these items are similar or shared between them. This will put weight on the assumption that the two objects are the same or two instances of the same type. By contrast, the new procedure for the analysis of analogies is compound. It starts with building the hierarchical network of types we think are relevant in the matter, continues with the addition of the instances as terminal nodes and only then, when the network is believed to be as complete as possible, does it terminate with the reasoning about the degree of their similarity by inspecting their common supertypes.

Top-down or bottom-up? The old procedure can be seen as a bottom-up approach (Aronson et al. 1994: 88), because we start from the bottom layer with the tokens, and we compare them directly by means of an unanalysed notion of similarity, in order to build a model which would accommodate all of them. Conversely, in the new top-down approach, the construction of a generalisation for natural selection should start by identifying types of processes that are as abstract as possible, building the hierarchy for the more ‘specialised’ types and then concluding with NS as a token inheriting from some of them. It is precisely this initial generalisation that allows for the domain of interest to be clearly defined from the outset. Thus, any other process commonly

A Type Hierarchy for Selection Processes

67

associated with NS – such as Cultural Selection or Clonal Selection or Genetic Drift – will be placed in the hierarchy linked to the types that, according to our present understanding of the phenomena, are only associated to properties that also apply to the instances that inherit from them. These processes will then be evaluated in relation to their closeness to NS, if required. Although this approach may be overly complex for evaluating analogies between simple physical systems, we argue that it is relevant to Hull’s attempt to defend his evolutionary analogy: Hull searches for a supertype of both natural selection and scientific/cultural selection, as shown in the following Figure.

Figure 5 – Hull’s abstraction of selection

However, because his task is limited to the direct comparison of these two phenomena, in order to identify the common supertype, his procedure is, partially, still the old ‘bottom-up’ approach of the comparison view. This will be clarified in the next section.

3.2 Existing generalisations Hull Hull’s aim is to “set out a general analysis of selection processes that is equally applicable to biological, social and conceptual development” (Hull 1988c: 13). That is, scientific change and evolution in the biological world are more than just analogous, they are the same kind of process. To describe this kind of process he develops a number of general categories and mechanisms that, abstracting from the entities involved, would explain evolution in different domains. The key entities for his construction and for which he gives exact definitions in an effort to reduce confusion over the nature of selection are the following (Hull 1988c: 408, emphasis in the original):

68

Chapter Three

replicator – an entity that passes on its structure largely intact in successive replications. interactor – an entity that interacts as a cohesive whole with its environment in such a way that this interaction causes replication to be differential. With the aid of these two technical terms, selection can be characterized succinctly as follows: selection – a process in which the differential extinction and proliferation of interactors causes the differential perpetuation of the relevant replicators. Replicators and interactors are the entities that function in selection processes. Some general term is also needed for the entities that result from successive replications: lineage – an entity that persists indefinitely through time either in the same or an altered state as a result of replication. By defining these entities in terms of their function in selection processes, any entities having “the relevant characteristics belong to the same evolutionary kind. Entities that perform the same function in the evolutionary process are treated as being the same, regardless of the level of organization they happen to exhibit” (Hull 1988c: 402). This generalisation allows non-biological processes – such as sociocultural change – to be treated as the same kind of process, i.e. a selection process, so long as it is possible to identify replicators, interactors, the selection process and the lineages in them.1

Hull is not general enough Although innovative and more focussed than the previously discussed authors on what really matters – that is, the general principles behind the selective phenomena – we still see a few problems with Hull’s actual implementation of this new method in tackling the evolutionary analogy. First, although the definitions he introduces for the entities involved are much more general than the concepts normally used, such as ‘genes’ and 1

Hull acknowledges that these distinctions are not entirely new. Dawkins introduced the distinction ‘replicator’/‘vehicle’ in 1976 but Hull improves the second concept into his own ‘interactor’.

A Type Hierarchy for Selection Processes

69

‘phenotype’ which would apply only to the biological domain, his procedure is still moving upwards from the instances of biological concepts to their generalisations. As a result, his definitions are not as general as would be desirable. For example, interaction is defined in terms of replication. Although this is quite acceptable in the context of biological selection, it should not be required in a generalisation of selective processes, where such a definition might depend on that of selection at most. In brief, this lack of generalisation causes his system to be still too closely related to the evolutionary biology side of the analogy.2 Secondly, the TH approach causes the positive analogies between tokens of a specific TH to reveal themselves in the topography of the diagram, the negative ones being those which are not represented in the hierarchy. This, however, is not the actual approach used by Hull to tackle the evolutionary analogy. Although he sets out to find the laws of a more general theory of selection, firstly he finds and analyses a number of neutral analogies, such as fitness, replication and lineage, and shows they are positive analogies, (i.e. that these features of biological evolution can be found in scientific change). Secondly, he then argues against the detractors of the evolutionary analogy by showing that their alleged negative analogies, such as the rate of change, are actually positive ones (Hull 1988c: 440-1). Finally he shows that alternative models, such as Lamarckism, are plainly wrong (p.452). So, although Hull is partially immune to the difficulties of the traditional approach to analogy mentioned above, in his “analysis of selection processes sufficiently general to accommodate all sort of selection processes” (Hull 1988c: 439-40) he still feels the need to reply point by point to all the disanalogies found by the critics of attempts to assimilate biological and sociocultural change to a common mode of explanation. In contrast, a properly built type hierarchy would make these point by point replies unnecessary, because disanalogies are not represented at the higher level of supertypes (see above). This is exactly where the arbitrium of the comparison’s view of similarity resides, in focusing on the bottom layer of the hierarchy and placing the entire burden of similarity comparisons on the local links between this bottom-most slice (Aronson et al. 1994: 106).3 2

In a different context, Skipper (1999) advances a similar criticism to Kitcher. At the end of their Introduction, Hull et al. (2001) seem to recognise the need of unbiased generalisation and their whole paper is an attempt to build it. In our view, they end with too general a theory though. 3 It could be argued that Hull’s defence is different from the statement of his theory. We believe, however, that his defence is just revealing the procedure he

70

Chapter Three

In light of the complexity of the TH approach to similarity, we believe Hull’s attempt should be further developed and its complexity unravelled. First of all, a really general theory of selection should comprise cases of selection which do not entail any evolution, such as, for example, the simple procedure of selecting a coin from my pocket, or the selecting action of a sieve on flour.4 Secondly, the bare derivation of both the selection theories to be analysed from a single supertype is still too reminiscent of the comparison approach to similarity, where individual analogies are one by one slotted into the positive, negative or neutral box. That is what Hull mainly does in his work, as we have illustrated by critiquing Hull’s focus on the bottom layer of the TH. While Hull always has biological selection and the organic world in mind when he tries to define the main features of selection processes, by following a bottom-up procedure, full generality is only achieved, we believe, by trying to define the more abstract types first, and then descending to further specify more specific cases of subtypes, down to the individual instances of the lowest subtypes.

Darden and Cain We also find that Darden and Cain (1989) take a similar line to Hull. They aim to “discuss abstract characterizations of selection theories” (p.106).5 Their main point is that selection processes solve adaptation problems, and in their search for an abstract characterisation of selection processes they identify a number of steps in which the selection process can be broken down, which are easier to abstract. These are: 1. Preconditions (e.g. a population in an environment and variation) followed to elaborate his theory, which is the one of the comparison view and which results in a very slim hierarchy. 4 This level, of course, is very general and these are extreme examples only given for illustration purposes. What we mean, here, is that in our search for generalisations of processes the details need to be really forgotten, in order to lose any bias deriving from the particular examples we have in mind. Clear evidence of this risk is, for instance, the fact that most people believe that evolution only happens by natural selection, forgetting other mechanisms such as genetic drift. This mistake occurs because they do not think of the general concept of evolution, but what they have in mind is the more specific ‘Darwinian evolution by natural selection’. Only later will more detail be needed to populate those regions of the hierarchy where our task will require more precision. 5 Although they do not use the more detailed treatment that Hull offers in his book, which wasn’t published until shortly after their paper.

A Type Hierarchy for Selection Processes

71

2. Differential interactions between individuals and environment 3. Short range effects, such as differential survival (or even more abstract: benefit or suffering of individuals) 4. Longer range effects (e.g. increased reproduction of something associated with more prolific individuals) 5. Very long range effects (e.g. accumulation of benefits in lineages of individuals) Following this schema, they provide abstractions for Natural Selection, Clonal Selection (Antibody Formation) and Neural Selection. The following, for example, is the abstraction for NS (p.116-7): 1.

2.

Preconditions a. b. c. Interaction a. b.

3.

Ys exist Ys vary as to whether they have property P Ys are in an environment E with critical factor F Ys, in virtue of possessing or not possessing P, interact differently with environment E Critical factor F affects the interaction such that

Effect a.

4.

5. 6.

the possession of P causes Ys with P to benefit and those without P to suffer. Longer-range effect a. 3 may be followed by increased reproduction of Ys with P or reproduction of something associated with Ys. Even longer-range effect 4 may be followed by longer-range benefits.

Darden and Cain are too general Although in the right direction, we have to express some criticism of this approach. The role of reproduction in NS is not emphasised enough, and this is revealed by their treatment of Neural Selection, where increased reproduction is replaced with stimulation and reinforcement. In their summary table at page 124 of their paper, where they provide abstractions for the various concepts involved, they do not have an abstraction for

72

Chapter Three

‘reproduction’,6 while a ‘metric for measuring differential effect’ is the abstraction for differential survival (NS), lymphocyte activation (Clonal S.) and stimulation (Neural S.). Here we reproduce a portion of the first two columns of that table (Table 3). In the first column are the variables used above as placeholders, in the second column are the abstractions of possible instantiations for those placeholders. For instance, ‘variant’ could be instantiated by ‘allele’ or ‘individual organism’, which are possible values for the variable Y. Variable Y P nature of P E F Z Benefit

Abstraction variant interactor benefitor variant property Range of variation7 environment critical factor level of sorting metric for measuring differential effect

Table 3 – Darden and Cain’s abstractions More importantly, below we will soon argue that as this abstraction can also be used for processes that nobody would like to place side by side with NS, it is, as a result, inadequate.

3.3 Type hierarchy of selection processes A provisional hierarchy We will now attempt to develop Hull’s work to take into account a more complete characterisation of selection processes, in terms of a suitable type hierarchy. Given that the main illustrative feature of this

6

We believe that, in a complete characterisation of selection processes, reproduction or its abstraction (replication) is needed to make a selective process evolutionary. This will be explained more fully later. 7 For example continuous or discrete.

A Type Hierarchy for Selection Processes

73

approach is the representation of a type hierarchy in a semantic network, we start with a provisional one.8

Figure 6 – A provisional type hierarchy for selection processes

If we generalise the meaning of the term ‘to select’ to a more abstract ‘to separate off’ or ‘to sort’, deprived of their connotations of intentional choice, we would place the type selection quite high up in the hierarchy. Of course Hull and all evolutionary epistemologists mean by the term ‘selection’ the type of selection acting on living beings and, allegedly, on cultural entities, but we need to enrich our lexicon in this domain if we aim to deepen the hierarchical description of selection processes. In other words, for a TH to include several types of processes such as biological evolution, sociocultural evolution etc. any more specific definition of ‘selection’ would only make it applicable to one of the processes. The point is that terms acquire specificity as we proceed towards the bottom of the TH, where they inherit from different combinations of more abstract supertypes. This is exactly where the power of this method resides. We need to unpack complex concepts into simpler 8

The term ‘selection’ in the lower nodes has been left for clarity reasons and historical continuity. It does not imply that selection, in those nodes, has a prominent role.

74

Chapter Three

(more abstract) concepts, which are of little use on their own, and their combination, explicitly shown in the TH, will produce the more familiar concepts we want to analyse. As a consequence, we believe the bare term ‘selection’, which is very general, is more suitable for basic processes of ‘separating off’. This opens up the possibility of deriving a very simple subtype which we call ‘mechanical’, which in turn is instantiated by the processes linked to it in the picture.9 We call these mechanical selection processes, because what is selected and where it is placed is governed by simple physical conditions: spatial dimension, shape, position and energy level.10

Mechanical Selection This, for instance, is the place where the selection mechanism provocatively proposed by Catania (2001) would find room, here labelled as gravitational selection. Catania’s suggestion is to “start with systems of selection in the broadest possible terms, and then to move toward a taxonomy of selective systems with the sorting depending on a number of different dimensions”. As an example he proposes a system of scattered particles in the vicinity of a gravitational field produced by a planet, which end up sorted by gravitational effects (sent into outer space, attracted into collision with the planet or put in a different orbit) in some time, as in the case of Saturn rings or the asteroid belt between Mars and Jupiter. Although Catania’s point seems to set off on the right foot, he does not think that generalisation should be trying to break selection into components. However, Hull et al. merely see Catania’s objection as an attempt to create too broad a concept of selection and are quick to dismiss it. In our view, Catania’s example involves a separation process, generated by the interaction of the scattered particles with the planet(s). Apart from the occasional collision between particles, which would introduce some sort of (uninteresting) ongoing variation in their population, no other change can be identified in the constituents, nor is any replication 9

Although these instances could be linked directly to the node ‘selection’, because the node ‘mechanical’ inherits its properties but does not add any new ones, we found it more clear to group together these processes in a subnode. 10 It could be objected, for example, that sorting coins from a pocket is not mechanical in the way gravitational or natural sieve actions are, because it involves intention. At this level of granularity, however, in which only the physical properties of the selected objects are counted, intention is not explicitly represented and it does not differentiate ‘coin sorting’ from the other two nodes. It will appear in a more refined version of this diagram, offered in one of the next pages.

A Type Hierarchy for Selection Processes

75

involved.11 Consequently, we feel confident in treating the gravitation selection process as an instance of mechanical selection. On the other hand, if we consider the abstraction suggested by Darden and Cain (1989), we can instantiate the variables in the following way, by extending Catania’s own remarks: Y: particle group of Ys: group of particles P: total energy of particle (mass and orbit) nature of P: continuous variation in allowed orbits12 E: surrounding gravitational fields F: resonance13 Z: individual particles’ orbits benefit: orbit stability14 longer-range effect: accumulation of particles even longer-range effect: strictly confined structures of particles (rings, belts) This, we believe, is enough to show that Darden and Cain’s abstraction is more abstract than both they and Hull would accept.

Evolution In further developing our hierarchy, we assume that selection processes become relevant to our discourse by the addition of the properties of another very general type – ‘evolution’ – meant as the “gradual development of something” (Soanes and Stevenson 2005).15 From the arcs in the diagram it is clear that we do not think that mechanical selection, as represented here, inherits the properties of evolving processes: that is, we 11

We cannot see any relevant “replication in the cyclic character of the orbits within the rings”, as Catania claims. 12 Because of perturbations, the particles keep changing orbit and the space of variation is not discrete. 13 Particles tend to stabilise into orbits which reinforce each other or get expelled to outer space. 14 As defined in this instantiation. 15 Analogously to what has been already pointed out in the case of the term ‘selection’, although ‘evolution’ brings to mind the biological process usually driven by NS, it is worth stating again that here it is used in the more general sense of ‘gradual change’ of something which conserves its identity. After all, physicists talk about the ‘evolution’ of states of a system, meteorologists discuss the ‘evolution’ of atmospheric phenomena and so on.

76

Chapter Three

do not see sifted flour or jumping electrons as ‘evolving’ in a sense that would be applicable to this domain. Drawing on this perspective, we also need to generalise further the other key concepts that Hull uses to build his abstraction of selective processes. We suggest we modify (generalise) two of Hull’s definitions as follows: interactor – an entity that interacts as a cohesive whole with its environment in such a way that this interaction generates a sorting process on the population it belongs to. selection – the process in which some of the interactors are sorted from the rest of the population. This further generalisation also has the effect of uncoupling replication and interaction, while in Hull’s definition the latter is unnecessarily defined in terms of the former. It is true that, in his view, the only interesting interactors are those which interact with their environment to cause differential replication – that is, the interactors which “function in selection processes” (p.409). On the other hand, if we truly generalise all concepts in the way we are proposing, interaction and replication should be treated as separate types. An abstract interaction, however, would become too vague a process to be given a status of its own in our hierarchy, because it would be a supertype of selection. As a result, only selection would inherit from it in our TH, so we prefer to tighten the diagram and represent it implicitly, just as a feature of selection. Now, if we conceive of an evolutionary process which is selective and develops by means of cycles of replication, we have the subtype ‘differential selection’, as it stands in our Figure 6. This is a crucial point, for if we take this central node, move all the labels from the supertypes to it, link natural selection and science/culture selection to it, then remove everything else from the diagram, we go back to Hull’s view in Figure 5. Our diagram, thus, does not provide a different picture of Hull’s view, it only adds the additional detail which emerged from our discussion. In our view, however, we need the greater level of complexity we have added to this description in order to: 1) reflect the different properties that apply to the phenomena identified by different nodes; 2) answer several of the objections still raised by Hull’s approach; and 3) provide a general analysis that is broad enough as a whole to comprise most phenomena involved in the subject, and narrow enough in its components to characterise them as much as possible (or, at least, to offer a more stable

A Type Hierarchy for Selection Processes

77

basis for their further analysis). The benefits of this will be clear in the final version of the TH.

The complete Type Hierarchy In order to complete the proposed TH, we also need to distinguish between a type of processes guided by some intentional mind and another type of processes where some sort of ‘genetic’ material is present. These crucial distinctions are generated by what, in our view, are the main problems all theorists of evolutionary analogies – including Hull – have not managed to solve. Only from these very articulated subtypes can we find full characterisation of the tokens in the diagram. The power of this approach is shown by the fact that different type nodes are not mutually exclusive in the inheritance relations to the instance-tokens. That is, most of the issues affecting other accounts of the problem, which we will discuss soon, just dissolve in the TH approach, by letting, for example, an instance of a selection process inherit from more than one type. This point will be made clearer by adding some more nodes to our previous graph into Figure 7, where the dotted lines represent links that might be more controversial and where Differential selection has been replaced by a finer hierarchy.

Intentionality and Artificial Selection Let us start with the problem of intentionality. This is, as already explained, one of the most difficult problems encountered when trying to unify the processes of selection in culture and biology, and Hull uses the counter example of the process of artificial selection which, though guided by intentional minds, is not conceivably different from the slower process of natural selection. As it stands, we believe our diagram offers a better solution to the problem by simply linking Artificial selection to Directed selection instead of bare Selection, to which Darwinian selection is linked. In other words, given a TH representing all types and links we believe are relevant to represent all the processes we are analysing, the three types of selection are necessarily represented at different leaves in the graph, because they do not share all their supertypes. As a result they do not coincide.

78

Chapter Three

Figure 7 – A type hierarchy for selection processes

A Type Hierarchy for Selection Processes

79

To provide a more concrete illustration, it is possible to show that artificial selection is, for instance, no more problematic than having amphibian vehicles without necessarily have to identify trucks with boats. Are amphibian vehicles trucks or boats? They inherit most of the properties of both types of vehicles, but this does not imply that trucks and boats coincide. A common supertype for both of them might be the type ‘vehicle’, or something intermediate such as ‘surface vehicle’ – the actual supertype here does not matter as long as it is correct and relevant, that is it helps linking all the entities we are analysing (trucks, amphibian vehicles and boats). However, if we detail the classification a little bit, we would end with a hierarchy similar to Figure 8: So, although the type ‘amphibian’ shares all the properties of trucks and boats, this does not mean that trucks and boats share the same properties between them. This is due to the presence of the intermediate types which introduce additional but distinct characterisations.

Figure 8 – A type hierarchy for boats and trucks

Analogously, the existence of artificial selection does not necessarily imply the identification of sociocultural with natural selection, and sociocultural selection does not represent, in this approach, a problem to be solved, because the features in common between NS and AS – that is, evolution and replication – are preserved.

Cyber Selection We can also think of an imaginary case of what we call Cyber selection, which would raise similar difficulties in the old approach. Imagine we produced very complex machines provided with an extensive knowledge and autonomy to ‘know’ what they are made of, with the ability to search for the components needed to build exact or better replicas of themselves, assemble the components and transfer to this next generation their own software. This we can imagine to be an expert system which changes over time to perform ‘better’, according to certain

80

Chapter Three

indicators. Finally, imagine that these changes are produced by generating slight random variations in the hardware/software setup of the machines which are judged, by its software, to be moving in a beneficial direction, according to the past history/experience of the machine lineage.16 This thought experiment is clearly designed ad hoc to provide another example of a selection process which would be difficult to discount as being analogous to NS but one which is definitely goal-driven by intentionality, if we agree with the intentional stance. Within their software, the machines would have a set of instructions which specified the list of their components, the way they are assembled, instructions on what specific procedures and tools are needed to assemble the components, how to transfer the software to the newly built machine, and so on. We can also imagine the machines operating in some sort of environment where, given certain conditions, the machines fail or ‘decide’ to halt and do not reproduce, or add all sorts of features such as auto-healing and information exchange between machines in order to optimise performance and spread the up-to-then advantageous features of some machines to others. Would these machines evolve by natural selection? Well, we can see that their selection process inherits all the properties of Darwinian selection. However, originally they were intentionally built by human beings, and in the following generations by other machines which try to ‘guess’ some improvements for the next generation. If we hold a position close to Dennett’s intentional stance (Dennett 1979) we should let cyber selection to inherit from directed variation, and even from directed selection as in the case of artificial selection, in case the failure to reproduce is ‘decided’ by the machine or by human intervention (dotted line in the diagram). If we don’t hold the intentional stance and if we also think of these machines as totally autonomous in the selective and variation processes after the first generation is built – that is to say, no human being intentionally gets rid of ‘bad’ machines and leaves around only the ones which are ‘good’ – then those links cannot be drawn, and cyber selection and NS would be two instances of the same type of process.17

16

Most of these features have already been implemented separately in different systems. We are not aware of any system currently integrating all of them. 17 Darden and Cain (1989: 111) think that ‘as long as preexisting diversity exists, the details of the mechanism of its production can be omitted from a selection type theory’. We see this judgment as a consequence of the reduced role of replication in their analysis.

A Type Hierarchy for Selection Processes

81

Drift The evolution of organisms over time is not driven by natural selection alone: other factors come into play. One of those is genetic drift. When the number of individuals in a generation is quite small, an advantageous characteristic can disappear if its bearers are unlucky. Conversely disadvantageous or neutral characteristics can prevail over others if their bearers are fortunate enough. Hull sees drift as differential replication in the absence of interaction; in other words, there is no selection process at work and he updates his definition of ‘selection’ from earlier versions by explicitly emphasising its causal character (Hull 1988c: 410). As a result, it is correct to assume that drift inherits from ‘replication’ and ‘evolution’ only, not from ‘selection’, hence the way it is represented in Figure 7. A more detailed characterisation of drift shows that its position in our TH is consistent with the current understanding of it. Genetic drift is “the random change in the frequency of alleles […] due to accidents of sampling caused by random variation in rates of survival or reproduction by different genotypes”.18 Genetic drift does not cause adaptation but only variation: it constitutes evolution by chance alone. Thus we can conclude that the TH we are proposing characterises properly the genetic drift phenomenon, and is also immune to the mistake of including drift within selection processes.

Immune system The theory most widely cited for explaining the operating of the immune system when fighting the attacks from infections is the clonal selection theory, which stems from the work of Jerne and Tonegawa (Cziko 1995). This theory states that, in mammals, B lymphocytes are constantly produced in large quantities by the bone marrow. The antibodies attached to these cells are all different because the genes relevant to their formation are inherited as shuffled fragments from the parent cells and, when “the DNA segments are combined to form the complete B lymphocyte gene, new DNA sequences are added at random to the ends of the fragments, ensuring even more antibody diversity” (Cziko1995: 44). The enormous, blind variety thus generated makes almost certain that at least one of the antibodies on one of the B lymphocytes will bind with a given antigen. When this happens, for example in the presence of an infection, the attached B cell is stimulated to 18

(Futuyma 1998: 304) Genetic drift is strictly related to the subdivision of species into local breeding units which exchange genes to a greater or lesser degree.

82

Chapter Three

reproduce itself into clones, which will grow exponentially in number, enabling the infection to be defeated. This selectionist theory of the antibody formation still has things to explain, but it is not affected by the major problems that the previous alternatives had encountered, and ever since its formulation all accounts of natural selection have incorporated it as a special or parallel case, as in Darden and Cain (1989) or in Hull et al. (2001). There is no problem in placing clonal selection just beside NS, that is in considering it as an instantiation of ‘Darwinian selection’. All the characteristics of somatic selection of B cells are definitely instantiations of all the characteristics defined by the type ‘Darwinian selection’ and its parents. Thus, we definitely agree with labelling the somatic selection of B cells as Darwinian selection according to our TH.

3.4 The usual objections How does our approach behave when tested against the main objections that Hull’s target papers raised? We will answer the most relevant.

Variation Cain and Darden (1988: 165) complain that Hull does not pay enough attention to variation which, in their opinion, should be given a welldefined role. They link this problem to the excessive emphasis that Hull gives to the role of individuals as opposed to the neglected role of the environment (see also Darden and Cain (1989)). Related to this difficulty is the problematic coupling of variation and selection in conceptual evolution, which is a notorious objection to the possibility of explaining conceptual evolution in full Darwinian selection terms. The coupling is the result of the scientists’ search for ‘variations’ on their ideas which would better pass the selective tests. In our TH we have reserved an exclusive place to variation, as a property of evolutionary processes. Variation occurs, for instance, when genetic mutations or modified ideas are produced, if we restrict ourselves to examples of biological or conceptual variation. This placement is what lets drift become part of/belong to the diagram without being classified as a selection process. It is only thanks to variation that lineages can develop as a continuity. Without variation we would only have mechanical selection, not evolving lineages. In sociocultural selection, however, variation is of a special sort – it is the one intended by actors to pass the

A Type Hierarchy for Selection Processes

83

selection stage as efficiently as possible. So, the introduction of ‘intention’ is actually constraining the way Sociocultural Selection (SCS) inherits from evolution. In other words, it is instantiating the main property of ‘evolution’, i.e. variation, in a specific manner, that is, in a way that is constrained by the presence of the supertype ‘selection’. This does not happen in the case of NS, which does not inherit from ‘intention’. Its inheritance of the properties of ‘selection’ is independent from the inheritance of the properties of ‘evolution’ and vice-versa. There is no constraint on variation.19 It is in this sense that we argue the introduction of the additional type ‘intention’ is necessary and not ad hoc.20 This is clearly a key point, given the fact that even Hull (1988c) is forced to admit that intentionality is the chief disanalogy between rational selection and NS. However, he emphasises that “this difference in kind does not make much of a difference in degree” (Hull 1988b: 262). We see this concession as an implicit acknowledgement that the two processes, after all, are of a different kind, even though some observations about them may be fairly close. This is exactly the opposite of what Hull is trying to prove, that the two selections are two instances of the same process. To add more confusion, Hull (1988c: 472) says that selective phenomena should be classified initially in terms of the character of the entities selected, and only then according to the character of the acting selection process. We do not believe this is tenable. Think about the widespread national lotteries, where 6 or more numbers out of 4921 are randomly extracted every week to be matched with those previously and blindly chosen by the players. The few matching combinations, if any, usually bring to the winners monetary peace of mind for the rest of their life. Now, imagine somebody who ‘slightly’ changes the extraction procedure and intentionally selects the numbers that match those printed on their coupon. Although the entities selected are still the same, we cannot say that the two processes are the same. This is quite a simplification of course, also because no variation is involved, but the example shows that although only intentionality distinguishes the two cases and everything else stays 19

Although there might be, as in the case of processes which inherit from both sociocultural selection and Darwinian selection. 20 It possible, of course, to conceive unintentional or quasi-unintentional selection or variation in sociocultural processes, for example in the form of transcription errors or unexpected consequences of intentional actions. We regard these as interesting exceptions that may be worth further analysis but which would not alter the main points of our argument. 21 In the UK.

84

Chapter Three

the same, nobody would classify the random extraction and the fixed one as the same process. And yet, artificial selection is universally mentioned (also by Hull (1988c: 473)) as a striking example showing that intentionality does not bring anything relevant to selection processes.

Progress Strictly related to the intentionality problem is the difficulty arising from the progressive nature of science. Hull’s claim that, after all, intentionality does not alter selection processes in a relevant manner is untenable because intentionality makes selection processes progressive. While there are meaningful ways to claim that science is progressive22 the opposite seems true for biological evolution (Futuyma 1998: 700), as seen in the previous chapter, and Henson (1988: 193) characterises this distinction by opposing the adaptive nature of biological evolution to the progressive nature of science. This distinction stems from the radically different characters of the selection benchmarks in biological and conceptual change: environment and natural laws. In the case we agree that if a selection process is acting on scientific ideas – i.e. that this selection is driven by the immutable laws of nature that those ideas are supposed to describe – then the selection outcome is progressive, at least in principle.23 Conversely, biological selection is acting on the basis of a mutable environment, so evolving entities cannot ‘progress’ in a fixed direction, but only adapt to or coevolve with whatever environment they happen to be immersed in. Hull’s answer to this difficulty is a ‘damage limitation’ one: only “if scientific change is both locally and globally progressive, [does] an important difference exists between the two” (Hull 1988c: 464). After reminding us that, after all, science is a relatively recent and initially spatially localised phenomenon he then suggests that, as such, its progress might count as just local. Eventually, he concludes that the source of this difference “lies in the relevant ‘environments’ of these two processes” (Hull 1988c: 466), not in the mechanism of the processes. First of all we see this claim as contradicting his already mentioned suggestion that selective phenomena should be classified initially in terms 22

However, there is disagreement about what these ways are. See Shapere (2000) for a short survey. 23 Although this is a controversial topic, the predominant view is that science progresses, by getting closer to truth, by becoming simpler, more explanatory or less erroneous. This has been briefly discussed in the previous chapter. We cannot treat this topic more fully in this space but we will assume the progressive nature of science in a general sense, which is based on the postulated regularity of nature.

A Type Hierarchy for Selection Processes

85

of the character of the entities selected and, only then, according to the character of the acting selection process. This contradiction arises if we consider what we believe is the key difference here: in biological change, the selected entities are part of the environment against which they are selected, so broadly speaking they act as ‘selectors’ against the others; conversely, ideas and theories, if we think of them as ultimately selected against the natural laws, do not take part in the environment.24 The real difference, then, is not just variable versus immutable environment, because this difference could result in a fuzzy distinction between local and global progress which could induce confusion by the fact that globality is conceivable as a limiting case of locality. For example, it could be argued that there is no difference in kind but only in degree, because an immutable environment is undistinguishable from an environment which does not change in a measurable way. Conversely, if we agree that, in biology, the selected entities are part of the environment against which selection occurs, then environment changes by definition at a rate that is comparable with the rate of change of the selected entities. In science, instead, if we agree that selected entities are distinct from the environment against which selection occurs – that is, natural laws – this difficulty does not arise and both the selected entities and the selection process are classified differently from the biological ones, leaving Hull’s argument without a point. The only way to save Hull’s account as it stands is to completely assimilate the scientific enterprise to the sociological realm,25 negating any special status granted to science as referring to discoverable regularities in nature, regarding it instead as the product of a constantly changing social and intellectual milieu, of which it is a part. From this perspective you would lose not only the key differences in the environments and so in the selection processes, but also the progressive nature of science. In the TH we are proposing, the problem of progress is not directly represented, because it is embedded in the way intention is located in the TH. Scientific progress is the result of the comparison of theories and 24

It could be argued that scientific theories compete with one another in some sense, as animals do. However, while in the distinction, for example, between one organism and the selecting environment it is possible to merge the rest of the organisms with the rest of the environment under the same type ‘selecting environment’, an analogous operation is not possible in the trichotomy theory/rest of theories/natural laws. 25 Campbell stresses the need to shift away from biological parallelisms but also that “a selection theory for conceptual evolution in successful science will have to be a sociology of science” (Campbell 1987: 177).

86

Chapter Three

ideas with the results of what we think are partially independent and reproducible tests carried out on nature itself, that is experiments, with the intention of making the former as consistent as possible with the latter. Thus, progress is embedded in the inheritance link of sociocultural selection from intention. Also note that there is a partially similar route from Intention to Artificial selection, which can be seen as progressive if we track the evolution of intermediate results of selection between the natural organisms and the final organisms displaying the desired target characters.

3.5 The evaluative step If we only look at the problem of producing a more intersubjective method to assess the claim that two processes are the same or similar at some degree, the need for characterising these processes as the instantiations of more abstract, general processes is recognised by most of the authors we have mentioned here, although in some cases not too clearly or explicitly. None of them, however, takes the further step of assessing whether there is a way to weigh the ‘closeness’ of the two processes: whether it is possible to express more clearly or even measure the claim that the two processes are similar or even the same. The work of Aronson et al. (1994) may offer an illuminating point. In their analysis of metaphor, Aronson et al. (1994) suggest that metaphor “takes place by establishing new semantic linkages as a result of coarse-grained masking” (p. 101) of a given TH. If we build a TH with all the relevant links between the relevant type nodes of a given utterance, the resulting semantic network can convey a number of meanings, depending on the context of the speech. The role of that context, according to the authors, can be represented as a ‘mask’ superimposed on the hierarchy, which will hide or reveal different links and nodes. A literal use of speech, where only the ordinarily understood aspects of the TH are communicated, corresponds to a mask revealing only a few of the elements of the hierarchy, i.e. those representing what is supposed to be the real state of things. Conversely, a ‘coarse-grained’ mask which exposes more links among semantic domains can represent a metaphorical speech: that is “the redescription of one domain in terms of the generated hierarchy and the associated beliefs from another”. Is it possible to define an operation on the TH that, instead of discriminating metaphorical degrees of speech down to the literal level, would allow us to distinguish levels of similarity, down to the identity?

A Type Hierarchy for Selection Processes

87

First of all we argue here that the TH approach is less controversial because it is easier to agree on a certain ontology than on the relative relevance of a set of similarities and dissimilarities. The ontology represented in the TH derives from our understanding of the world; that is what we think are the general laws that are relevant to the area investigated and how they relate to one another. Let us suppose we possess an ontology-based means for evaluating analogies. Once agreement has been reached on the ontology that best describes the domain, the resulting analogies between objects of the domain are less questionable, so the problem has shifted from the choice of suitable analogies to the choice of the correct TH. Though less arbitrary, this process is of course still open to debate, with the risk being that one will introduce only the types and links that support the analogy you may already have in mind, and omit those that would not. This risk is unavoidable because the TH, as an abstraction and simplification of the subject, is itself a theoretical construct. However, We agree with Aronson et al. (1994: 106) that the setup of a suitable ontology is an empirical matter, which should eventually be tested against what we believe is the current status of knowledge of the world; as such it is debatable, as is any theory based on empirical evidence.26 For the next step – that is, the decision whether two selection processes are two tokens of the same type of process – the evaluation is now possible on the basis of quantitative measures, in which lies the real power of the approach we are proposing. We suggest the following criterion: two processes are the same if and only if they have all their direct supertypes in common. So NS and Clonal Selection (CS) are the same process, while NS and Artificial Selection (AS) are not. This criterion is easily justified: in a given TH, only instances inheriting from the same types can be assimilated to one another and be regarded as the same. If they differ even in just one supertype, they will display differences in their inherited properties which are considered to be relevant in the chosen ontology because they follow from the resulting classification.27 On the other hand, to measure the degree of similarity between two processes we propose the use of a ‘collapse’ operation. A collapse in a TH is the removal of a type node followed by the creation of direct links between each of its subtypes/instances and each of its supertypes (if any), unless a direct or

26

It is worth stressing that arbitrariness is not removed, only one of its layers is; this is the layer where similarities and dissimilarities are decided and weighted. 27 This justifies the criterion as necessary. If we add that the chosen ontology is detailed enough for our aim, this justifies the criterion as sufficient.

88

Chapter Three

indirect link already exist between the given instance and supertype.28 From this operation, we can formulate the following definition: in a given TH, two processes A and B are more similar to one another than two processes C and D if A and B become the same process by collapsing a smaller portion of the TH than is needed to make C and D the same process. Now, to make AS and SCS the same we have to collapse one node (Directed variation); to make Darwinian Selection (DS) and SCS the same we need to collapse three nodes (Directed variation, Directed selection and Intention); to make AS and DS the same we have to collapse two nodes (Directed selection and Intention). Thus, in this metric, the similarity between AS and SCS (1 node collapsed) is stronger than the similarity between DS and SCS (3); the similarity between AS and SCS (1) is stronger than the one between AS and DS (2).

A brief note on ontological and pragmatic approaches The approach we are proposing here is very general, in the sense that it can be extended and applied to evaluate all cases of proposed analogy or identity between processes or any other couple of entities which are the subject of investigation. It is also in this sense that we call it an ‘ontologybased’ method for evaluating analogies. The method, however, does not aim at the provision of an evaluation of the fruitfulness of analogies, which we believe would be strongly domain-dependant and less prone to intersubjective analysis. The ‘standard’ approach to the analysis of analogies in science (Montuschi 2000: 306-307) distinguishes between realist and anti-realist views. In this context, anti-realist views regard models as heuristic devices, with no direct relation with reality. Analogies, then, can be helpful in discovery and explanation, but no claim is derived from those analogies about the real status of things, and good models are those bringing new discoveries and better explanations. Realist views, conversely, regard models as analogous to reality in some way – that is, a good model has to represent approximately the real world. Both these views regard theories as “static ‘textbook’ entities” (Montuschi 2000: 306), formal systems whose meaning is provided by their models. A 28

For instance, the collapse of the node Directed variation in the diagram in Fig. 7 entails its removal and linking both Sociocultural selection and Cyber selection directly to Intention and Evolution.

A Type Hierarchy for Selection Processes

89

different approach is provided if we take into account the dynamic, historical dimension of theories, which themselves are seen as models or analogues of reality. Here, however, we have a different problem. We are concerned with the analogies or identity, if any, between theories themselves; in other words, we are building a metamodel of reality which is also comparing models of two different domains, i.e. scientific and biological change. Most of the work published on analogies stems from the analysis of linguistic metaphors and the distinctions between literal and metaphorical discourse. One of the features of the recent analyses, performed by the Interactive View and by Way herself, has been to account for the asymmetry of linguistic metaphors, because it is one thing to say ‘surgeons are butchers’, but quite a different thing to say ‘butchers are surgeons’. Way calls the first term of the metaphor the ‘tenor’, the second the ‘vehicle’ which ‘carries’ the metaphor.29 This asymmetry is mirrored in the scientific use of analogy, where an uncontroversial or very clear system or theory is taken as a model for a system or theory to be investigated or developed. In the evolutionary analogy this translates into the attempts to derive simple assertions about the nature of scientific change, seen as a process analogous to or coincident with natural selection, or into the attempts to clarify our treatment of scientific change with insights from biological evolution. The two approaches are not mutually exclusive, though usually one is more emphasised than the other by any given author. We would like to note, however, that in the way we are using the TH approach here, this asymmetry is necessarily lost, our aim mostly being the analysis of the two processes at issue from an ontological point of view, in order to show whether the processes are the same, or, if not, what degree of similarity can be assigned to them. If we take the TH just as the way we represent our knowledge of the world on the matter, without the additional step of accounting for an asymmetrical metaphor, the resulting type hierarchy is neutral to any traversing direction between two given tokens. This is in the full spirit of the Cain-Darden and Hull approaches, because if we characterise a process as a more abstract process of both cultural and natural selection, any asymmetry is lost in the operation. However, in real terms, this asymmetry is still present in their approaches, with no attempt to apply insights from studies on scientific change to explain how biological change works. This would, in principle, be perfectly legitimate: if we assume that ‘atom’ and ‘solar system’ are analogues because they both inherit from the type ‘central force system’, 29

This terminology is from Richards (1936). Alternative dichotomies are, among the others subject/modifier (Beardsley), frame/focus (Black), primary/secondary system (Hesse).

90

Chapter Three

we can assert their similarity in both directions and say that ‘an atom is a miniature solar system’ as well as ‘the solar system is a giant atom’. It could be objected that this does not really happen, because the actual scientific process is the use of analogy to provide explanations and insights from a well known area into a new or still unclear one. However, once the explanandum has been explained the full symmetry should be restored; more importantly, Hull argues for the identity of the two processes as instances of the same supertype, so the asymmetry of any analogical approach should be long removed. While most of the earlier defenders of the evolutionary analogy restricted themselves to listing and analysing analogies and disanalogies,30 without going into any detail on the actual advantages of doing so, Hull – who reveals above all others an ontological interest – is the first who tried to show that this analogy can actually be used as a working tool to explain specific cases of scientific change. His work, then, is committed both to the ontological approach, to show the identity of the two processes, and to the practical one, that is, to use the analysis to explain real cases. However, these two approaches seem to clash in Hull’s book. For example, in the section where he deals with the problem of progress, he writes that “[i]f this difference were due to differences in the mechanisms involved in biological versus conceptual change, then it would count against any attempt to produce an analysis of selection processes that is equally applicable to both sorts of change” (Hull 1988c: 466), which reveals that the identity of mechanisms is the principal issue. A few pages later, however, he dismisses the intentionality problem by claiming that “[e]ven if one interprets the presence of intentionality in rational selection as introducing a difference in kind between it and natural selection, it is a difference in kind that does not produce much of a difference in degree” (p.474).31 This seems quite a U-turn to us, because if intentionality introduces a different selection mechanism, then this prevents the subsumption of “both sorts of change” under a common heading in the analysis. From the point of view of our TH, we see this as an acknowledgment of the crucial role that intentionality plays in our diagram. Thagard (1993), in his chapter ‘Against Evolutionary Epistemology’, takes a pragmatic approach and claims that “a good analogy is simply one that contributes to the solution of a given set of problems” (Thagard 1993: 102). In other words, an analogy or identity may be weak and unsound on 30 31

In response to their critics, who used the same method. The same remark is in Hull (1988b: 262), see section Variation above.

A Type Hierarchy for Selection Processes

91

ontological terms but still somehow useful in practice. The two questions are distinct. If we put aside the ontological problem for a moment and we look at the practical use that Hull made of his analogy/identity, what conclusions can we draw, at the end of this chapter? Hull illustrates a very specific case study, that is, the change within the science of taxonomy in the last few decades arising from the disputes between pheneticists and cladists. Although we do not want to enter into the taxonomy debate, we see in Hull’s long and interesting account of the dispute a clear example of what we were fearing above: the reduction of the methodology of science to nothing but a sociological phenomenon and the negation of any special status to science and its products.32 That is, if we forget about the ontological standing of Hull’s account of the evolutionary analogy and focus on what problems it may solve, we actually find that it is rather a source of further problems than a solver. Ruse (1989) blames this on Hull’s area of expertise – taxonomy – in which the “way taxonomists classify is much more a construct of the taxonomist than a reflection of reality”. Although this analysis of the taxonomy debate takes a substantial portion of Hull’s book and, we believe, took a substantial portion of his time, it is also revealing that very few have commented on this specific investigation. In other words, the problemsolving power of his analogy has never been the main focus of commentators’ analysis, while most have gone on to comment on what we would call the ontological issues.

3.6 Conclusion “… Darwinism, if it exists, must surely be more than selection from an unimaginably huge pile of possibilities. As we argue here, Darwinism involves multiple cycles of selection; that is, interaction, mutation and replication” (Hull et al. 2001: 519).33 In this passage Hull et al. seem to hint at a departure from the usual confusion between the two terms ‘Darwinian selection’ and ‘selection’, a confusion which we believe we have partially resolved here. We have argued that a better way of testing the hypothesis that what appear to be two processes are in fact one and the same is not by the listing of analogies and disanalogies between them, but by the construction of a hierarchy of types which represents our present understanding of the world and, if possible, would link those processes. 32 33

See also the sections Toulmin and Campbell, in the previous chapter. Here ‘Darwinism’ means ‘evolution by Darwinian selection’.

92

Chapter Three

Only then is it possible to analyse the structure of the hierarchy and test the claim. Our analysis has shown that an exhaustive rendering of the common knowledge about processes involving selection, replication and evolution would not identify natural selection and scientific change – and would even distinguish processes that are commonly (and from our view uncritically) regarded as the same. Even a very simple hierarchy, which would collapse many of the details of our analysis into a more generic TH, would not be enough to identify the two processes. Here we have granted that there can be some sort of selective process in sociocultural change, but it could not be called Darwinian. If, after all, we decided that our TH is somehow defective, it can still be shown that Hull’s and other analogous views of Darwinian selection in scientific change would turn science into a purely sociological phenomenon. We believe Hull is right in pointing out that Darwinian selection may, after all, not exist, in the sense that it is too complex a phenomenon to be thought of as a simple, clear-cut law of nature. That is, Darwinian selection occupies a very low position in any meaningful type hierarchy and it is better thought of as a name for grouping together more basic phenomena, such as selection and evolution. So Darwinian selection should not be thought of as something that exists out there, as Hull et al. would seem to imply. Instead, given a plausible grouping of more general natural laws, Darwinian selection is only the label we attach to the node of the type hierarchy and, as such, it is so defined.

CHAPTER FOUR SCIENCE AS A SPECIAL CASE OF SOCIOCULTURAL ENTITY

In this final chapter the two remaining evolutionary processes investigated by Hull – replication and interaction – will be analysed. We shall argue that conceptual replication should be renamed, or at least rethought of, as conceptual ‘copying’. We shall also argue that lineages of concepts are critically different from biological lineages, mainly due to the referential value of concepts, as opposed to the functional value of biological entities. These differences will prove crucial for the different characterisation of the two processes, as biological evolution is undirected, unpredictable and progress-neutral. Scientific change, on the other hand, is convergent, partially predictable and globally progressive. In order to slightly expand the scope of our analysis, and to show routes of further research which could consolidate the basis of the claims of this book, we shall eventually argue that selective evolutionary processes better characterise types of sociocultural change other than scientific change. Furthermore, other largely unacceptable – for the biological realm – evolutionary models, such as the orthogenetic hypothesis, are more likely to describe a kind of change process closer to the one we believe is in place in science.

4.1 Direction and convergence in science Replication So far we have mainly focused on the analysis of the processes of variation and selection and we have left Hull’s definition of lineage almost untouched, along with the related notion of replication. Replication is one of the abstract processes we have included at the top of our TH (Figure 7), and its role is crucial in Hull’s determination of what counts as a lineage. Lineages are invested with a pivotal meaning in his analysis and in what follows we shall argue that the concept of lineage complicates further the

94

Chapter Four

investigation and full analysis requires singling out scientific change from the rest of processes that inherit from sociocultural selection in our TH. Hull’s main point is to treat “[scientific] concepts as historical entities in the same way as evolutionary biologists treat taxa and traits. We organize term-tokens into lineages, not into classes of similar term-types” (Hull 1988c: 16-17). Lineages are central in Hull’s organisation of termtokens, because in his view two term-tokens can count as the ‘same concept’ only if they are part of the same conceptual lineage.1 Just as the wings of insects and birds are to be considered as different structures, because their lineages have always been disjoint, term-tokens are part of historical entities – term-token lineages – and different lineages refer to different individual concepts. Species and concepts are historical entities, evolving through lineages as a result of selection (Hull 1988c: 17). Biological replication What counts as a lineage in biological evolution, according to Hull? “Lineages are formed by sequences of replicators” (Hull 1988c: 410),2 so genes, the chief replicators, and sometimes organisms form lineages while species are lineages.3 The matter, however, is complicated because the concept of ‘replicator’ is, on its own, not easy to characterise, and on page 424 of his book (Hull 1988c) we find the result of this confusion. On this page, Hull summarises his results from an argument about the complexity of biological evolution, which critics of the evolutionary analogy seem to forget – Hull says – when opposing the complexity of scientific change to the alleged simplicity of the biological one. Hull notes that critics of the efforts to see conceptual evolution as a natural selection process consider the theory of biological evolution as much more “neat and clean” than any available 1

This is an application of Mayr’s population thinking to the conceptual world. Hull notes that organisms rarely function as replicators (see below). So they should rarely form lineages and usually form lineages only indirectly, because they contain the sequences of actual replicators, the genetic material. 3 At least in a gradualistic version of evolutionary theory (Hull 1988c: 410). Hull’s concept of ‘species’ is a refinement of ideas first argued by Ghiselin (1966: 208-9). Mayr, who prefers the term ‘population’, argues that the species taxon is an individual, whose parts are the organisms, while the species category is a class, whose members are the species taxa (Mayr 1976: 192). Hull-Ghiselin’s view considers species not as kinds but as individuals, because they are entities that evolve, come into existence and go out of existence like organisms. Each species, as an evolving lineage, is a whole; its parts, the individual organisms, are related by the ancestor-descendant relationship and (among sexual forms) by gene exchange in reproduction. 2

Science as a Special Case of Sociocultural Entity

95

description of conceptual change. This, he objects, is not the case and the entities of biological evolution are as broadly and fuzzily defined as the entities of conceptual change. In particular, he reminds us that replication, interaction and evolution occur at a range of different levels in the living realm, and these levels are not sharply distinguishable. Replication, however, is more restricted than the others: “[r]eplication as a causal process is confined primarily to the genetic material” (Ibidem, p. 428) and special circumstances are needed for organisms to function as replicators.4 Populations and species cannot be replicators “very often or very well” while avatars are better equipped for that function. This shows that the theoretical apparatus behind biological evolution is not as clear-cut as most people think.5 Why is “replication […] concentrated at the lower levels”? One main problem for Hull is “the sense in which [alleged replicators] can be said to pass on [their] structure largely intact”. The main characteristics that Hull explicitly states for entities to be classified as replicators are the following: to exhibit structure to be able to pass on this structure largely intact to allow only minimal change before ceasing to exist6 By putting replication in the context of the remaining evolutionary processes, however, we believe it is possible to identify a further requirement, that is the autonomous potentiality to replicate. Lineages can be identified, in biological evolution, because replicators induce the generation of structurally similar copies of themselves. “In selection processes both descent and structural similarity are required” (Ibidem, p. 408). While structural similarity is covered by the three characteristics listed above, identity by descent is ultimately guaranteed by the replicators being responsible for their own replication, so that a causal chain can be identified to form the lineage.7 While almost anything can be copied over 4

This is the case, for instance, of unicellular organisms. Also individual cells within a multicellular organism can function as replicators. 5 A few lines below we find that asexual organisms form organism lineages and that in “some situations […] lineages are composed of sequences of […] populations”. 6 Each of these characteristics are open to several problems and would deserve further analysis. Here, however, we prefer to accept them as primitives and concentrate on what we think is the most relevant difficulty in the context in which we are arguing. 7 ‘Autonomous’ here is not meant in the sense of ‘self-sufficient’ or ‘selfcontained’. Any manual of molecular biology (such as Turner et al. 1997) reminds

96

Chapter Four

and over, without forming any lineage, only special entities can replicate, that is those entities which possess the ‘instructions’ for the process to be performed and these ‘instructions’, which are passed over to the resulting entities, assure a lineage can be identified. Conceptual replication What counts as a lineage/species in conceptual evolution, according to Hull? The choice here seems much more varied: research groups (Hull 1988c: 23-4), particular members of research groups (Ibidem, p. 243), research programs, theories and ideas (Ibidem, p.25), concepts and terms (Ibidem, p.142). Again, what is crucial in his account is the presence of a sequence of replicators, in this case what Hull calls contents: ideas, beliefs, data. Scientists, the primary interactors in conceptual development of science (Ibidem, p. 434), are the vehicles through which scientific contents indirectly interact with the world and with each other. Scientists primarily use their individual experiences and contents from other scientists to produce their explanations of natural phenomena and these explanations, in their turn, will be passed on to other scientists and continue the lineages of which they are part. Some difficulties can already be identified at this stage. While Hull mainly confines replication to the genetic material, conceptual change in science admits a vast range of replicators, as they are defined as anything that can be classified as ‘content’.8 Although this may be considered a mere accident, because what really counts is the process of replication itself and not the multiplicity of entities that implement it, the difference becomes more meaningful when we consider replicators only to be entities us that genetic (DNA) replication needs a whole range of cellular components to occur. However, the enzymes responsible for the actual replication (helicase, polymerase, ligase) have been produced starting from genes encoded in the DNA itself. Although there are no step-by-step, algorithmic ‘instructions’ for replication in the genetic material, actual instances of replication of genetic material occur because of its properties and the properties of the structures surrounding the material itself. And, again, it can be argued that the whole surrounding apparatus has been ultimately set up by the ‘instructions’ contained in the genetic material itself. Significantly, at the boundary between life and non-life viruses are usually placed, which do not have the capability to produce the whole machinery for their own replication and need to rely on host cells to replicate. 8 It could be argued that ‘content’ is a single category, no more problematic than the notion of ‘gene’ for instance. However, all life forms share the same genetic alphabet (the four nucleotides) and the same genetic code for amino acids (with minor exceptions). There is nothing as unifying as this for what counts as conceptual ‘content’.

Science as a Special Case of Sociocultural Entity

97

which have the potentiality to replicate. Hull admits that this potentiality is necessary,9 but he does not qualify it further. We believe this is a key property of replicators involved in Darwinian selection processes, that is the autonomous potentiality. As we have just noted above, the chief biological replicator is the genetic material because the ‘instructions’ for the replication are in the material itself. Replication processes at higher levels, i.e., at the level of organisms or populations, are possible, but only in special cases as the genetic material is ‘embedded’ in them. The same cannot be said of contents of books, brains or papers. Hull insists (Ibidem, p.438-9) that both genetic material of living beings and linguistic artefacts of scientists count as information.10 He takes a few pages to emphasise the distinction between information (which replicates) and what information is about (which does not replicate). For instance, a dinosaur footprint, when discovered by a scientist, can initiate a replication sequence of information about it, but it is not a replicator itself. A key distinction, however, is not mentioned, namely the distinction between replicating information and autoreplicating information. In our view, for an entity to count as a replicator it needs its replication process to be triggered by its own structure, otherwise the causal link for the formation of a lineage is missing. The issue is not the level at which replication occurs, but the capacity of the replicator to initiate its own replication. We have already pointed out that ‘autonomy’ here does not mean ‘self-contained capability’ because a variety of constraints, additional resources and conditions would be needed – and the higher the level the greater the variety which would be required – for the replication to occur. Within these constraints, however, genetic material instructs its own replication and, more indirectly, that of the higher level replicators in which it is contained. The nature of the entities we label as ‘scientific content’ is radically different in this respect. The replication process of the scientific content is not instructed or even described by the content itself. Popper (1994) suggested the metaphorical view according to which organisms embed ‘expectations’ about the environmental conditions in which they live. If we refine this view we can argue that ‘theories’ or ‘expectations’ about the world are actually embedded in the genetic material, in its structure, which codes for the organisms that will survive and replicate if those ‘expectations’ turn out to be acceptable. However, if we turn to what we regard as actual scientific theories, conceptual objects, 9

Although not sufficient. Replication must actually occur for replicators to be given this name (Hull 1988c: 412). 10 Information about something, that is. The problem of reference, however, will be addressed below.

98

Chapter Four

they lack the ‘instructions’ for their own replication. Scientific content gets copied from a computer file to another, gets replicated via discussions, via the production and the reading of papers, through manuals. In none of the replication processes does the scientific content ‘initiates’ its own replication or is involved in the production of the medium which will enact its replication, as in genetic material and organisms respectively.11 In the building of our TH in Fig. 7 we decided to follow Hull’s own principle and abstract from the entities involved in evolutionary processes and concentrate on the processes themselves, which are to be precisely characterised in order to establish whether different kinds of entities follow the same evolutionary regularities or not. From our distinction between the biological ‘replicate’ and the conceptual ‘being copied over’ we feel confident we can distinguish conceptual replication from biological replication. To generalise the process of conceptual replication, which might be renamed as ‘copying’,12 we can start by noting that the entities (contents) being copied over are more or less stably stored in all different kinds of media, from brains to magnetic disks to paper. The actual process of copying is initiated by some interactor who intentionally causes this content to be copied over from one medium to another, for example by reading, talking, transcribing and so on. There is no direct causal connection between the content itself and its copying process in the sense that the initiation and the modalities of the process do not originate from the content but from the interactor, the scientist who wishes to express, capture or distribute that content. We believe this is sufficient to claim it is appropriate to distinguish two processes of conceptual copying and biological replication and this is the reason why we represent the inheritance of sociocultural selection from replication as a dotted line in our TH. On the other hand, the process we label as Cyber Selection in our TH definitely inherits from the same replication process which Darwinian Selection inherits from. In this case, as in the biological one, the replication is initiated and largely governed by internal information coded in the entity which replicates. In the next section, after the introduction of some relevant points about biological and conceptual interaction, we will further develop this 11

Apart from trivial cases, such as the content of documents designing the construction of a photocopier, which get eventually copied by the resulting machine itself. This example, however, is still debatable because those documents are not properly part of the photocopier. Moreover, the machine can copy any document and might never get a chance to copy those containing its design. 12 A more appropriate term would be ‘re-production’, in the sense of ‘producing again’, if only it were not already taken by the biological terminology.

Science as a Special Case of Sociocultural Entity

99

argument and show that the differences just noted contribute to a related argument, namely that the two products of biological replication and conceptual copying, lineages, also show different properties.

Interaction Hull and conceptual interaction In Hull’s book (1988c) a whole chapter is dedicated to conceptual interaction, a key issue in the general question of conceptual evolution. Individual scientists, and/or research groups are the main conceptual interactors in Hull’s view. Most of his account of scientists, mainly those operating in the field of systematics, revolves around the following pivotal points: society influences scientists’ activity; there is no sharp distinction between theoretical and observation terms; the only identifiable continuity is in conceptual replication. Although Hull stresses that social forces have always had a strong influence on the scientific research of individuals or groups, from funding organisations to wars to political regimes, he also recognises that this influence is far from clearly obeying general laws. Scientists can have the same background of poverty or richness or political persecution and still hold opposite views; social factors that we might think strikingly diverse can give rise to scientists with similar views. Hull concludes that we have yet to find “significant correlations between kinds of social forces and kinds of substantive scientific belief” (Ibidem, p.480) in order to give some ground to externalist claims. However, he notes the same arguments hold for biological evolution, if we take into account that the same environment allows for a huge variety of adaptations for organisms. Thus if we accept adaptationist hypotheses and explanations in biology, we should also do so in science. From this argument we isolate a specific point, namely that in terms of their unpredictability, conceptual and biological change are of the same sort: it would be impossible to predict the future development of any lineage by observing the environment it is in. Later on, while discussing the role of observation and theory in science, Hull notes the difficulty in making the classic distinction between theoretical and observation terms. This difficulty mainly arises because, if we define observation terms as those denoting entities to which we have direct access,13 it happens that our sense organs display a certain degree of

13

Such as things or events we can perceive (the Moon, a dog, a whistle) as opposed to “those descriptive terms that appear in the most fundamental

100

Chapter Four

variability between different individuals or even at different stages of individual development, such as the acoustic acuity which varies with our age. However, Hull finds this contingent variability, as well as the contingent status of our sense organs due to evolutionary factors, irrelevant: in spite of this variety “scientists do seem to be able [to] converge on remarkably similar conceptions of natural phenomena. Phenomenologically, red light and infrared light are quite different. Most people can see with one but not the other. Scientifically this difference is of no consequence. We know as much about one sort of light as the other”. In order to reach the same conclusions, scientists “need not have precisely the organs that they happen to have” (Ibidem, p. 487). For our purposes, all these remarks are quite important, because they show a crucial difficulty which is related to the previous paragraph. On the one hand, the course of scientific change is unpredictable on the basis of environmental (social) factors, as its biological counterpart. On the other, scientists’ views eventually converge towards a broad consensus. These two claims are contradictory. Some objections may be raised to our argument here. Firstly, being able to predict the eventual convergence of lineages of theories and hypotheses may be considered a weak form of predictability, because we are unable to predict the actual course the lineages will go through. However, convergence is an indication that conceptual interaction cannot be the same process as biological interaction, because its coupling with conceptual copying produces lineages behaving differently from the biological counterparts. Moreover, in some cases a stronger form of conceptual predictability can also be expected. Imagine different research groups working on partially overlapping areas of research, and holding irreconcilable views on the overlaps. If one of these groups produces very convincing results in these common areas, we will be able to predict the occurrence of eventual rerouting of the other groups’ views and lines of research in order to accommodate those results and, in some cases, we might also be able to guess in what direction the rerouting will occur. By contrast, an adaptive advantage acquired by a lineage of organisms in an environment will not necessarily be acquired by the others. It is still possible to object, however, that convergence is also observed in biological evolution, so this might not be a difficulty after all. For example it is possible to mention the striking number of homoplasies between fishes and mammals that happen to share the same watery propositions of a theory, such […] as ‘mass,’ ‘acceleration,’ and ‘force.’” (Hull 1988c: 486).

Science as a Special Case of Sociocultural Entity

101

environment.14 Thus, it may seem perfectly acceptable to expect unpredictable but convergent evolution for both scientific and biological change. We do not believe we can treat the two convergences as the same, though, and a number of examples will show that the two processes cannot be the instantiation of the same general process type. As in the case of insects’ and birds’ wings, which will be discussed more fully below, convergent evolution in biology is the result of uniformity of constraints on structures which happen to perform the same function.15 We can expect organisms of similar size in the sea to have developed similar structures to perform the same functions, such as fins to swim around, and the same applies to insects’ and birds’ wings. However, fins are not the only structure providing motion for sea creatures: jellyfish, shellfish and squids all happened to get very different locomotion organs in the lottery of evolution, because life functions do not imply unique solutions. Additionally, ‘convergence’ is a misleading term to describe the evolution of functionally similar structures into similar anatomies, because the respective lineages do not join into a single one. ‘Convergence’, in this case, is about structure, not lineage. The same cannot be said of conceptual evolution. The same phenomenon can be given different explanations but, together with Hull, we are not surprised at scientists’ tendency to converge towards a single explanation, which will represent the merging of the previous lineages.16 Structural convergence in biological evolution is observed when structures eventually happen to perform similar functionalities, while conceptual convergence in scientific change meets an already existing expectation of scientists, on the basis of the preexistence of natural regularities that scientific hypotheses try to explain. Another example will show in more detail why the interaction processes in biology and in science cannot be equated. If we stick to Hull’s example of wings, we have to accept the possibility that different 14

Homoplasy is very common in biological evolution, involving even very complex structures such as the vertebrate and cephalopod eyes (Futuyma 1998: 110-111). 15 More exactly, when “a similar character (or character state) in two organisms has not been derived from a corresponding character (or state) in their most recent common ancestor” (Futuyma 1998: 111) we have homoplasy, and “[h]omoplasious features are often (but not always) adaptations by different lineages to similar environmental conditions” (Ibidem, p. 111). Convergent evolution is a special case of homoplasy, occurring when “independently evolved features are superficially similar” (Ibidem, p. 110). 16 Of course this convergence is not towards explanations which correspond in a definitive way to the corresponding phenomena. Explanations will change and multiply again, until the next stage of agreement is reached.

102

Chapter Four

wings can coexist forever. Birds and insects exhibit different ways to perform the function of flying, as do different species within their respective groups, but there is nothing like ‘the best wing’, neither we can expect that, eventually, one of the wing structures will replace all the others in the interaction game. Biological functions can be performed in totally different ways and we cannot really define a meaningfully definitive way to compare them, in an evolutionary context. Some wings can be more efficient in the ratio velocity/power produced by the organism or in the ratio distance covered/energy consumed and so on. As long as a species carries them, however, any pair of wings will do. Even in a prey/predator situation, extinction is one of the possible outcomes but coexistence is the norm. In most ecosystems we know of, different species continue to survive because of the presence of the other species, as required by the numerous interactions occurring between them. In most cases, for example, predators cannot survive if all their prey are eventually captured, and in the absence of predators, some species would eventually grow too big and die from starvation, because they would consume all the resources they need at too fast a rate. Other examples of complex interactions include parasitism, symbiosis and other less clear forms of interdependence between different species. The point to make here is that interaction in the life realm brings equilibrium. In other words, there is no direction of development of any species towards any specific target or along any specific direction, but the interaction and selection processes bring the lineages in the ecosystems along a dynamic equilibrium.17 What happens in the case of scientific concepts and theories? Rival theories do not carry on side by side for ever. Although several comparison criteria exist and they can favour different rival theories, science – as we know it – will eventually get rid of all alternatives but one, because different theories are still supposed to describe the same phenomenon and this abundance cannot be sustained for long. How long, of course, is a relative quantity but the nature of scientific research makes it a finite one. One set of explanations, for any given domain of reality, will eventually replace all the others, if only to be replaced in its turn at some later time. All of this is the result, in our view, of the referential role of scientific content,18 which will be treated soon. Here it is enough to note that 17

This has already been pointed out when the Red Queen Hypothesis was mentioned, in Chapter 2. 18 It refers to the domain of reality it is supposed to explain, that is. We shall regard this as intuitively clear: in scientific theories and ideas proper nouns refer to individual objects (such as ‘Venus’), the description of laws of nature refer to natural regularities and so on.

Science as a Special Case of Sociocultural Entity

103

scientists believe that when there are different theories which explain roughly the same group of phenomena, these theories compete with each other for the position of ‘best’ explanation, which will be won according to certain criteria. These criteria may be historical, not well defined or even contradictory, but the basic assumption of the need for unitary explanations always accompanies scientists’ activities, whatever their ‘social bias’ might be. Scientists work with the aim of sorting all possible conceptual lineages relevant to an area of interest, the result being some lineages continue and the rest are terminated. In nature there is no such aim and competition of biological entities is different. Such myths as the ‘struggle for survival’ or ‘survival of the fittest’ only persist in popular accounts of evolutionism. Homoplasious wings only ‘happen’ to be similar because of constraints derived by their function. Or, better still we happen to call ‘wings’ all organs that we see animals use for flying, but there was no aim to fly in the first place, only the casual emergence of structures apt for flying.19

Identity by descent Related to both the problems of the autonomy of replication and of the referential role of conceptual lineages is the issue of identity of scientific content by descent. As we have already pointed out, Hull stresses that ideas in selection processes are individuated in terms of descent (Hull 1988c: 434). Science, as a social process, is made up of historical lineages of concepts (Ibidem, p.448) and conceptual lineages can be disjoint even though the replicated concept is formally the same. Even the same discovery, made independently by two different scientists, really counts as being “the same” only if those scientists belong to the same group or the lineages of relevant ideas and theories they hold have common ancestors. In our view, it is correct that distinct, independent discoveries of natural regularities, such as a year ~365 ¼ days long, objects, such as a certain mountain, or abstract entities, such as the number ‘2’, count as different discoveries from a historical point of view. If we confine ourselves to the historical level, however, we forget about the additional level that discoveries possess, that is the referential level. Discoveries, like concepts, refer to something.20 Hull’s preferred example, illustrated above, 19

It may also be noted that Hull’s example of wings is particularly ad hoc, because other structures that we regard as performing the some function, such as breathing or digesting, are much more diverse across different lineages. 20 At least the realist view we are holding. Discoveries, of course, refer in a different way than concepts but we regard this as intuitively clear as before: the

104

Chapter Four

which he uses to explain the identity by descent in nature, is that of wings. We have shown that there is a relevant difference between lineages of organs or organisms and lineages of concepts. Organs perform functional roles, concepts perform referential roles. If in a lineage of organisms a certain structure evolves into a flapping structure that allows the organisms to fly, then we say that the function of this structure is to make flying possible regardless of its origin, and – in English – we give it the broad name ‘wing’, because the name is associated with the function rather than the particular structure.21 However, from an evolutionary point of view, it is correct to see eagles’ and robins’ wings as the same structure which, at a certain point in time, evolved into two slightly different sets of organs. It is also correct to distinguish evolutionistically these structures from insects’ wings, whose origin is totally distinguished from birds’ wings and whose only feature in common with them is the function they perform (and some secondary features given by the constraints associated with that function, such as shape or mobility), but the case for lineages of concepts is again different. Let’s restrict ourselves to just one type of discovery and think of two disjoint lineages of ideas or theories, say in two reciprocally isolated ancient civilisations, both of which eventually culminate in the conclusion amongst the observers that the Sun and the stars22 return to the same relative positions after about 365 ¼ days. And as a result they accordingly compile their calendars. The two lineages of concepts are historically distinct, of course, but they refer to the same presupposed regularity of nature, namely what we would call now the ‘mean year’. Even if we looked back at the occurrence of these discoveries from a historical perspective, we would certainly be looking at two particulars or tokens, as Hull would call them, however we would still need to treat them as particulars which refer to the same phenomenon. This need arises by the nature of the lineages in question, conceptual lineages, which in our view can be so called exactly because they show a referential property. At the root of the problem is what we think is an ambiguous use of the notion of ‘sameness’ by Hull. When wings are treated as evolutionary homoplasies, he says, it is correct to consider the wings of birds and insects as instantiating the same structure. However, as discovery of an island refers to that island, the discovery of a natural law to that natural regularity and so on. 21 We are not trying to endorse any sort of nominalism here. Our focus is on functions, not on names. 22 Apart from terminological issues of course, the definitions of ‘Sun’ or ‘stars’ are taken as commonly understood.

Science as a Special Case of Sociocultural Entity

105

they do not derive from a common ancestor, they are not homologous and an analogous treatment, in his view, should be applied to concepts. What he means here is that individual wing-tokens of birds are parts of a higher level individual, the lineage of bird wings, which is distinct from the lineage of insect wings. The comparison, then, is not between individual wing-tokens. To this we object the following: if we consider independent discoveries of Mendel’s laws, for instance, when we say that they are the same discovery we actually mean they are discoveries of the same regularity in the world. We can appreciate that discoveries and concepts are historical entities, but this does not entail that we should forget that we think of them as referring to something that is not historical, such as regularities in nature. On page 17 Hull writes: A consistent application of what Mayr (1963, 1969) has termed "population thinking" requires that species be treated as lineages, spatiotemporally localized particulars, individuals. Hence, if conceptual change is to be viewed from an evolutionary perspective, concepts must be treated in the same way. In order to count as the "same concept," two termtokens must be part of the same conceptual lineage. Population thinking must be applied to thinking itself. This change in perspective is radical, so radical that some readers are liable to dismiss it out of hand. Terms are not important. Concepts are. Some scientists can use the same term and mean different things, while other scientists can use different terms and mean the same thing. The gene concept is the gene concept, regardless of space and time, regardless of conceptual replication sequences. All those who have the same views about genes are using the gene concept in the same way, accidents of terminology and history to one side. As a result, several scientists can make the "same" discovery independently. Just as wings have evolved several times independently, Mendel's laws were discovered independently by several workers. If "Mendel's laws" or "the gene concept" are treated as conceptual evolutionary homologies, then independent discoveries are ruled out by definition. If the discoveries are genuinely independent, then they are not the "same" discoveries.

Hull is talking about ‘terms’ being used by scientists to mean ‘concepts’, and we can agree that these are historical entities which change and are transmitted. He focuses on the opposition between ‘term’ and corresponding ‘concept’, while we believe the relevant distinction is between ‘concept’ and ‘what the concept refers to’, and the latter is not historical in the sense that we think of natural regularities as independent of human activity, whatever accessibility we want to grant to them. One possible objection is Hull’s claim that “if an evolutionary account of conceptual change is going to have any chance of succeeding, the basic

106

Chapter Four

units of both evolutionary biology and conceptual evolution must be viewed as the same sort of thing – either as spatiotemporally unrestricted classes or as spatiotemporally connected lineages”. However, we take it as an admission of failure if this uniform view is not possible. It could also be argued that, after all, concepts are what are evolving anyway and they should be given an evolutionary account as species are. However, in our approach, which is an extension of Hull’s approach after all, what is important is to analyse the individual nodes of the TH and corroborate or dismiss the claim that Natural Selection and Sociocultural Selection in science can inherit from the same nodes. Lineages In the last section of his chapter on conceptual interaction, Hull briefly deals with the problem of reference by using Kitcher’s work on the reference potential of term-types (Kitcher 1982). The main distinction noted there, for our aim, is between lineages of term-tokens referring to historical entities, such as a specific person or a particular species, and those referring to non-historical entities, such as gold or water.23 In the first case for instance, two lineages of tokens referring to the same entity can be shown to converge, by historical investigation, at some point in space, although at two different times in the past. For example, the discovery of an island by an expedition can be shown to be about the same island found at a different time by a different expedition by reconstructing the routes of the two crews or reading the descriptions of the island’s features they might have recorded. This would show that the same spatiotemporal entity, the same island, originated two distinct lineages of terms used to designate it. In that case the convergence proves that termtokens in the two lineages are actually coreferential, and some clarification can be put in place, if needed, for example eliminating one of the currently used alternative terms or recording the coreference for posterity. In the case of non-historical entities, lineages of term-tokens referring to them need not converge at any time in the past. For instance, lineages of termtokens referring to the chemical element gold need not have a common ancestor which was applied to a certain lump of gold. However, other sorts of investigation, such as chemistry experiments, can still reveal the coreference and allow for clarification (Hull 1988c: 500).

23

We acknowledge this might be seen as not too sharp a distinction. For instance, it could be objected that gold and water have not existed for ever. The distinction, however, is fairly intuitive and further analysis of it is not our aim here.

Science as a Special Case of Sociocultural Entity

107

We believe one of the main limitations of this account is its focus on scientific terms denoting substances and on definitions denoting historical objects, sidelining theories and explanations. This downplaying of the role of more complex entities in Hull’s analysis of lineages is also mirrored in his use of the idea of ‘convergence’ only in a backward modality: conceptual lineages can be converging back, in the past, to a common ancestor, to a common token in the sequence of terms. Only in very special cases, for Hull, is it possible to decide otherwise on the coreference of lineages, such as in independent lineages of terms referring to a chemical element. To all this it can be objected that scientists are not interested nor are they driven by this search for backward convergence.24 Historians will normally be, but only in the reconstruction of theory change partially detached from the causal processes that are responsible for the change itself. If we are to characterise scientific change we have to account for what scientists are aiming towards, which is onward convergence. Scientists’ utterances, conceptual tokens, and words written on a piece of paper by the scientists are definitely historical entities, but what the scientists think they are describing are universal laws and this affects the changes they are going to apply to these conceptual lineages. Hull sees the question in different terms, as he states “[s]cientists intend to transmit term-types, but all they actually transmit are term-tokens, which are immediately interpreted as types” (Hull 1988c: 506). Biological selection occurs at the level of tokens, not types, says Hull, and thus the same should be expected for conceptual selection. This is a key point. Undeniably, by writing down a word on a piece of paper a scientist has represented a term-token from her mind. When another scientist eventually reads that word another term-token occurs in her mind and the lineage continues through her if the term-token continues to get copied. What is missing from this account is that both the intention of the scientist to convey information about what she thinks is a natural type – the one she intended by what she saw as a term-type – and the interpretation of the received information as a type are driving the transmission and the interaction. This intentional, goal-directed layer of conceptual interaction, we claim, accounts for convergence and directionality of scientific change. Discourses about the provenance of the tokens and about the genealogic sameness are only marginal in the explanation of scientific change, because scientific interaction is not 24

Unless this convergence is, on its own, the object of their research, as in the case of evolutionary research. Which is, of course, Hull’s preferred example of research.

108

Chapter Four

driven by what these tokens actually are but by what scientists think they refer to. To summarise, it is possible to identify two types of lineages in Hull’s view: lineages of entities which do not have a reference (e.g. species) and lineages of entities which are supposed to refer to something (e.g. concepts). The latter can refer to historical entities, such as in the case of the lineage of term-tokens ‘tiger’, or to entities spatio-temporally unrestricted, such as the lineage of term-tokens ‘gold’. To the latter type we add a related one: lineages of more complex entities, such as theories and explanations. It is in these more comprehensive entities, as opposed to simple ‘terms’, that the conceptual convergence manifested in scientific interaction – which is what we are interested in – is more clearly in action. For instance, if we follow Kitcher’s example of reference change (Ibidem, p. 500) proposed by Hull, the term ‘planet’ has been under debate until recently (IAU 2006). By contrast, the orbital regularities of planets have caused theories and explanations to eventually converge much earlier to common descriptions expressed by Kepler or Newton laws. Terms like ‘mass’ and ‘gravitational field’ may have changed their reference, however Einstein’s theory has become the latest point of convergence in the field of celestial mechanics.25 These convergences are dictated, if we keep a realist point of view, by existing regularities which, however inaccessible, are able to constrain scientific change in a way which is not occurring in biological change, that is in a directional way.

Convergent realism Two points need to be addressed to support our claim. Firstly, it must be shown that (forward) convergence and directionality can really be found in science. Secondly, that these phenomena arise because tokens in scientific conceptual lineages refer to regularities in the real world. We will avail, again, of some suggestions from Aronson et al. (1994), namely of their argument in support of convergent realism. The argument for convergent realism, the main point of the whole book and culminating in its last chapter, can be summarised as follows. Firstly, we have to reverse the traditional definition of verisimilitude in terms of truth and “characterise truth as a limiting case of verisimilitude” (Aronson et al. 1994: 123). Verisimilitude, in turn, is defined in terms of similarity of types described by the model of a theory and the actual type 25

In other words, the scientists’ representations of the natural entities and laws thought to be involved in celestial mechanics are now largely homogeneous and correspond to the representation given by Einstein.

Science as a Special Case of Sociocultural Entity

109

in the reality: “verisimilitude is [not] a matter of comparing different sets of propositions […] but weighted similarity comparison between a model system and the real system it intends to capture” (Ibidem, p. 122). Secondly, we define the plausibility of a theory in terms of two criteria: a logical one, judged by its capacity to produce correct predictions and retrodictions (empirical adequacy); and an ontological one, which is its being modelled on a chunk of the type-hierarchy of the world which the scientific community accepts as the common ontology (ontological adequacy). We think of the content of a theory as the model, or models, embedded in a TH. Different theories then can be compared regarding their ontological adequacy by estimating to what degree the TH in which they are embedded is similar to the commonly accepted TH, which we believe is descriptive of the states of affairs. The key point then is the authors’ argument that increasing plausibility is an inductive ground for increasing verisimilitude, i.e. progress. (Ibidem, pp. 191-2) The claim revolves around the inductive assumption that a functional relationship exists between verisimilitude and plausibility. In particular “that as our theories are getting closer to the truth we are reducing the error of our predictions and measurements and vice versa”. This is based on the principle of epistemic invariance, exemplified by Fig. 9: When it comes to gathering evidence for our beliefs, the epistemological situation remains the same for observables and unobservables alike, no matter whether we are dealing with observables, possible observables or unobservables. (Ibidem, p. 194, emphasis in the original)

On the right hand side, the figure graphically represents an historical finding about science, the basis of the inductive claim: increased plausibility, that is better pre/retrodictions provided by a theory within an accepted ontology, results a posteriori in increased verisimilitude, that is better resemblance of the model provided by the theory with reality. In other words, each time hypotheses are formulated on the basis of plausibility about unexamined aspects of the world, when examination becomes possible – the ‘veil of perception’ moves to the left – the assumed model turns out to match what reality proves to be like. An example is provided by Boyle’s argument in favour of the reality of corpuscles on the basis that chemistry, at the time, was intelligible only by assuming that scientists were manipulating unobservable corpuscles in

110

Chapter Four

Figure 9 – A graphical representation of convergent realism26

matter. Eventually, tunnel-effect microscopes and X-ray diffraction made ‘observable’ what had been only manipulable (Ibidem, p.199). On the left hand side is the result of the inductive claim: there is no reason why we should think this relationship between plausibility and verisimilitude does not hold for presently unexamined aspects of the world, even for those which are so in principle. We believe the general principle of convergent realism, adequately amended, is sustainable.27 As a consequence, we claim that convergence and directionality can be found in science. If this is accepted, it must also be accepted that the ultimate reason for the existence of this convergence is the existence of natural regularities which scientists, by constantly modifying the conceptual lineages they work with, try to model. It is also worth stressing that, for our aim, no hypothesis is needed on the absolute ‘values’ of the regularities which cause convergence (they are inaccessible anyway), as opposed to the weaker assumption that regularities do exist 26

Adapted from Aronson et al. (1994: 195). Amendments or refinements are needed, for instance, to take into account that some revolutionary scientific changes cannot be directly represented in the diagram, which should allow for some sort of ‘discontinuity’ in the line representing the state of science. This and other criticisms, however, are not our aim here because they do not seem to affect our main argument.

27

Science as a Special Case of Sociocultural Entity

111

and have effects on our theories and on scientists’ way of working (interacting) with them and one another. Finally, even a much weaker assumption will do. We can decide that convergent realism is false, that the models of the world that our theories provide are not getting closer to any ultimate representation of the true structure of the world in any meaningful sense. We can still claim, though, that scientists behave as if that was actually the case, and pre/retrodictions are really getting more accurate, and scientists’ models are converging towards more inclusive explanations on which the scientific community agrees. Thus, without the stronger assumptions of convergent realism it is still possible to claim that conceptual change is driven by the same forces that would be in place if convergent realism were true. *** A final remark will introduce the topic of the following sections. Because plausibility and verisimilitude, as defined here, are only found in science as we know it, change described by convergent realism is characteristic of science only. No other social processes can be described, in terms of the change they produce, the way science is described. As a result a new TH would be needed to represent the node of scientific change as distinct from other sorts of social change. Furthermore, two questions arise: do we have a process model that would be closer, when represented in the TH, to scientific change? If scientific change cannot be thought of as a case of natural change, do other social processes exist which might be?

4.2 Natural Selection and Sociocultural Selection Science as a special case of sociocultural entity Hull claims that scientific change is just a case of sociocultural change and in his analysis he aims to show that, as with the rest of the social processes of change, it is of the same type as biological change. In the previous section we acknowledged that although scientists are not exactly the truth-seekers that popular science books depict, they are largely committed to the search for unitary explanations. It is true that science is a social entity, and scientific change is a social process, but we have argued that the directionality of the scientists’ activity and the referential value of scientific content makes it a social process of a particular sort, in the sense that it is of a different type from biological change when analysed by

112

Chapter Four

means of a TH approach. Is it possible, though, that Hull’s general analysis of selection processes is more applicable to the domain of other sorts of entities which undergo changes under the pressure of social interaction, for which reference or directionality play a minor role or no role at all? We will briefly attempt to show here that this may not be the case. In particular – although this is not the place for a full analysis – we find that the matter is very likely to be more complicated than Hull’s view, and that the processes explaining the change over time of different social entities involved in selection-like processes still end up slotted into nodes of our TH differently from the node occupied by Darwinian Selection. Candidates for such an attempt may be found in domains where the required evolutionary entities – replicators, interactors, lineages – and associated processes can be identified. As a concrete example we can think of hats (i.e. objects which have the general use to cover part of persons’ heads). In a naïve approach to analogy, the design and ideas producing a specific model of hat can be thought of as analogous to the genetic information which ‘produces’ a specific species of organisms. From this point of view, it is possible to find selective processes at several levels, as Hull points out for any selective process. For example at the level of individual hats, when a customer browses in a shop and eventually chooses one; or at the level of design proposal, when the marketing department decides to send some designs to production and not others. These levels of selection would be analogous, respectively, to selection at the level of individuals and species in biological evolution. However, what happens when we try a deeper approach, more consistent with the one we have proposed in this volume, and search for a location for these processes in our TH? How should we arrange the various links, particularly in relation to Science selection? Ordinary hats28 are valued differently in different places and times and the corresponding design ideas are affected accordingly. The value that hat-users give to hats is different from the value that scientists would give to theories and concepts, regardless of their justification or belief for doing so. Instead of truth, verisimilitude or predictive power, hats are valued in terms of fashionableness, warmth, colour, price and so on, most of which are entirely subjective and impossible to decide for very large sets of

28

One you would normally buy from a shop because you like it, as opposed to hats which are worn for special purposes, such as safety helmets, sport or military caps.

Science as a Special Case of Sociocultural Entity

113

users.29 Ordinary hats are used just for pleasure and none have any referential value.30 As a result, the process of hat-selection inherits, in our view, directly from Selection in our TH, and not from Directed selection. This brings it closer to Darwinian selection in the metric we specified in the previous chapter. Similarly, the process inherits directly from Evolution (variation), if we think of the changes introduced in successive designs for a certain model as pseudo-random and not directed towards a specific target design.31 As a result, some of the arguments made for scientific change do not apply any more and convergence or directionality cannot be found in the evolution of hats. In the case of ordinary hats, however, it is not clear how to identify lineages in a way which is consistent with the concept of lineage which has been used here. Certainly, at the level of individual hats, all the items of a certain model have the same structure but nobody could argue that this structure ‘is passed’ from one item to another. At the level of the design we might regard the specification of a certain model as an entity that is transmitted from one designer or computer to another. However, a designer could come up with a totally new design every year, and it would be difficult to identify any continuity over different models. In this case it is the crucial entity behind the concept of lineage – replicator – that is difficult to find: there is nothing which is passing its structure largely intact over different generations of hats. Thus the inheritance from Replication is removed from the diagram.32 However, if we focus on hats which are chosen for more intersubjective reasons we can find a sort of continuity in their design. For instance, let us think of hats – helmets – used on construction sites. Around the world different companies produce different models and, over time, they tend to modify those models with continuity. Their aim is to 29

From the list we suggested, ‘warmth’ may be seen as the least subjective criterion. However, some people prefer hats which keep warm and some value hats which do not. 30 An argument could be made regarding unordinary hats, for instance on those which are part of uniforms, but we will not embark on such an argument here. 31 It could be objected that the variations are actually introduced with a goal in mind, that of selling more items. However, there is no ideal hat which would sell the best and designers can only perform ‘blind variations’ and hope for the market to be favourable. 32 This may seem too strong an objection. It might be argued, for instance, that it does not apply to models of hats which are designed as slight modification of previous ones. In that case, however, a weaker objection can be raised which is illustrated in the following paragraph.

114

Chapter Four

improve the functional characteristics of the products, which is usually achieved by progressively introducing modifications into their design. This is efficient and cost-effective, as compared to a strategy of complete ex novo redesign. Additionally, if we focus on the functional value of those hats, we find that some of the arguments we raised to object to Hull’s analysis cannot be sustained any more. In particular, those related to the referential content of scientific concepts as opposed to the functional value of organic evolutionary objects. Safety helmets need to be highly visible to human beings, which constrains the colours that can be successfully used due to human physiology of vision and the main range of wavelengths of sunlight and of artificial lighting of the sites where they are used. Analogous constraints derive from other functional requirements, such as degrees of flexibility and elasticity for impact absorption, reasonable weight for comfort and so on. Moreover, in contrast with scientific theories, different lineages of helmet designs can coexist indefinitely: ‘the best’ safety helmet on the market is a much more subjective notion than 'the best' scientific theory for a certain domain of phenomena. However, we believe we are in the same situation as scientific concepts in terms of the distinction between ‘replication’ and ‘copying’. If we think of the designs of helmets as objects which get copied over, largely intact, from one (teaching) designer to another (learning) or from one design to a following, slightly enhanced one, then we can argue we have found replicators and associated lineages. Nonetheless, as for scientific concepts, it could be objected that 'being copied over' is not the same as 'replicate' and we will leave the inheritance from Replication represented as a dotted line. When compared to ordinary hats in our TH their changing process still inherits directly from Selection, because there is no static, ideal 'helmet' towards which all designs are directed. Over time, they are selected on the basis of their degree of safety, for example on the basis of laboratory tests or actual accidents that occurred on sites. There is no implicit directionality in this selective process. On the other hand, direct inheritance from Evolution (variation) is no longer sustainable. The variations introduced in the helmets' design are directed towards what are thought to be good features which the products should have. If one of the desirable features is the visibility of the helmet, changes in the colour and reflective properties of the material used, or the varnish to cover it, will be intentionally directed towards that goal.

Science as a Special Case of Sociocultural Entity

115

Figure 10 – A type hierarchy for hat selection

4.3 Orthogenesis and scientific change We shall conclude by exploring a possible answer to a question symmetrical to that of the previous section. There the question was, “Is there another sociocultural process of change, different from scientific change, which is of the same type as biological change?” Here the question will be, “Is there another type of evolutionary change which is the same as scientific change?” In other words, is it possible to find a theory of organic evolution different from Darwinism which describes the same process occurring in science? We shall argue that the hypothesis of orthogenesis might provide an answer to the question.

The Orthogenetic Hypothesis In the nineteenth century, the orthogenetic hypothesis of biological evolution gave rise to a cluster of theories which shared the assumption that evolution is driven by directed variation:33 orthogenesis could be characterised as non-adaptive, directed evolution.34 This cluster of theories

33

Futuyma (1998: 24) claims that orthogenetic theories “held that the variation that arises is directed toward fixed goals, and that a species evolves in a predetermined direction without the aid of selection”. ‘Goal’ and ‘predetermination’, however, are rejected by most of the main proponents of orthogenetic theories (see below). 34 Bowler (1983, 1984) uses the term ‘linear’ as well, but we agree with Grehan and Ainsworth (1985) who see in the Greek word ‘orthos’, in the context of the

116

Chapter Four

was the principal challenge to Darwinism in the years following the publication of The Origin of Species, as they set aside or played down the role of natural selection and incorporated a notion of progress. A strict orthogenetic view was not held for long by many writers, though, and a precise mechanism for orthogenesis was never put forward. This cluster of theories ceased to exist when the phenomena that orthogenesis attempted to explain were shown to be better accounted for by employing natural selection, Mendelian genetics and a better reading of the fossil records (Futuyma 1998).35 We will summarise here the features that most authors associate with orthogenetic development. Wilhelm Haacke, Theodor Eimer and Thomas Henry Huxley are among the most prominent scholars who embraced orthogenesis.36 Haacke37 points out the existence of a trend in evolution to progress from lower forms to higher forms of complexity in nature. Organisms’ variations follow a ‘predestined’ direction: evolution can proceed in one direction only, permanent reversions in the trend, such as mammals reverting back to ancestral amphibians, are not possible. When reversions occur they are only temporary because, according to Haacke, the overall trend still persists: An example of a well-known evolutionary trend […] is the reduction of toes during evolution of the horse. Haacke’s view was that, although there may be the occasional throwback in present day horses (i.e., doubling the number of toes), it was out of the question that the modern horse will become five-toed again. (Grehan and Ainsworth 1985: 181)

After Haacke, Eimer popularised the term ‘orthogenesis’ at the end of the nineteenth century and it soon became prevalent among palaeontologists to account for phenomena which could not be fully explained by Darwinism (Bowler 1979: 40). Eimer considered orthogenesis as a universal law of variation in a limited number of definite directions actual orthogenetic theories, the meaning of ‘direct’ (Bowler 1983: 119) and (Bowler 1984: 233). 35 Again, Futuyma does not seem too precise in these remarks: more sophisticated (and diluted) orthogenetic views are still supported by some researchers (Grehan and Ainsworth 1985), although they do not appear to have many followers. 36 The origin of the term ‘orthogenesis’ is debatable. According to some biologists its origin should be attributed to Theodore Eimer. This opinion is held by Bateson (1910), Lang (1923) and Gould (1977). Other scholars, among them Rensch (1966) and Mayr (1982), attributed the source to Haacke. 37 Haacke, W (1893) Gestaltung and Vererbung, O.W. Nachforger, Leipzig quoted in Grehan and Ainsworth (1985).

Science as a Special Case of Sociocultural Entity

117

(Ibidem, pp.50-51). He argued that factors both external and internal to the organisms were important to orthogenetic development. He claimed that variations are not random. They are directed by forces regulated by the internal constitution of the organism, so they tend to occur more easily in some directions than in others and are essentially the organisms’ response to external stimuli from the environment without being an adaptation to it. For instance, Eimer states that when a whole species experiences modified environmental conditions, all the organisms would respond in the same way because they have the same constitution.38 The causes of definitely directed evolution are contained, according to my view, in the effects produced by outward circumstances and influences such as climate and nutrition upon the constitution of a given organism. […] In my view development can take place in only few directions because the constitution, the material composition of the body, necessarily determines such directions and prevents indiscriminate modification. But through the agency of outward influence the constitution must gradually get changed. The organisms will thus acquire more and more in a manner harmonising with their specific individuality – and so new 39 directions of development will be produced.

However, a new form in the original population might appear in response to the environmental changes: “[t]he species will divide itself by ‘genepistasis’–the tendency of the more sensitive individuals to progress to the next stage of variation ahead of their fellows.” (Bowler 1979: 51). This mechanism, in Eimer’s view, shows the superiority of orthogenesis over Darwinism because genepistasis can explain speciation without isolation. Thomas Henry Huxley (1978) also maintained that the direction of variations is constrained by constitutional factors in the organisms. The environmental variables might elicit change but it is the internal constitution of the species that predetermines the pattern of development. And all the species with the same predisposition will change in the same way to the same stimulus. Again, although the environment can influence modifications in a species, it is the internal constitution of the species that leads the pattern of its development. Thus, in both these authors’ views the external factors act as stimuli, the environment acts on the existing constitution of the organism and the 38

Or, in other words, they obey the same ‘laws of growth’. Eimer G. H. T. (1898) On orthogenesis and the Impotence of Natural Selection in Species Formation, trans J. M. McCormack, Chicago: Open Court, p. 22 quoted in Bowler (1979: 50-51).

39

118

Chapter Four

constitution, in turn, predisposes the organism to develop in certain ways. This inclination can be considered the main directing element. Both scholars consider orthogenesis as the main process of evolution, put natural selection aside and hold that the evolution of a species proceeds in predetermined internally directed routes.40 Other non-Darwinian alternatives suggested in the second part of the nineteenth century promoted different versions of orthogenesis and supported the idea of a linear trend in evolution, but not necessarily a progressive interpretation of the evolutionary process (Grehan and Ainsworth 1985).

The Orthogenetic Analogy In the previous chapters, we have argued that one of the problems arising from selective models of scientific change is the different types of processes accounting for the occurrence of biological variations and the emergence of new ideas in the scientific community. In Darwinian selective models, variations are unrelated to the pressures of the organism’s environment or to its needs, while changes in scientific ideas are the trials to solve problems, the results of an argument, the answers to the specific needs of a scientist or group of scientists. An individual or a group of individuals have sought a solution to a problem and by trying to find it have generated a new idea. This, in brief, is the core reason why in our TH (Fig. 7) Science selection inherits from Directed variation. If a TH is built to represent orthogenesis, however, orthogenetic change would also inherit from Directed variation (if we replace Intention with a more generic Direction) and bring orthogenesis closer than Darwinian evolution to scientific change. This is correct because occurrences of orthogenetic variations are constrained by constitutive factors in organisms which establish a developmental direction. In science, existing knowledge constrains and directs the pattern of development. Furthermore, scientific knowledge gradually changes and new directions of development become available. In other words, the same mechanism described by Eimer in the quotation above is in place because, although variation is directed, new directions can be produced. If we 40

Initially, natural selection played a purely negative role in Eimer’s view, performing the elimination of harmful directions (Bowler 1979: 40). Eventually, he became more and more hostile towards selectionism, without falling into teleology though.

Science as a Special Case of Sociocultural Entity

119

accept Lakatos’ standpoint we can find additional features in common between the two processes. Lakatos (1970) argues that the basic unit of scientific evaluation is not a single theory or idea, but a whole research programme. The leading ideas of the research programme constitute the hard core, defining a set of commitments which cannot be abandoned without giving up the whole research programme. The hard core is shielded by a protective belt of auxiliary hypotheses whose function is to protect it from falsification. This belt is modified continuously, as a response to empirical developments and under the direction of the heuristic of the programme, which belongs to the core. Although a single negative test does not refute a whole research programme, it might lead to modifications in the protective belt of auxiliary hypothesis in order to accommodate the anomaly. The heuristic, like the laws of development in organisms, sets up a direction: it suggests to the scientists what paths of research to pursue or to keep away from. Embedded in the programme and shielded by the protective belt, it is the most stable developmental drive for scientific change, like the organisms’ constitution for evolution. The biological laws of growth, however, do not alone direct the course of evolution because orthogenesis is the result of the interference between the embodied laws of development and the contingent external environment. By the same token, heuristic and core ideas do not alone direct the development of the research programme, as it also has to account for the results of tests and experiments performed on nature. Also, when external conditions dictate, both the biological constitution and the scientific hard core can get (eventually) to be changed. When this occurs, a new research programme, or a species, is then generated. Finally, in Chapter 3 we also argued that the progressive nature of the scientific enterprise raises problems when we try to accommodate both scientific and biological evolution into the same mechanism of Darwinian selection. These problems are dissolved in the case of orthogenesis, as the overall trend of orthogenetic evolution is progressive, for instance because of the production of progressively more complex structures in place of simpler ones.

CONCLUSION

The major proponents of evolutionary analogies have converged, over time, towards a common approach. This approach is meant to provide a general theory of selective processes, with the objective of showing that scientific change and biological evolution are two instances of the same process and solve all disputes around the subject. This approach is relatively recent but not employed by all philosophers of science. Kuhn did not follow it, he used fragmentary evolutionary analogies to support some details of his view of scientific change. We showed that his arguments were based on a flawed view of evolutionary biology. Furthermore, the analysis of his view provided some insights on related issues and an important point for the rest of the exposition, namely concerning the possibility to find progression in scientific change but not in organic evolution. Campbell, by contrast, made a great contribution to the new approach but the analysis of the difficulties of his ‘blind variation’ revealed other important issues, namely the coupling of variation and selection in scientific change and its directionality, both absent in organic evolution. The scrutiny of Toulmin’s view, usually considered the precursor of the current approach to the evolutionary analogy, showed two risks of using a generalisation of selective processes. The first risk is the provision of too abstract an account of selective processes, so that the mechanisms in place in organic and scientific change look the same but all the details responsible for the explanatory power of the analogy are lost. The second risk is that the natural history origin of the idea of natural selection biases the researchers to use an abstraction still too reminiscent of the organic domain, with the consequence that all details of scientific change are forced into resembling organic ones. David Hull has attempted the most detailed generalisation of selective processes but we have shown that he did not succeed in solving the problems and avoiding the risks just mentioned. Our use of type hierarchies has also shown that it is possible to measure the relative strength of analogies between processes. The addition of this evaluative framework has allowed deeper and less subjective analysis of different processes classically associated with organic evolution. In particular, it has provided the conclusion that scientific and organic evolution cannot be regarded as two instances of the same general

122

Conclusion

process. The type hierarchy approach has also shown that it is misleading to conceive all social processes, scientific change included, as the same. Moreover, scientific change has been shown to be more closely analogous to orthogenesis than to Darwinian evolution.

REFERENCES AND BIBLIOGRAPHY

Achinstein, P. (1968) Concepts of Science, Baltimore, Maryland: John Hopkins Press. Ageno, M. (1984) ‘Importanza della Concezione Darwiniana nella Biologia Moderna’, La Nuova Critica, Quaderno III-IV(71-72): 7-31. Appleman, P. (ed.) (2001) Darwin, New York: Norton & Company. Aristotle (1941a) Poetics, translated by Ingram Bywater, in McKeon (1941). Aristotle (1941b) Rhetoric in McKeon (1941). Ayala, F. (1968) ‘Biology as an Autonomous Science’, American Scientist 56 (3): 207-221. —. (2000) ‘Evolution and the Autonomy of Biology’, Aquinas, XLIII(2): 284-311. Aronson, J.L., Harré R. and Way E.C. (1994) Realism Rescued, London: Gerald Duckworth & Co. Bateson, W. (1910) ‘Hereditary and Variation in Modern Lights’ in Seward (ed.), pages 85-101. Baghramian, M. (2004) Relativism, London: Routledge. Barlow, N. (ed.) (1967) Darwin and Henslow. The Growth of an Idea. Letters 1831-1860, London: John Murray. Barsanti, G. (1979) Dalla Storia Naturale alla Storia della Natura, Milano: Feltrinelli. Bird, A. (1999) ‘Scientific Revolutions and Inference to the best Explanation’, Danish Yearbook of Philosophy, 34: 25-42. —. (2000) Thomas Kuhn, Bucks: Acumen. Black, M. (1962) Models and Metaphors, Ithaca (NY): Cornell University Press. Bowler, P. J. (1979) ‘Theodor Eimer and Orthogenesis: Evolution by “Definitely Directed Variation”’, J Hist Med Allied Sci XXXIV: 40-73. —. (1983) The Eclipse of Darwinism, Baltimore: The Johns Hopkins University Press. —. (1984) Evolution: The History of an Idea, Berkeley: University of California Press. —. (1988) The Non-Darwinian Revolution, Baltimore and London: The Johns Hopkins University Press.

124

References and Bibliography

—. (1989) Evolution: the History of an Idea, Berkeley: University of California Press. —. (1992a) ‘From “Savage” to “Primitive”: Victorian Evolutionism and the Interpretation of Marginalized Peoples’ Antiquity, 66: 721-9. —. (1992b) The Fontana History of the Environmental Sciences, London: Fontana Press. —. (1993) Biology and Social Thought: 1850-1914, Berkeley: Office for History of Science and Technology University of California at Berkeley. —. (1996) Charles Darwin: The Man and His Influence, Cambridge: Cambridge University Press. —. (2001) Lecture Notes for ‘History of Science: Science and Religion’ Module, Belfast: Queen's University of Belfast (unpublished). —. (2002) Lecture Notes for ‘History of Science: Biology, Race and Society’ Module, Belfast: Queen's University of Belfast (unpublished). Bradie, M. (1986) ‘Assessing Evolutionary Epistemology’, Biology and Philosophy 1: 401-459. —. (1997) ‘Without Miracles: Universal Selection Theory and the Second Darwinian Revolution’, Philosophical Psychology Vol. 19, Issue 3: 399- 402. Brockman, J. and Matson, K. (1996) Science, Mind & Cosmos, London: Phoenix. Bush, S.F. and Kulkarni, A.B. (2002) ‘Genetically Induced Communication Network Fault Tolerance’, SFI Workshop: Resilient and Adaptive Defence of Computing Networks. Cain, J.A. and Darden, L. (1988) ‘Hull and Selection’ in Biology and Philosophy 3, 165-171. Callanan, J.J. (2008). Kant on Analogy. British Journal for the History of Philosophy, 16(4), 747-772. Calvin, W.H. (1996) How to Think What No One Has Ever Thought Before, in Brockman and Matson (1996). Campbell, D.T. (1974) Evolutionary Epistemology, in Schilpp (1974). —. (1985) ‘Toward an Epistemologically-Relevant Sociology of Science’ in Science, Technology, & Human Values 10(1), 38-48. —. (1987) ‘Blind Variation and Selective Retention in Creative Thought as in Other Knowledge Processes’ in Radnitzky and Bartley (1987). —. (1997) ‘From Evolutionary Epistemology Via Selection Theory to a Sociology of Scientific Validity’ (edited by Heyes, C. M. and Frankel, B.) Evolution and Cognition 3: 5-38. Campbell, M. Hoane Jr., A.J. and Hsu, F.-H. (2002) ‘Deep Blue’, Artificial Intelligence, 134(1-2), 57-83.

Evolutionary Analogies

125

Cassirer, E. (1950) The Problem of Knowledge: Philosophy, Science, and History since Hegel, New Haven: Yale University Press. Catania, A.C. (2001) ‘Selection as a Cause Versus the Causes of election’, in Hull et al. (2001). Chalmers, A. F. (1982) What is this Thing called Science?, Milton Keynes: Open University Press. —. (1990) Science and its Fabrication, Milton Keynes: Open University Press. Charlesworth, B. and Charlesworth, D. (2003) Evolution: A Very Short Introduction, Oxford: Oxford University Press. Cziko, G. (1995) Without Miracles, Cambridge (MA): The MIT Press. Cohen, L.J. (1973) ‘Is the Progress of Science Evolutionary?’, Brit.J.Phil Sci. 24: 41-61. Darden, L. and Cain, J.A. (1989) ‘Selection Type Theories’ Philosophy of Science 56: 106-129. Darwin, C. (1985) The Origin of Species, London: Penguin. —. (1987) The Correspondence of Charles Darwin. Vol. 2: 1844-1846, Cambridge: Cambridge University Press. Davidson, D. (1978) ‘What Metaphors Mean’ Critical Inquiry 5, no.1: 3147. —. (1984) Inquiries into Truth and Interpretation, Oxford: Oxford University Press. Davies, M.W. (2000) Darwin and Fundamentalism, Cambridge: Icon books. Dawkins, R. (1989) The Selfish Gene, Oxford: Oxford University Press. —. (1995) ‘A Survival Machine’, The Third Culture, available online: http://www.edge.org/documents/ThirdCulture/j-Ch.3.html (last accessed 18 August 2011). —. (1999) The Extended Phenotype, Oxford: Oxford University Press. —. (2001) River Out of Eden, London: Phoenix. —. (2003) A Devil’s Chaplain, London: Weidenfeld & Nicolson. Dennett, D.C. (1979) Brainstorms: Philosophical Essays on Mind and Psychology, Hassocks: Harvester Press. —. (1991) Conscious Explained, Boston: Little Brown. —. (1995) Darwin’s Dangerous Idea: Evolution and the Meaning of Life, London: Allen Lane, the Penguin Press. Dupré, J. (1994) ‘Species: Theoretical Contexts’, in Fox Keller and Lloyd (1994) Edelman, G. (1987) Neural Darwinism. The Theory of Neuronal Group Selection, New York: Basic Books. Eldredge, N. (1996) Reinventing Darwin, London: Phoenix.

126

References and Bibliography

Endler, J.A. (1986) Natural Selection in the Wild, Princeton: Princeton University Press. Ereshefsky, M. (2002) ‘Species’, in The Stanford Encyclopedia of Philosophy (Fall 2002 Edition), Edward N. Zalta (ed.), http://plato.stanford.edu/archives/fall2002/entries/species/ (last accessed 11 April 2007). Ferguson, A. (2002) Lecture Notes for ‘Molecular Ecology and Evolutionary Genetics’ Module, Belfast: Queen's University of Belfast (unpublished). Feyerabend, P. (1975) Against Method, London: NLB. —. (1978) Science in a Free Society, London: NLB. —. (1993) Against Method, London: Verso. —. (1996) Killing Time: The Autobiography of Paul Feyerabend, Chicago: Chicago University Press. —. (1999) ‘Realism and Instrumentalism: Comments on the Logic of Factual Support’ in Bunge, M., (ed.) Critical Approaches to Science & Philosophy, New Brunswick, NJ: Transaction Publishers. —. (2000) Conquest of Abundance: a Tale of Abstraction Versus the Richness of Being, Chicago: University of Chicago. Forrest, S., Hofmeyr, S.A. and Somayaji, A. (1997) ‘Computer Immunology’, Communications of the ACM 40(10): 88-96. Fox Keller, E. and Lloyd, E.A. (eds.) (1994) Keywords in Evolutionary Biology, Cambridge, MA: Harvard University Press. Fuller, S. (2001) ‘Discussion Note: is There Philosophical Life after Kuhn?’, Philosophy of Science Vol. 68, No.4: 565-572. Futuyma, D. J. (1995) Science on Trial: The Case for Evolution, Sunderland, MA: Sinauer Associates. —. (1998) Evolutionary Biology, Sunderland (MA): Sinauer Associates. George, S., Evans, D. and Davidson L. (2002) ‘A Biologically Inspired Programming Model for Self-healing Systems’, Proceedings of the First ACM SIGSOFT Workshop on Self-Healing Systems (WOSS '02): 102-4. Ghiselin, M. T. (1966) ‘On Psychologism in the Logic of Taxonomic Controversies’, Systematic Zoology 15: 207-215. —. (1974) The Economy of Nature and the Evolution of Sex, Berkeley, Cal.: University of California Press. —. (1994) The Imaginary Lamarck: A Look at Bogus “History” in Schoolbooks, http://www.textbookleague.org/54marck.htm (last accessed 18 August 2011). Gould, S. J. (1968) ‘Evolution and Triumph of Homology, or Why History Matters’, American scientist, 74(1): 60-69.

Evolutionary Analogies

127

—. (1977) Ever since Darwin. Reflections in Natural History, New York: W.W. Norton and Co. —. (1979) ‘Another Look at Lamarck’, New Scientist 84: 38-40. —. (1996) Three Facets of Evolution, in Brockman and Matson (1996) —. (2001) The Pattern of Life’s History, available online: http://www.edge.org/3rd_culture/gould/gould_p2.html (last accessed 18 August 2011). Gould, S. J., and Lewontin, R. C. (1979) ‘The Spandrels of San Marco and the Panglossian Paradigm: a Critique of the Adaptationist programme’, Proc. R. Soc. Lond. B 205: 581-598. Gould, S. J., and Vrba, E.S. (1982) ‘Exaptation: A Missing Term in the Science of Form’, Paleobiology 8: 4-15. Greene, J. C. (1991) ‘Progress, Science and Value: A Biological Dilemma’, Biology and Philosophy 6: 99-106. Grehan, J. R. and Ainsworth R. (1985) ‘Orthogenesis and Evolution’, Systematic Zoology, Vol. 34, No. 2: 174-192. Grene, M. (1986) ‘Philosophy of Biology 1983: Problems and Prospects’, Logic, Methodology and Philosophy of Science VII: 433-452. Hahlweg, K. (1986) ‘Popper versus Lorenz: An Exploration into the Nature of Evolutionary Epistemology’ PSA, Vol. 1: 172-182. Hobbes, T. (1962) The Leviathan in The English Works of Thomas Hobbes of Malmesbury, edited by Sir William Molesworth, Aalen (Germ.): Scientia. Howard, J. (1982) Darwin (1st edition), Oxford: Oxford University Press. Hesse, M.B. (1966) Models and Analogies in Science, Notre Dame (IN): University of Notre Dame Press. Henson, P.M. (1988) ‘A Short Note on Hull’s “A Mechanism and Its Metaphysics: An Evolutionary Account of the Social and Conceptual Development of Science”’, Biology and Philosophy 3: 192-193. Hull, D.L. (1965) ‘The Effect of Essentialism on Taxonomy--Two Thousand Years of Stasis (I)’, The British Journal for the Philosophy of Science, 15(60): 314-326. —. (1974) Philosophy of Biological Science, London: Prentice-Hall. —. (1982) ‘The Naked Meme’ in Plotkin (1982). —. (1988a) ‘A Mechanism and Its Metaphysics: An Evolutionary Account of the Social and Conceptual Development of Science’, Biology and Philosophy 3: 123-155. —. (1988b) ‘A Period of Development: A Response’, Biology and Philosophy 3: 241-263. —. (1988c) Science as a Process, Chicago: The University of Chicago Press.

128

References and Bibliography

—. (1990) The Metaphysics of Evolution, New York: State University of New York Press. —. (2006). The Essence of Scientific Theories. Biological Theory, 1(1), 17-19. Hull, D.L., Langman, E.R. and Glenn, S.S. (2001) ‘A General Account of Selection: Biology, Immunology and Behavior’ Behavioral and Brain Sciences 24: 511-573. Hume, D. (1996) An Enquiry concerning Human Understanding, Oxford: Claredon Press. Huxley, T.H. (1978) ‘Evolution in Biology’, reprinted in (1983) Collected Essays II, Darwiniana, London: Macmillan, 187-226. IAU – International Astronomical Union (2006) Resolutions 5 and 6: "Definition of a Planet in the Solar System" AND "Pluto", available online: http://www.iau.org/static/resolutions/Resolution_GA26-5-6.pdf (last accessed 18 August 2011). Jacob, F. (1997) La Souris, la Mouche at l’Homme, Paris: Odile Jacob. Jordanova, L. J. (1984) Lamarck, Oxford: Oxford University Press. Johnson, M. (1981) (ed.) Philosophical Perspectives on Metaphor, Minneapolis: University of Minnesota Press. Kauffman, S. (1993) Origins of Order: Self-Organization and Selection in Evolution, Oxford: Oxford University Press Kitcher, P. (1984) ‘Species’, Philosophy of Science, 51(2): 308-333. Kreuzer, F. (1983) Offene Gesellschaft - offenes Universum (Franz Kreuzer im Gespräch mit Karl R. Popper), Wien: Franz Deuticke, Krieger, N. and Davey Smith, G. (2004) ‘”Bodies Count” and Body Counts: Social Epidemiology and Embodying Inequality’, Epidemiologic Reviews 26: 92-103. Kuhn, T.S. (1970) The Structure of Scientific Revolutions, Second Edition, Enlarged. Chicago: University of Chicago Press. —. (1990) ‘The Road since Structure’, in PSA: Proceedings of the Biennial Meeting of the Philosophy of Science Association 1990, Volume Two: Symposia and Invited Papers, 3-13. —. (2000) The Road Since Structure, Chicago: University of Chicago Press. Lakatos, I. (1970) ‘Falsification and the Methodology of Scientific Research Programmes’, in Lakatos I., Musgrave, A.E. (Eds.) Criticism and the Growth of Knowledge, Cambridge: Cambridge University Press. Lakoff, G. (1993) ‘Contemporary Theory of Metaphor’ in Ortony (1993). Lakoff, G. and Johnson, M. (1980) Metaphors We Live by, Chicago: University Press.

Evolutionary Analogies

129

—. (1999) Philosophy in the Flesh: The Embodied Mind and Its Challenge to Western Thought, New York: Basic Books. Lamarck, J.B. (1914) Zoological Philosophy, London: MacMillan and Co. Lamport L. (2003) The Future of Computing: Logic or Biology, talk given at the Christian Albrechts University, available online: http://research.microsoft.com/en-us/um/people/lamport/pubs/future-ofcomputing.pdf (last accessed 18 August 2011). Lang, W.D. (1923) ‘Evolution: a Resultant’, Proc. Geol. Assn. 34: 7-20. Lewontin, R.C. (1978) ‘Adaptation’ Scientific American 239: 212-228. —. (1993) Biology as an Ideology. The Doctrine of Dna, London: Penguin. —. (1995a) ‘Il Potere del Progetto’, Sfera 43: 11-34. —. (1995b) Human diversity, New York: American Scientific Library. Lloyd, E. A. (1994) ‘Unit of Selection’, in Fox Keller and Lloyd (1994). Locke, J. (1975) An Essay Concerning Human Understanding, ed. P.H. Nidditch, Oxford: Claredon Press. Lorenz, K. (1977) Behind the Mirror: a Search for a Natural History of Human Knowledge, London: Methuen. Lorenz, K. (1982) ‘Kant’s Doctrine of the a priori in the Light of Contemporary Biology’ in Plotkin (1982). Losee, J. (2003) ‘Philosophy of Science and the Theory of Natural Selection’, ȆǹǿǻǼǿǹ, http://www.bu.edu/wcp/Papers/Scie/ScieLose.htm (last accessed 12 June 2003). McKeon, R. (ed.) (1941) Introduction to Aristotle, New York: Random House. Magee, B. (1974) Popper, Fontana Press: London. Martin, E. and Hine R.S., (eds.) (2000) A Dictionary of Biology, Oxford: Oxford University Press. Martinich, A. P. (1984) ‘A Theory for Metaphor’, Journal of literary Semantics 13: 35-56. Maynard Smith, J. (1988) ‘Mechanism of Advance’, Science 242(4882), 1182-1183. Mayr, E. (1942) Systematics and the Origin of Species, New York: Columbia University Press. —. (1970) Populations, Species, and Evolution, Cambridge, MA: Harvard University Press. —. (1976) ‘Is the Species a Class or an Individual?’, Systematic Zoology, 25(2): 192. —. (1977) Evolution and the Diversity of Life, Cambridge, MA: Harvard University Press.

130

References and Bibliography

—. (1981) La Biologie de l’Èvolution, Paris: Hermann. —. (1982) The Growth of Biological Thought: Diversity, Evolution and Inheritance, Cambridge, MA: Harvard University Press. —. (1991) ‘The Ideological Resistance to Darwin’s Theory of Natural Selection’, Proceedings of the American Philosophical Society, 135(2): 123-139. —. (1993) One Long Argument: Charles Darwin and Genesis of Modern Evolutionary Thought, London: Penguin. —. (1996) ‘What is a Species, and What is Not?’, Philosophy of Science, 63: 262-277. —. (1998) This is Biology – The Science of the Living world Cambridge, MA: Harvard University Press. Millstein, R.L. (2002) ‘Evolution’ in Machamer, P. and Silberstein, M. (ed.) Philosophy of Science, Oxford: Blackwell. Montuschi, E. (2000) ’Metaphor in Science’ in Newton-Smith (1994) pp. 277-282. Mullis, K. (1998). Dancing Naked in the Mind Field. New York: Pantheon Books. Murphy, G. L. (1996) ‘On Metaphoric Representation’, Cognition 60: 173-204. —. (1997) Reasons to Doubt the Present Evidence for Metaphoric Representation’, Cognition 62. Newton-Smith, W.H. (ed.) (2000) A Companion to Philosophy of Science, Oxford: Blackwell. Nozick, R. (1975) Anarchy, State and Utopia, Oxford: Blackwell. Numbers, R. L. (1999) Darwinism Comes to America, Cambridge, MA: Harvard University Press. O’Grady, P. (2002) Relativism, Bucks: Acumen. O’Hara, R. (1988) ‘Homage to Clio, or Toward an Historical Philosophy for Evolutionary Biology’, Systematic Zoology 37(2): 142-155. —. (1998) ‘Population Thinking and Tree Thinking in Systematics’, Zoologica Scripta 26: 323-329. O’Hear, A. (1980) Karl Popper, London: Routledge and Kegan Paul. —. (ed.) (1995) Karl Popper: Philosophy and Problems, Cambridge: Cambridge University Press. Ortony, A. (1993) (ed.) Metaphor and Thought, 2d ed., Cambridge: Cambridge University Press. Plotkin, H. C. (1982) Learning, Development and Culture, London: John Wiley & Sons. Pollard, J.W. (ed) (1984) Evolutionary Theory: Paths into the Future, Chichester: John Wiley & Sons Ltd.

Evolutionary Analogies

131

Popper, K. R. (1957) The Poverty of Historicism, London: Routledge and Kegan Paul. —. (1972) The Logic of Scientific Discovery, London: Hutchinson. —. (1974) Conjectures and Refutations: the Growth of Scientific Knowledge, London: Routledge and Kegan Paul. —. (1976) Unended Quest: An Intellectual Autobiography, London: Fontana. —. (1984) ‘Evolutionary Epistemology’, in Pollard (1984). —. (1978) Objective Knowledge. An Evolutionary Approach. Revised Edition, Oxford: Oxford University Press. —. (1992) In Search of a Better World: Lectures and Essays from Thirty Years, London: Routledge and Kegan Paul. —. (1994) Tre saggi sulla mente umana, Roma: Armando (Italian translation of the following papers: ‘Natural Selection and the emergence of Mind’ (1978), ‘Natural selection and the Place of Mind in Nature’ (1982), ‘Die Erkenntnistheoretische Position der Evolutionären Erkenntnistheorie’ (1986), ‘Kommentar zu “Die Erkenntnistheoretische Position der Evolutionären Erkenntnistheorie”’ (1987)) —. (1995) A World of Propensities, Bristol: Thoemmes. —. (2002) Alles Leben ist Problemlösen, München: Piper. Popper, K. R., and Eccles, J.C. (2003) The Self and Its Brain, London: Routledge. Pust, J. (2001) ‘Natural Selection Explanation and Origin Essentialism’, Canadian Journal of Philosophy 31(2): 201-220. Quine, W.V. (1969) Ontological Relativity and Other Essays, New York: Columbia University Press. Radnitzky, G. and Bartley, W. W. (1988) Evolutionary Epistemology, Rationality, and the Sociology of Knowledge, La Salle, Illinois: Open Court. Radick, G. (2000) ‘Two Explanations of Evolutionary Progress’, Biology and Philosophy 15: 475-491. Rensch, B. (1966) Evolution above the Species Level, New York: John Wiley and Sons. Renzi, B. G. (2009) ‘A Type Hierarchy for Selection Processes for the Evaluation of Evolutionary Analogies’, Journal for General Philosophy of Science, DOI: 10.1007/s10838-009-9099-8. —. (2009) ‘Kuhn’s Evolutionary Epistemology and its being Undermined by Inadequate Biological Concepts’ Philosophy of Science 76(2): 143159.

132

References and Bibliography

Richards, R.J. (1977) ‘The Natural Selection Model of Conceptual Evolution’, Philosophy of Science 44(3): 494-501. Ricoeur, P. (1977) The Rule of Metaphor, Toronto: University of Toronto Press. Rose, S. (1997) Lifelines: Biology, Freedom, Determinism, London: Allen Lane. Rose, H. and Rose, S. (eds.) (1976) Ideology of/in the Natural Sciences, New York: Macmillan Press. —. (eds.) (2000) Alas, poor Darwin: arguments against evolutionary psychology, London: Jonathan Cape. Rummelhart, D.E. (1993) ‘Some Problems with the Notion of Literal Meanings’ in Ortony (1993). Ruse, M. (1986a) Taking Darwin Seriously, Oxford: Blackwell. —. (1986b) ‘Biology and Values: A Fresh Look’, Logic, Methodology and Philosophy of Science VII: 453-466. —. (1989) ‘Great expectations’, The Quarterly Review of Biology 64: 463468. —. (1995) Evolutionary Naturalism, London: Routledge. —. (1999) ‘Evolutionary Ethics: What Can We Learn from the Past?’, Lygon 34(2): 435-451. —. (2000) ‘Is Evolutionary Biology a Different Kind of Science?’, Aquinas XLIII(2): 252-282. Ruse, M., and Wilson, E.O. (1985) ‘The Evolution of Ethics’, New Scientist 108(1478): 50-52. Salzmann, D., Eliezer S., Krumbein and Gitter L. (1983) ‘Lazer-driven Schock–wave Propagation in pure and layered Targets’ Physical Review Vol. 28, Number 3: 1738 -1751. Schilcher, F. and Tennant, N. (1984) Philosophy, Evolution & Human Nature, London, Boston, Melbourne and Henley: Routledge & Kegan Paul. Schilpp, P.A. (ed.) (1974) The Philosophy of Karl Popper, La Salle, Illinois: Open Court. Steward, A.C. (ed.) (1910) Darwin and Modern Science, Cambridge: Cambridge Univ. Press, Shapin, S. (1996) The Scientific Revolution, Chicago: University of Chicago Press. Shapere, D. (2000) ‘Scientific Change’, in Newton-Smith, W.H. (ed.) (2000), pp. 413-422. Sharrock, W. and Read, R. (2002) Kuhn. Philosopher of Scientific Revolution, Cambridge: Polity Press.

Evolutionary Analogies

133

Skipper, R.A.Jr. (1999) ‘Selection and the Extent of Explanatory Unification’, Philosophy of Science 61 (Proceedings): S196-S209. Smith, R. E. and Taylor, N. (1998) ‘A Framework for Evolutionary Computation in Agent-Based Systems’, Proceedings of the 1998 International Conference on Intelligent Systems: 221-224, ISCA Press. Soanes, C. and Stevenson, A. (2005) Oxford Dictionary of English, Oxford University Press, Oxford. Sober, E. (1984) The Nature of Selection: Evolutionary Theory in Philosophical Focus, Cambridge, MA: MIT Press. Sokal, R.R. and Crovello, T.J. (1970) ‘The Biological Species Concept – A Critical Evaluation’, American Naturalist, 104(939): 127-53. Sowa, J.F. (1984) Conceptual Structures, Reading (MA): AddisonWesley. SSR: see Kuhn (1970). Sterenly, K. (2001) Darwin vs Gould: Survival of the Fittest, Duxford, Cambridge: Icon Books Ltd. Thagard, P. (1993) Computational Philosophy of Science, Cambridge (MA): The MIT Press. Toulmin, S. (1961) Foresight and Understanding, London: Hutchinson. —. (1965) The Philosophy of Science, London: Hutchinson. —. (1967) ‘The Evolutionary Development of Natural Science’, American Scientist 55(4): 456-471. —. (1972) Human Understanding. Volume I, Oxford: Clarendon Press. —. (1974) ‘Rationality and Scientific Discovery’, in Cohen, R.S. and Wartofsky, M. (eds.) Boston Studies in the Philosophy of Science, Dordrecht: Reidel. Turner, P.C., McLennan, A.G., Bates, A.D., White, M.R.H (1997) Molecular Biology, Oxford: BIOS. Von Neumann, J. (1966) Theory of Self-Reproducing Automata, Chicago: University of Illinois Press (Originally Published in 1953). Way, E.C. (1994) Knowledge Representation and Metaphor, Oxford: Intellect Books. Watkins, J. (1995) ‘Popper and Darwinism’, in O’Hear (1995). Wilson, D. S. (1990 ) ‘Species of Thought: A Comment on Evolutionary Epistemology’, Biology and Philosophy 5: 37-62. Wilson, E. O. (1986) Biophilia, Cambridge, MA: Harvard University Press. —. (1998) Consilience : the unity of knowledge, London: Little, Brown. Worrall, J. (1995) ‘”Revolution in Permanence”: Popper on TheoryChange in Science’ in O’Hear (1995).

134

References and Bibliography

Wuketits, F. M. (1990) Evolutionary Epistemology and Its Implications for Humankind, Albany: State University of New York Press. Wynne-Edwards, V.C. (1962) Animal Dispersion in Relation to Social Behaviour, Edinburgh: Oliver and Boyd. Zimmer, C. (2002) ‘“Digital Biology”: Is This Chip Educable?’ The New York Times, March 10th.

INDEX

Achinstein, 14, 123 adaptation, 5, 6, 14, 26, 30-39, 49, 70, 81, 117 Ainsworth, 115-118, 127 analogy, 1, 5-8, 11-14, 19, 22-34, 37-39, 42, 45-49, 53-59, 65-69, 87-91, 94, 112, 121 Aristotle, 8, 9, 123, 129 Aronson, 15-18, 66, 69, 86, 87, 108, 110, 123 Artificial selection, 77, 86 asteroid, 48, 74 autonomy, 79, 97, 103 bioepistemology, 5 Bird, 27-29, 33, 123 Black, 13, 14, 89, 123 blind-variation-and-selectiveretention. ^ĞĞ^sZ bottom-up, 66, 67, 70 Bowler, 115-118, 123 Bradie, 5-7, 124 BVSR, 39-41, 45, 48 Cain, 54, 70-72, 75, 80-82, 89, 124, 125 Campbell, 4, 6, 28, 39-52, 56, 58, 85, 91, 121, 124 Catania, 74, 75, 125 Clonal Selection, 67, 71, 87 coevolution, 35-38 Cohen, 56, 125, 133 comparison view, 12, 13, 16, 67, 70 complexification, 30, 32 complexity, 13, 26, 31, 32, 58, 70, 76, 94, 116 conceptual change, 44, 46, 49, 53, 55, 84, 90, 95, 96, 105, 111 content, 6, 9, 14, 18, 47, 50-52, 96, 98, 102, 103, 109, 111, 114 content selection, 51, 52

convergent realism, 60, 108-111 copying, 93, 98-100, 114 course, 5, 28, 29, 32, 37, 43, 48, 5052, 59, 61, 70, 73, 82, 83, 86, 100-104, 107, 119 crystals, 48 Cyber selection, 88 Cyber Selection, 79, 98 Cziko, 81, 125 Darden, 54, 70-72, 75, 80-82, 89, 124, 125 Darwin, 1, 7, 53, 123-127, 130-133 Darwinian selection, 77, 80-82, 91, 92, 113 Deep Blue, 46, 47, 124 Dennett, 80, 125 Directed selection, 77, 88, 113 direction selection, 52 ecological niche, 22, 30, 33-38 EEM, 5, 6, 21, 36 EET, 5-7, 21, 36-39, 48 Eimer, 116-118, 123 Endler, 4, 126 environment, 2, 3, 7, 11, 27-40, 44, 52, 55-59, 68-72, 76, 79-84, 99101, 117-119 evolution of epistemological mechanisms. ^ĞĞD Evolutionary Analogy, 1 Evolutionary – Analogy View, 5 evolutionary biology, 1, 7, 22, 29, 30, 33, 34, 37, 69, 106, 121 evolutionary epistemology, 1, 5, 6, 13, 21, 25, 29, 39, 52 strong evolutionary epistemologies, 7, 19 weak evolutionary epistemologies, 7, 66

136 evolutionary epistemology of theories. ^ĞĞd experiment, 27, 28, 51, 52, 80 explanandum, 13, 14, 90 explanans, 13, 14 externalist selection, 51 fitness, 3, 14, 32, 49, 57-60, 69 Fuller, 21, 126 Futuyma, 1-4, 27, 35, 36, 81, 84, 101, 115, 116, 126 genes, 3, 4, 68, 81, 94, 96, 105 genetic drift, 1, 2, 70, 80, 81 genius, 9, 44, 58 Gestalt, 23 giraffe, 28, 29 global, 30, 31, 50, 52, 54, 85 Grehan, 115-118, 127 group selection, 4, 49 Haacke, 116 Hahlweg, 5, 127 Hesse, 13, 14, 89, 127 Hobbes, 10, 127 Hull, 7, 41, 49-56, 60-62, 65-77, 8085, 89-114, 121, 124-128 Huxley, 116, 117, 128 identity, 1, 7, 47, 75, 86-90, 95, 103, 104 immune system, 81 incommensurability, 23, 26 inheritance, 16, 18, 33, 36, 77, 82, 85, 98, 113, 114 intentionality, 14, 50, 56, 65, 77, 80, 83, 90 interaction, 14, 33, 34, 43, 51-54, 68-71, 74, 76, 80, 91-95, 98-101, 106-108, 112 interactive view, 12-14 interactor, 68, 72, 76, 98 interpretation, 5, 10-13, 22, 24, 3438, 107, 118 Jerne, 81 Johnson, 9, 11, 128 Kasparov, 46 Kitcher, 69, 106, 108, 128 knowledge, 5-7, 11, 15, 24-26, 30, 36-48, 58, 79, 87-91, 118, 133

Index Kuhn, 13, 21-38, 58, 121, 123, 126, 128, 131-133 Lakatos, 119, 128 Lakoff, 11, 128 Lewontin, 32-38, 127, 129 lineage, 68, 69, 80, 93-108, 113 local, 23, 30-32, 36, 40, 50, 54, 69, 81, 84, 85 Locke, 10, 129 Lorenz, 4, 5, 127, 129 Losee, 5, 129 Martinich, 11, 129 Maynard Smith, 49, 129 Mayr, 2, 3, 55, 94, 105, 116, 129 mechanical selection, 74, 75, 82 Mendel, 105 Mendelian, 1, 116 Mercury, 28 metaphor, 7-13, 22, 24, 39, 44, 48, 49, 53, 86, 89 milieu, 57, 85 mind, 6, 11, 34, 42-49, 70, 75, 77, 83, 87, 107, 113 Modern Synthesis, 1 Montuschi, 12, 88, 130 Mullis, 45, 130 mutation, 2, 3, 8, 22, 26, 91 natural selection, 1-8, 12, 14, 21, 25, 32, 33, 36, 40, 41, 50, 53, 66, 67, 70, 76-81, 89-91, 94, 116, 118, 121 Newton, 28, 108, 130, 132 ontology, 24, 86-88, 109 orthogenesis, 115-119, 122 paradigm, 23-25, 28, 30, 58 phenotype, 2, 3, 69 philosophy, 4, 12 plausibility, 47, 54, 109, 111 Plotkin, 5, 127, 129, 130 Polymerase Chain Reaction, 45 Popper, 4, 5, 7, 21, 25, 36, 39, 40, 97, 127-133 population, 2, 3, 4, 33, 35, 38, 44, 55, 60, 61, 70, 74, 76, 94, 105, 117

Evolutionary Analogies populational approach, 55 progress, 14, 22-32, 39, 52, 54, 58, 59, 84, 85, 90, 93, 109, 116, 117 punctuated equilibrium, 26, 29 Radick, 30-32, 131 rationality, 55 Read, 24, 34, 38, 132 reality, 7, 27, 29, 33, 34, 38, 46, 47, 50-52, 55-58, 88, 91, 102, 109 recombination, 3 Red Queen hypothesis, 32 replication, 14, 48, 49, 68-80, 9199, 103, 105, 114 replicator, 68, 94, 97, 113 Richards, 13, 39, 89, 132 Ricoeur, 9, 132 Rummelhart, 11, 132 Ruse, 5, 37, 91, 132 Saturn, 48, 74 Schilpp, 39, 41, 124, 132 science, 5-8, 12-14, 18, 21-29, 32, 39, 41, 47-50, 54-60, 66, 76, 8385, 88-93, 96, 99, 101, 106-111, 115, 118, 121 scientific change, 6, 12-14, 21-24, 29-33, 37, 39, 49-69, 84, 89-94, 100, 101, 107-122 scientific method, 50 scientists, 4, 7, 12, 13, 23, 28, 33, 34, 37, 44-47, 50-61, 66, 82, 96112, 118, 119 selection, 1-7, 12, 25-32, 39-57, 6597, 102, 103, 107, 112-121, 131 semantic network, 15, 18, 73, 86 Sharrock, 24, 34, 38, 132 simile, 9, 12 Skipper, 69, 133 Soanes, 75, 133 Sober, 4, 133

137

software, 79 Sowa, 15, 16, 133 specialisation, 26, 30, 32 speciation, 22, 27, 117 species, 1-8, 26-39, 53-62, 81, 9496, 102-108, 112-119, 126 Stevenson, 75, 133 structure, 10, 18, 37, 48, 55, 68, 91, 95, 97, 101, 104, 111, 113 survival, 2-6, 25, 35, 53, 71, 72, 81, 103 taxonomy, 24, 31, 60-62, 74, 91 tenor, 13, 89 term-tokens, 94, 105-108 term-types, 94, 106, 107 test, 24, 27, 28, 57, 91, 119 TH, 15-19, 66-93, 98, 106, 109-114, 118 theory change, 26, 107 theory-ladenness, 51 Tonegawa, 81 top-down, 66 Toulmin, 4, 52-63, 91, 121, 133 Traditional View, 9 trial-and-error, 40, 44, 48 Turner, 11, 95, 133 type, 15-19, 39, 55, 65-92, 101-111, 115, 121 type-hierarchy. ^ĞĞd, Van Valen, 32 variation, 1-3, 8, 39-42, 46-49, 5359, 70-75, 80-83, 88, 93, 113118, 121 veil of perception, 109 verisimilitude, 108-112 Way, 15-18, 65, 66, 89, 123, 133 Wilson, 6, 132, 133 Wuketits, 5, 134