Biology and Philosophy 13: 479–504, 1998. c 1998 Kluwer Academic Publishers. Printed in the Netherlands.
The Evolutionary Structure of Scientific Theories JOHN S. WILKINS 3 Peel Grove Mt Martha 3934 Australia1 E-mail:
[email protected]
Abstract. David Hull’s (1988c) model of science as a selection process suffers from a two-fold inability: (a) to ascertain when a lineage of theories has been established; i.e., when theories are descendants of older theories or are novelties, and what counts as a distinct lineage; and (b) to specify what the scientific analogue is of genotype and phenotype. This paper seeks to clarify these issues and to propose an abstract model of theories analogous to particulate genetic structure, in order to reconstruct relationships of descent and identity. Key words: Dawkins, Hull, meme, gene, species, theory, selection, semantic
1. The theory as genotype the replicator-interactor distinction is often used in ways that subvert much of its potential usefulness. As a remedy I suggest that these two terms should refer to two mutually exclusive domains of selection, one that deals with material entities and another that deals with information and might be termed the codical domain. : : : The material and the codical are separate domains because of a dearth of shared descriptors. : : : A message is always coded in some medium, but the medium is not really the message. (Williams 1992: p. 10) :::
In a departure from previous attempts to apply evolutionary models to conceptual change in science, David Hull (1988a, 1988c) eschewed traditional epistemological concerns and applied a general model of selection to cover biological and epistemic evolution equally. Hull developed some general categories to explain evolution by selection irrespective of the entities evolving. As a contribution to the debate over genic selection2 he introduced the terms interactor and replicator to refer respectively to the interface entity between the environment – what Sober (1984: p. 279f) called the benchmark entity, the phenotype – and to the entity that is transmitted in replication, in biology, the genotype. Dawkins (1977) had introduced the term meme to describe the equivalent in social evolution of genes, which he believed were the basic units
480 of selection (‘genic selectionism’). Hull treats biological entities as parts of an inclusively nested hierarchy, and, like Sober, allows selection to occur at any level of that hierarchy so long as there is interaction of the entity with its ecology and replication of an underlying structure that causally codetermines the interactive traits of the entity.3 Hull summarises his analysis of science thus: I argue that one of the fundamental mechanisms in the conceptual development of science is conceptual inclusive fitness. Scientists can pass on replicates of their ideas as their ideas directly to later generations of scientists, but they can also cooperate with their contemporaries in promoting their collective goals. Initially, in the history of science, scientists worked in relative isolation from their contemporaries. They built on past work but did not band together to pursue joint research. Rather rapidly, however, the demic structure of science materialized and continues to characterize science to the present. As the years have gone by, it has grown increasingly difficult for scientists to work in social isolation from their contemporaries. However, just as the advent of sexual reproduction introduced a new set of partially conflicting goals into biological evolution, the formation of research groups in science introduced a comparable set of partially conflicting goals into science, as scientists had to cooperate with their conceptual competitors. (1988c: pp. 22f. Cf. 283, 287f, 310 on conceptual inclusive fitness; pp. 159ff and 310ff on competition and cooperation.) :::
The concept of conceptual inclusive fitness is a Hamiltonian-like theoretical concept to account for the value to a (self-interested) professional player of views not developed by themselves. They share in the credit by promoting a view that is ultimately successful. The mechanisms of this demic (small-group) process of competition and cooperation are straight undirected selection processes, and the similarities between biological and conceptualsociocultural evolution are more than vague analogies. They are due to the fact that, according to Hull (1988c: chapter 11) a general analysis of selection applies equally to both types of evolution by selection, and neither is more basic than the other vis a´ vis selection (1988c: p. 424). Since Hull had amended Dawkins’ earlier vehicle/replicator distinction, this is now known as the HullDawkins distinction (cf. Eldredge 1989; Lloyd 1989; Williams 1992). At this quite general level, both types of evolution by selection are exactly the same kind of process, in every relevant detail (Hull 1988c: p. 287). The replicators in conceptual evolution are the substantive content of science – beliefs, goals, practices, data (1988c: p. 434). The interactors are the scientists themselves (1988c: pp. 438, 478ff). The equivalent entities in biological evolution are typically genes (replicators) and organisms (interactors), although as a result
481 of Hull’s analysis, he takes the view that selection occurs at no specific level, but that instead, inverting the usual formulation of the units of selection controversy, one defines an individual in selection processes by the fact that it and its kind are causally relevant as replicators or interactors (1988c: pp. 411, 428). Hull therefore has a general theory of selection processes that applies equally to all evolutionary processes irrespective of the substrate or ontological level – according to Hull, the biological is no more fundamental than the conceptual in terms of the processes of change by selection. He draws a number of his theses from debates over various forms of biological evolutionary theory – genic selectionism, punctuational/graduational theories, and attempts at general analyses of evolution of species. Some of the ideas he appropriates are those of Mayr (population thinking), Ghiselin (spatiotemporally individuated species, lineages) and Dawkins (vehicles, interactors and memes), all turned to new purpose. He applies these insights to the problem of conceptual change in science, arguing that intentions and internal methods notwithstanding, what accounts for the spread and adoption of theoretical change is the success of one scientist’s idea in developing promising lines of research in the eyes of other scientists, in the seeking and gaining of credit for ideas and in their being used by other scientists (conceptual inclusive fitness), in local cooperation in demes and competition from outside the deme but within the invisible college, and so on. Dawkins argued in his 1977 (cf. also his 1982) that memes underwent an evolution-by-selection process independent of, and in some ways in opposition to, the biological process. In parallel with his ‘selfish gene’ theory – in which the selection process maximises the spread of successful genes (actually of some complex of molecular DNA that codes for selectable phenotypic traits) in such a way that it is useful, if sometimes misleading, to speak of genes ‘acting’ in such a way so as to maximize their own benefit – Dawkins also speaks of memes seeking to increase their differential survival against their competitors and to increase their frequency in the population. Dawkins often rejects intentional interpretations of this ‘selfish gene/meme’ theory, treating it as merely a shorthand mode of expression of game-theoretic (‘bookkeeping’) aspects of evolutionary processes, following such approaches as those of Maynard Smith (1983) and Axelrod (1981, 1984), applying zero-sum and iterated Prisoner’s Dilemma analyses to evolutionary interactions. This game-theoretic approach assumes a ‘rational egoist’ individual agent – that is, the ‘unit of selection’ is purely self-interested. Somewhat surprisingly, macroevolutionary effects such as the evolution of cooperation (Axelrod 1981, 1984; Hardin 1971, 1982; Ullman-Margalit 1977) and altruism (Hamilton 1964; Dawkins 1977) are not only modelable this way, but
482 often show up as robust and stable strategies that tend to dominate certain initial populations, and having done so reach an equilibrium state (or at least a stable state that is unlikely to change in the medium term). Following Maynard Smith, Dawkins terms these evolutionarily stable strategies (ESS), applying the term to macroevolutionary-effect-resulting strategies such as the now famous ‘Tit-for-tat’ strategy in iterated Prisoner’s Dilemma games. According to Dawkins, we can explain the development of a range of apparently counter-Darwinian phenomena in game-theoretic terms – parasitic interdependencies, biased sex ratios in social insects, predator-prey ‘arms races’, and so forth. Hull’s analysis is an extension of Dawkins’ suggestion of memic evolution. He argues that it is the demic structure of science (a ‘deme’ is a local breeding population of organisms) that acts to further the interests of theories (as memes) and also the professional careers of individuals who originate or adopt successful theories. Demic science, says Hull, is a strategy that evolved very early in the history of modem science, allowing a rapid increase in the complexity and rate of change. In fact, like the evolution of scientific reproduction, demic isolation, internal cooperation and inter-demic competition has raised the variability and selection rates enormously. Does this ensure that science is progressive? Is science adapting as a result of the demic ESS? Sexual reproduction appears to hedge against extinction and facilitate adaptation (Dawkins 1986: p. 268) so perhaps the demic nature of science is an analogue of sexual reproduction, in the sense that it allows ‘crossing over’ of memes within a professional lineage in the deme, allowing for more variation than naked selective environments would. 2. Evolutionary genes and memes I use the term gene to mean ‘that which segregates and recombines with appreciable frequency.’ : : : A gene could be defined as hereditary information for which there is a favorable or unfavorable selection bias equal to several or many times its rate of endogenous change. (Williams 1966: p. 44, cited in Dawkins 1982) The memes are being passed on to you in altered form. This looks quite unlike the all-or-nothing quality of gene transmission. It looks as though meme transmission is subject to continuous mutation, and also to blending. It is possible that this appearance of non-particulateness is illusory, and that the analogy with genes does not break down. (Dawkins 1982: p. 195) The double helix DNA strand that makes up the genetic material of living things on earth maybe seen logically as a long line of points (allele bases)
483 that may take one of four values (the amino acids represented by the letters A, C, G, and T). The human genome has around three billion base pairs. Cistrons, the ontogenetically active clusters and sequences of these values that code for proteins when transcribed by RNA, are made up of these four values as written words are made up of letters in alphabetical languages. There are millions of these active sequences in any specific species, and they are constantly changing through recombination, crossover and point mutations (allelic transcription errors at meiosis). This is the raw variation that natural selection acts upon, for many of these changes affect the gross phenotypic traits of the specific individual and its descendents (many also don’t – perhaps the majority of DNA is this ‘neutral’ genetic material). We might therefore represent the genome of a species as a structured set of coordinates, each axis of which takes one of four values, and clusters of coordinate values make active sequences that affect the interactive traits of the individuals they code. (This is a gross oversimplification of both the nature of genetic structure and of the developmental process by which genes guide ontogenetic development. For a start, the helix is a helix, that is, not only does each allele have its linear neighbours, but it lies adjacent to the next twist of the strand and its properties are in part a result of this topology. Nevertheless, we can usefully continue to deal with the primary structure of the DNA molecule with that caveat. More important is the fact that ontogeny is guided by the genetic code, but not determined by it. Changes to the developmental environment will affect how a gene sequence is expressed. It is often convenient to say ‘this gene codes for that trait’ but this is sometimes misleading. As Oyama (1985; cf. Dennett’s 1995 discussion of “readers”) notes, genes code only in a context, and only for a norm of reaction.) What Dawkins is saying above is that memes may be replicated with a blending inheritance rather than a particulate inheritance like genes. He is not saying that memes do not form structures. He is not saying that memes cannot be replicated because they are not particulate. He is merely drawing out what appears to be a disanalogy with genes. The structure of genes is not essential to the success of Darwin’s evolutionary theory (except that if genes are subject to blending then as Jenkin and Mivart observed selection would be unable to function, because variations would be swamped), and it is not essential to Hull’s. Although Darwin’s ‘gemmules’ theory of inheritance toyed early on with the idea that sexual fertilization might involve mixing rather than fusion of hereditary factors (Fisher 1930: pp. 1–2), overall he saw heredity as the result of the blending of the variation of parental traits. The criticism of Fleeming Jenkin that caused Darwin so much heartache showed a difficulty with blending inheritance, that as Fisher later calculated, is that in the absence of rapid mutation, variation in a population will decrease
484 by roughly half each generation (Fisher 1930: p. 5). In consequence, less than one thousandth of existing variation can be ten (or in the sexual case, twenty) generations old.4 Where cultural fashion is blending, as perhaps in styles of artistic endeavor and recreational behavior, this may account for the rapidity of the development of novelty, for a blending inheritance model requires a ‘mutation’ rate many orders of magnitude greater per generation than a particulate model: “To maintain a stationary variance (under blending inheritance) fresh mutation must be available to supply half of the variance so lost (per generation) : : : Almost every individual of each generation must be a mutant : : : and moreover must be a mutant in many different characters.” (Fisher 1930: pp. 4–5). When Dawkins (1995) used the metaphor, or analogy, of ‘digital’ as opposed to ‘analogue’ inheritance, he was making effectively the same point. As any communications engineer knows, signals degrade, and boosting the signal for repropagation introduces noise – a digital transmission can be more cleanly filtered of noise than an analog transmission. The integrity of the ‘signal’ of cultural transmission relies strongly on its modularity (in words, syntax, statements, concepts and practices). A matter that this sort of model of cultural change has to establish in general is whether there are modules that tend to persist over many generations and iterations. The ‘meme’ meme is vague and slippery. As Dawkins uses it, the term applies to a range of transmitted cultural practices from snatches of tunes to ideas. In the hands of his disciples, it unfortunately has become a catchall bucket in much the same manner ‘paradigm’ did for post-Kuhnians. Clearly Dawkins wishes to make the analogy with Williams’ ‘evolutionary gene’ and therefore requires that something cultural with a determinate structure is differentially transmitted showing homology and convergence or their antitheses. That is probably enough for the purpose of his comments in 1977 and 1982, but to avoid the indiscriminate and ultimately vapid use of the term, we need some clearer specification of what is to fall under the rubric. For Dawkins, following Williams, a gene must be information that has a selection bias applied to it that exceeds its endogenous rate of change. A meme, by extension, is also information that has selection biases applied greater than its own tendency to change anyway. In other words, it must be information that leads to ‘ecological’ success, that codes for adaptations. Dawkins wants to define something upon which evolution acts, and the high fidelity nature of genes in replication makes them an obvious choice in some ways. However, one can just as easily say that the evolutionary process requires heredity and interactivity (traits), and go looking for things that satisfy these a priori stipulations. Assuming that the evolution of sociocultural traits occurs, one posits memes and goes looking for them.
485 If a Hull-Dawkins analysis does apply to scientific change, some geneanalogue has to be made out, at least in the sense that a replication process must occur with some of the properties that make evolution by selection possible. As the quotation above makes plain, Dawkins feels that the non-particularity of culture involves a difficulty for a Darwinian analysis of culture, and he is not alone in this.5 I would therefore like to be able to present a general model of theoretical change that both clarifies the nature of what is undergoing that selection process and develops a closer analogue with allelic competition and particulate evolutionary genetics in general. To make Hull’s case, we need to find some close genotype analogue, which (as Heyes 1988 asks) constitutes the phenotype analogue, the professional scientist. I have assumed that the appropriate entity is the theory (without excluding the possibility of other levels of epistemic entity, such as a ‘conceptual system’ – Richards 1987). But there is a very great difference between particulate inheritance in genetics and the transmission of semantic structures, or so it seems. If a science is transmitted as blending inheritance, what use is an evolutionary model, to an historian, say? We could not model scientific change the way population genetics can model particulate inheritance of traits, and such a model would anyway suffer from the problems of blending inheritance models generally: ‘swamping’ of novelties and the high rate of creativity required to maintain or direct stasis. Hull does not think, and I do not think, that sociocultural change involves a constantly high rate of novelty. So I will attempt to make out an abstract model for the particulate nature of epistemic inheritance, and show how such a model might be of use to an historian of epistemic change. Generalising the functional nature of the genotype, we might say that it is an information structure which (variably) generates interactive traits through replicative transformation mechanisms. The interactive traits that result are the phenotypic features that are causally involved in the selection process. At least at the more formal and communal level, a theory is a structure of information; tokens of it generate ‘traits’ (applications, interpretations, experiments) that vary according to the individual scientist’s understanding (according to Hull, even co-workers usually have different theory interpretations; 1988c: p. 493); and it is replicated with high fidelity (through transformation mechanisms of cultural transmission). Like genes, theory tokens are populational entities, and the state of a theory is a profile of the population of those entities (scientists) that are constituted through those mechanisms. In our analysis of the evolution of scientific knowledge, then, the theory is the fundamental genetic structure of the ‘organism’. What, then, are the appropriate analogues of ‘species’, ‘organism’, ‘phenotype’ and so forth? Addressing this requires some preparatory investigation into the nature of theories and explanations.6
486 3. The semantic conception In his introduction to the state of the philosophy of science at the time, Frederick Suppe (1977a) described two major viable alternative views of science to the Received View – the Weltanshauungen Analyses such as those of Toulmin, Kuhn and Hanson, broadly in the tradition of the later Wittgenstein; and what he terms the Semantic Approaches.7 The latter represent theories as mathematical models with determinate structures (1977a: p. 222) that have an intended scope the parameters of which assume that the scope of phenomena are isolated systems, rather like Ellis’ (1979) ‘ideal systems’. According to Suppe’s version (1977a: p. 224), one can represent the parameters of a theory as values (coordinates) in a Euclidean phase space – a term borrowed from thermodynamics8 – for an n-dimensional (in Suppe where n is the number of parameters of the physical theory) coordinate system or space.9 The major advantage of such a model is that if one can locate a theory as a set or cluster of coordinates, its position relative to other theories or its own ancestors can be specified and analysed using standard tools of taxonomic analysis. One of the seminal discussions that more clearly explicates the spatial metaphor is provided through consideration of contrast spaces. The concept of a contrast, or possibility, space is due to Alan Garfinkel (1981). He considered the anecdotal joke about bank robber Willie Sutton, who, when asked by a priest why he robbed banks, replied that was where the money was. Garfinkel comments: The difference between (Sutton and the priest) is that they have two different contrasts in mind, two different sets of alternatives to the problematic: Sutton robs banks. They are embedding the problem to be explained in two different spaces of alternatives, which produces two different thingsto-be-explained, two different objects of explanation. The object of explanation here is therefore not a simple object or state of affairs, but more like a state of affairs together with a definite space of alternatives to it. (1981: p. 21, italics original) The Sutton case is in effect posing at least two different questions: one alternative space is that “Sutton
banks”; the other is that “Sutton robs ”. Garfinkel continues: The effect of such differing spaces of alternatives is not always a joke; what aspect of a given state of affairs we take to be problematic radically affects the success of failure of potential explanations. For an explanation to be successful, it must speak to the question at hand, whether explicit or implicit, or else we will have failures of fit like Sutton and the priest. What we need, therefore, is some way of representing what is really getting
487 explained in a given explanation, and what is not. The contrast spaces gives us such a representation of one basic way in which explanation is “context relative”. My claim is that this relativity-to-a-contrast-space is quite general; I will call it explanatory relativity. (1981: p. 22, italics original)10 A contrast space is the semantic metric in which a statement is located; a question seeks a semantic coordinate in that (or a set of) contrast space(s). Contrast spaces are neither global nor absolute. According to Garfinkel, they are small-scale, local spaces (1981: p. 40), essentially similar to the state spaces of physics; geometric representations of the possibilities of a system. He says that they are not simply the spaces of possible worlds, but ‘more like equivalence classes of possible worlds (under the relation ‘differs inessentially from’) with almost all possible worlds excluded altogether from the space’. In physics, a phase space represents the variables of the system, within which a trajectory surface is described by the equations governing possible trajectories. The trajectory of any actual body (it need not be a physical trajectory, but a transformation of some kind) will be inscribed as a path on the surface. A phase space operates as a metric, or coordinate measuring system, within which a matrix, or set of constrained pathways, describes the nature of the objects that occur within it.11 The contrast space is, as it were, an explanatory metric; the range of possible alternatives, the explanatory matrix. The point of the Sutton joke is, of course, that the explanatory matrix of paternalistic social and spiritual theology excludes the alternatives of the pragmatic considerations of bank robbing. Sutton has at least one more contrast (and at least a few less) than the priest. The contrast space in Garfinkel’s account represents the presuppositions of the explanation, and the questions that seek it. When a contrast is not offered for a specific explanation, the ‘missing’ contrast forms part of the relative presuppositions of that explanation. Some of these arise due to the structural constraints of the system under explanation – Garfinkel calls these the structural conditions. They serve to limit the degrees of freedom or the possibility states of the system. Structural conditions are analogous to the kinematical conditions of analytical dynamics, and form the trajectory surface mentioned above. They constrain the explanations available and simplify the available alternatives (1981: pp. 44–47). Systemic structural conditions are the internal relations between the possibility spaces of individual elements of the system, and explanations of them presuppose the overall structure of the system and cannot explain it. Garfinkel rejects reductionism on the grounds that explanation is level-sensitive. Moreover, he argues (1981: p. 63f) that not all identical perturbations of a system’s microstates result in the same macrostates. There are ‘critical points’ that are unstable, at which change
488 is qualitative rather than incremental.12 In ironic phrasing that would sound much better in German, he says the underlying space is partitioned into equivalence classes within which differences do not make a difference but across which differences do make a difference. (1981: p. 65) The phase transitions that result from crossing these catastrophic boundaries render reduction untenable, in that any complete explanation of such macrolevel changes requires as a presupposition a topographical map of the system marking the phase change boundaries. He enunciates a principle – Whenever a global property is not simply the sum of N individual properties : : : the explanation of that global property will involve structural presuppositions. (1981: p. 72) Establishing the global independence of a property is achieved by posing generalised contrary-to-fact conditional questions of the form: ‘if individual i had not exhibited the specific property Ni , would the global outcome have differed?’ If the answer is no, or probably not, explanation of the individual state requires a global structural condition.13 The salient features of a contrast space model of explanations for our present purposes, are (i) every question presupposes a set of contrasts that may be offered in explanation; (ii) contrasts that are not available for a specific explanation are part of the structural constraints on that explanation; (iii) explanation is sensitive to the level of the semantic system within which it occurs; and (iv) because microstates do not always result in the same macrostates, simple reductionism in explanations of macrostates to microstates is impossible. The application of these points to scientific explanation is apparent: (i) scientific theories presuppose a domain of viable explanations and exclude some types of explanation as non-viable; (ii) certain assumptions are made that form part of the scientific background to a theory; (iii) scientific explanation is level sensitive (so that the explanation of meiosis, for example, is not found in the physical properties of the atoms that make up the cells, at least not in biology); and (iv) explanation is stochastic and non-reducible, because phenomena at one level are not determined rigidly by phenomena at lower levels.
4. The phase space gene analogue One of the oft-repeated criticisms of the Hullian and similar programs to model conceptual change in evolutionary terms, especially in science, (Dawkins 1977: p.112; Mosterin 1988: p. 206; Mun´evar 1988: p. 212; Smith
489 1988: p. 215) is that there are no obvious analogues to such ‘molecular’14 particulate genetic structures as ‘evolutionary genes’, alleles (or cistrons), genetic loci, or species’ karyotypes (the chromosomal structures of species). Despite the fact that Darwin himself could provide no satisfactory molecular or particulate theory of heredity (his own theory of pangenesis was rejected by early Mendelians and is a soft-inheritance theory, cf. Ghiselin 1969; Mayr 1982; Desmond and Moore 1991), Darwinian evolutionary theory did meet its objectives in explaining the range and mechanisms of biological change at the macro level. Likewise, Hull’s program may also meet its objectives vis a´ vis scientific change. Nevertheless, an idealised model of the memetic structures of scientific theories and disciplines may be sketched in such a way that could show that Hull’s basic Darwinian approach is feasible. I say ‘idealised’ because in practice the range of variables would be extremely large and complex, and to make matters even worse, it is unlikely that specific and precise values can be given to memes of individuals. We begin with the by no means uncontroversial view that a theory is a set of propositions, embedded in a discipline (cf. Suppe 1977a: p. 121). Let D be the domain of a discipline: a set of propositions at issue, which partitions into subsets of elementary issues, which are sequences of exclusive coordinates. The range of a domain is determined by the conventional agreements of the disciplinary community as to what is at issue, and may vary over time. D is the issue space of a discipline. It is the contrast space in which the structure of the discipline occurs, the disciplinary metric of the disciplinary matrix. Elementary issues are mutually exclusive sequences and not further decomposable in terms of the discourse of the discipline. An issue might be a theoretical question – does an electron have a determinate radius? Is evolution directed? It might be a question of empirical data, or methods or aims. An issue admits of a singular position for any holder of a view in the theoretical community, so that, if a disjunct of views may be coherently held open by any one individual, the issue is not elementary, and may be further decomposed into constituent issues. An issue is ‘live’, or viable, relative to a discourse domain if not all members of the discipline agree on the position to be taken, or, if there is consensus in fact, it would not exclude the holder of that view from membership of the discipline if a contrary view were taken. (‘Taking’ a view means here accepting a position as true, or likely to be true, and cognate approaches – believing in a view, and so forth – to the exclusion of other positions.) A position may be a fuzzy coordinate, but we will ignore that complication here. Non-viable issues are elided from D and newly viable issues added, resulting in a domain shift for that discipline. Note that not all possible issues, past or future as well as logically possible, will occur in the
490 domain, but only real or implicit issues. Likewise, not all the domain will be or may be occupied, since some theories will have been rejected as not credible or will never have been considered at all. A theory T is a set containing coordinates from some live elementary issues, that is, it is a set of propositions chosen from D where each proposition is represented by a coordinate on an issue sequence. Therefore T is a set of coordinates: < a; b; c; d; : : : n; t >
where each coordinate represents a position that may be taken on that issue. Since the Hullian approach is an historical theory and is therefore necessarily time-indexed, t presents the theory at a time. A global theory relative to a discipline (one that purports to deal with the totality of a discipline’s domain) is a theory that contains one coordinate from each issue of D. Global theories are a special case of theories. Assuming that a scientific theory enters the discipline by being a set of propositions considered by one or more scientists, each theory will be indexed relative to those individuals: T t i represents a theory held by i at t, where i is an individual scientist, a group of individuals, or the collective discipline. Since an individual who is a practising scientist may instantiate several conceptual ‘memotypes’ (Tennant 1988), or more accurately, may have different theoretical commitments at different times in his/her career, we can therefore treat the conceptual scientific instantiations (Mayr as adherent of soft inheritance and Mayr as adherent of hard inheritance (Mayr 1982), for example) as distinctly different individuals – that is, as professional profiles. A professional profile is a practitioner of a discipline, instantiated in an epistemic agent at a time or for a duration of time, and is constituted by the set Pt i of theories held by that individual. That is, a professional profile is an epistemic entity,15 whose theoretical commitments causally determine its nature as a professional scientist. It follows that at any one time, an individual person may instantiate only one professional profile within a discipline, but may adopt a series of professional stances over time (and may be a professional in a number of distinct disciplines). Let us further assume that the developmental process of becoming a professional consists in acquiring theoretical commitments from other professionals, immediately or proximately. This will involve the transmission of the great majority of the theoretical commitments of those professionals. In order to count as an ‘intellectual descendent’ under this extension of Hull’s account, there must be sufficient isomorphism of structure between teacher and student, founder and follower, for differentiation between a member of a given school of thought and other schools within the discipline. A professional is
491 a descendent of another professional if there is the appropriate causal relationship between them and the former resembles the latter enough to be distinguished from other lineages. If a putative descendent has no expressed views in a range of the domain, they may be taken as holding their ancestral views, since that is how they will be interpreted by their colleagues. However, intellectual ancestry is a matter of demonstrated (actual) influence. Care must be taken, in memetic phylogeny as in genetic, to separate homology from analogy through historical investigation of citations and personal contact and instruction. Theoretical positions, which are the coordinates taken by epistemic entities, are thus in effect the cultural analogues of genes in science. I say genes, and not traits, because a theoretical position does not by itself interact with the scientific analogue of an ecological system. Theoretical interaction, to follow Plotkin (1994), occurs when these positions are translated into behaviour, such as experimental activity, argument and publication, grant acquisition to continue investigation, and so forth. Selection – that is to say, differential success at resource acquisition closely correlating with differential success at propagation of these traits – can operate only upon expressed genes; likewise with memes. In order for selection to occur, theoretical points must give rise to, if not determine, interactive behaviour. Expressed views will be subject to selection insofar as their resource acquisition is closely covariant with their ability to spread. Under these assumptions, there is likely to be a Darwinian process of conceptual evolution.16 Theories and other conceptual structures will change over time to gain or lose selective advantage over competitor theories, etc., or be eliminated from the domain as a viable alternative. The mechanism Hull proposes as sufficient to explain theoretical change, conceptual inclusive fitness, works on my analysis through professionals: the self-interested desire of a professional for credit, either directly or vicariously through the promotion of a lineal ancestor’s or descendent’s views, will ensure that cooperation within the deme drives the articulation of theories, and the attempted elimination of competitor theories. It is an identifying trait of a professional that he or she is self-interested, to some degree, in attaining credit. Those who do not have this mark, are not professionals, but rather, dilettantes. It should be noted that there needs to be at least some conventional standards by which theories may be tested and the relative merits of theory change assessed. In a mature discipline, these standards will be consensual to all or most of the disciplinary community and the institutions of which the deme is a part. Typically, they will comprise methodological considerations at the operational and meta level (testing procedures and theory structure rules). Testing through physical experiment is also a hallmark of science in a way
492 that is not true of other intellectual domains such as theology or literature. In other respects, though, similar analyses could be given of them. Interpreting a domain as an n-dimensional hyperspace comprised of orthogonal axes for each elementary issue, the theoretical commitments of professionals, measured in a hypothetical census at t, would be represented by points (to some degree of exactitude) within the domain issue space. Given the constraint of lineal descendency of theories, the history of a discipline will likewise be constrained by past developments. A theoretical lineage may not abandon most or all of its positions and still count as a lineage. Kuhnian ‘paradigm shifts’ at the theory level will not occur. A new lineage may arise in parallel (as a competitor), and the census may show a sudden shift in the relative frequencies, but ex hypothesi lineal descendency will be seen to have occurred. A deme will be a cluster (according to Hull, a multimodal cluster) of partial coordinate sets representing a census of the views of some individuals of a discipline. This is entirely a posteriori: it will require a censual mapping to delineate, whatever initial intuitions may be. Typically, we should expect a discipline to include numerous demes in subsets of the issue space – demic structures will appear depending upon which axes are mapped. If we map the theoretical commitments Pi tl : : : tq over time of a scientist or science along this n-dimensional set in this way:
= T2 =
T1
< t1 ; l
< t2 ; l
1
1 1 1 1 1; l 2; l 3; l 4 : : : l n
>;
2
2 2 2 2 1; l 2; l 3; l 4 : : : l n
>;
:::
Tq
=
< tq ; l
q
q
q
q
1; l 2; l 3; l 4 : : : l
q
n >;
where lj n is some value determined by the semantic content of the sentence that represents it (cf. Tennant 1988), we will be able to locate (with varying degrees of fuzziness) that individual entity’s theoretical commitments at tq in a Cartesian space of n-dimensions. The general theoretical ‘profile’ of a scientist, deme, research tradition, or professional science will be mapped by a cluster of such individual points, which can be averaged to represent the ‘state’ of the practitioner, science or research program at tq .17 Mapping these averages over a range of times will provide a trajectory of the development of the views of that person/science/speciality.18 The variability of the views in a deme or specialisation cluster will be represented by the size of a cross section of that trajectory at tq and the rate of conceptual change will be represented by the distance of the space traversed by the trajectory between the sampled times. Furthermore, the creation or existence of rival research programs in a
493 given domain will appear as branching or separate trajectories, overcoming some difficulties of subjective historical interpretation. This method, as with Osgood’s semantic differential method, provides a useful research tool in that, by selecting some distinguishing conceptual axes, one can model change rates and variation spread empirically and, if not objectively, then intersubjectively (depending upon the evidence available) irrespective of the researcher. It would also assist in the recognition of the ‘active’, rapidly changing, concepts in a case of theoretical change, in that those axes would show the greatest change. Knowing this would assist in identifying both the ‘selective pressures’ (Sober 1984) that brought about the change in question and those concepts that are analogous to ‘neutral DNA’. With this apparatus in hand, we can then measure the relative distance in a disciplinary issue space between two propositions, and the rate of conceptual change. This gives us some handle on ‘conceptual distance’ and on the speed of conceptual change in science, allowing us to locate and examine relatively stable and unstable conceptual elements and to posit and test motivating causes for that change or lack of it. It does not matter what the dimensional scale is, so long as the differences between populations or disciplines or theoretical coordinates can be compared. Rates of change should be commensurate. This position may seem in some ways to depend on a holistic theory of meaning. I don’t think that it actually commits any historian of science19 to any particular theory of meaning at all. Two things need to be clear at this point: one, that an evolutionary model is not founded upon this (or any other) model of cultural change; two, that this approach is intended to serve as a heuristic and a methodological starting point for comparison and contrast of theoretical views. I don’t expect it to do duty as any kind of semantic theory as well. It may, however, fit in with such theories as the Semantic View. Insofar as they involve some theory of meaning, this model may be congenial to it. Now, on this view, rival theories may be compared within the matrix of the domain of the discipline, and are therefore not even partially incommensurable, since they represent alternative positions on common axes.20 Even if an issue does not occur for a given theory A, since it competes within a discipline against another theory B that does have a position on an issue, the disciplinary domain includes the issue, and so as a member theory A must take an implicit position on it, even if only to deny its relevance – for example, while the acts of special creation by God claimed by creationism are no longer accepted as credible explanations within biology, the constant proposal of such explanations by religiously motivated biologists suffices to ensure the viability (the ‘contentiousness’) of that issue within the discipline. Here are Garfinkel’s relative presuppositions in play, and the issue arises from a contrast space not internal to the theory. This model is therefore not a theory
494 of meaning view of the relationship of theories to each other, although it does not preclude such a model. Whether or not a dispute is internal or external to a domain will depend upon the criteria used to delineate a domain. This is entirely an artefact of the research concerns of the investigation. In other words, whether creationism is a religion/science political dispute or a dispute within biology, is determined by whether deity is part of the theoretical equipment of the biological community. Who is fit to judge? Obviously the views expressed by the professional associations and institutions has enormous influence, but in a specific (e.g., historical) context, it may be that professional opinion is overturned. There is no easy solution to this matter, as there hasn’t been since it was raised. A similar problem faces Kuhn, Lakatos and Hull: what defines a community? If it is the views it holds (behaviour it evinces/socioeconomic properties it has), then there is a certain circularity in pointing to those criteria to assess whether a given individual is properly part of that community. All I can say is that there are social and natural clumps about, and we can observe that clumping in various fallible and correctable ways.21 Finally, a sequence of domain shifts will determine a tradition space, and a lineage inscribed within a sequence of domain issue spaces will comprise a tradition. Thus, a tradition, over long periods, may have few or no elementary issues in common at the start and finish of the lineage segment, and yet still be related by descent. All lineal sequences in this model are assumed to be highly constrained in the range of alternative positions they may take at any time, by the previous history of the lineage – if a lineage is marked by certain theoretical commitments for a long time, it is very unlikely to drastically alter those commitments, and they may be taken to be rather central, and relatively immune from revision. However, Sober’s (1981) claim that embarking upon a revision process will position some statements as relative a priori is unnecessary. The revision rules themselves may be revised, and no statement is absolutely immune from revision, as Quine maintained. It is clearer, on this account, what may evolve and for what kinds of reasons. An individual’s professional profile may evolve through the intentional selection of theoretical commitments in a Popperian fashion. A deme’s views may evolve through the conventional articulation of a promising research program. A theory may evolve through disciplinary acceptance and emendation by application of conventional assessment methodology (a sufficiently strong theory, especially a global theory, may force the emendation of the common methodological considerations). And a discipline itself may evolve as issues cease to be relevant or become part of the wider background of the disciplinary tradition. In all these cases, change occurs as the less successful variants
495 are selected against, and the more successful achieve greater distribution in the population. This simple model takes no account of the competition of disciplines, nor of the milieu of a discipline within a social or cultural tradition (more precisely, set of traditions), but it may be extended to cover those cases where relevant. For our purposes, it is assumed that a discipline is crosscultural, relatively robust in its isolation from other disciplines, and stands in an ‘ecosystem’ of its own. The interaction of disciplines will occur as they share issues and either concur or conflict on those matters. The isolation of a discipline – its ‘speciality’ – will depend upon it having a unique cluster of issues not substantially addressed by other disciplines. Where this method would not assist researchers is in the ordering of the set. The same elements may be arranged as allotropes to form different theories, and some kind of semantic net structure analysis would be necessary to distinguish them at that level (cf. Quine and Ullian 1978). The semantic phase space conception of theories provides a useful tool for distinguishing between different traditions and sciences, identifying rates of conceptual change and evolutionary homologies or synapomorphies (shared derived characteristics) and identifying and modeling the theoretical structure of the professional communities that hold them and are characterised by that possession. The question whether theories change in gradualistic, episodic or catastrophic fashion will be decided upon the results of historical investigation using this or a similar method of quantifying the theoretical and disciplinary rates of change. If, as is now being accepted in biological evolutionary studies, variable rates of change and periods of stasis occur in different ways, the leaps of development required by early Kuhnian views will be seen to characterise only the most extreme kind of change if any. Toulmin (1970: p. 45) writes: (following Kuhn’s renunciation of radical theoretical change) the occurrence of a ‘scientific revolution’ no longer amounts to a dramatic interruption in the ‘normal’ continuous consolidation of science: instead it becomes a mere ‘unit of variation’ within that very process of scientific change : : : And, once we acknowledge that no conceptual change in science is ever absolute, we are left only with a sequence of greater and lesser conceptual modifications differing from one another in degree. The distinctive element in Kuhn’s theory is thus destroyed, and we are left looking beyond it for a new sort of theory of scientific change. This theory will have to go beyond both Kuhn’s concept of ‘revolutions’ and the naive uniformitarian views which he renounced, just as Darwin’s evolutionary reinterpretation of palaeontology went beyond both the catastrophism of Cuvier and the uniformitarianism of Lyell. (Italics original.)22
:::
496 Eldredge and Gould’s (1972) punctuational model of variable rates of rapid evolutionary change in biological speciation processes, followed by long periods of relative stasis and extinction, is like Kuhn’s, becoming less heterodox with time, but also less radically un-Darwinian (cf. Eldredge 1985,1995). It treats natural selection as usually more of a stabilising process than a strictly adaptive process, except during allopatric speciation, and the jury is still out after nearly a quarter of a century as to what admixture of views will prevail amongst biologists. However, being able to quantify change rates in Toulmin’s sense would enable a better historical understanding of theory change and the development of new sciences and specialisations. We should expect that cultural change, like biological change, will range from the slow to rapid, and from the continuous to episodic. Specific ‘atomic’ commitments (elementary issue positions) may be ‘borrowed’ across traditions in a kind of cross-fertilisation – conceptual hybridisation of this kind prevents a need for close Tractarian correspondence of atomic ideas with atomistic ‘facts’. There is also likely to be the same adaptive information lag (‘load lag’) conceptually as in biology, limited degrees of movement due to historical and structural constraint, and conservative retention of the bulk of theoretical structure over time. The question of what concepts are adapted to is not really one of realism/anti-realism (despite van Fraassen’s concerns in his 1980, cf. Suppe 1989), interesting though those questions are in themselves. There is a clear sense in which one wants to say our concepts are adapted to the world even if one is vague about whether that adaptiveness is merely instrumentalistic or has some other metaphysical import. Since Hume it has been a presupposition of inference that the nature of the world is not undergoing random variation itself, so once a science has reached a certain level of ‘material’ adaptedness, in contrast to ‘social’ adaptedness, it ought not to change in its broadest structure (which partly explains, for example, the retention of Newtonian structures of explanation and formulae in contemporary physics at given levels and boundary conditions, despite the supposed incommensurability of that theory with its successor theories which led Kuhn and Feyerabend to suppose that there was no conceptual lineage as such – Kuhn 1970; Feyerabend 1969, 1975a, 1975b). According to Ghiselin, Darwin held a broadly similar view.23 But since a science interacts through essentially social vehicles, changes in the social environment also count as ecological challenges to concepts. Concepts are adapted, as part of theoretical structures, to the social/natural environment of their vehicle, the individual. If therefore, the social environment varies, as it generally will, the aptive fitness of existing theories will vary also, without any necessary decrease in the empirical fitness of those theories.
497 How might this ordered data set be collected for a study of a theory’s phylogeny? As with all historical investigations, a certain amount of reconstructive work will be required, usually from the writings (published or personal) of the practitioner/s, but in the case of a current research program, a ‘census’ may be taken of all or some representative sample of the practitioners, perhaps also over time at a period of instability. This would amount to a subjective weighting of opinions considered acceptable, and to a profile of the conventions of the time. It is important to note that the initial choice of operant axes will be one of convenience for the researcher, but that one should expect that ‘active’ axes will show themselves upon analysis. A further complication is the appearance of novel active axes and the disappearance of the inactive or irrelevant ones from the domain of discourse (the existence of God and his theodicy is no longer a matter for scientific discourse, New Age writers and creationists notwithstanding24). Inactive concepts may form part of a more inclusive tradition or set of traditions, and thus eventually drop out of the specific phase space of a tradition. Whether a conceptual allele is evolutionarily active will depend upon the nature of the change occurring. This ‘higher level’ change may or may not be evolutionary. Mere change of live issues does not of itself constitute evolutionary change – there would need to be adaptive advantage and replicative differential success over competitor clusters for this change to be evolutionary. What gross change of axes does indicate is the rise of a new tradition – a new science, Weltanschauung, or paradigm in the sense of ‘disciplinary matrix’ (Kuhn 1977a). It is to be expected that there will also be borderline cases in which it is unclear whether a new tradition has resulted or merely a development of an existing one. In this situation, we are in similar straits to a biologist in the midst of the pheneticist/cladist debates in taxonomy (Hull 1988c: pp. 200–276; Gould 1983: pp. 356–365; Panchen 1992) or the reception of punctuational views (Dawkins 1986: chapter 9; Ruse 1989: essay 5; Eldredge 1995; Dennett 1995: pp. 282–303). It is clear also that not all change is to be imputed to selective retention due to relative adaptive advantage. Conceptual drift, as with genetic drift, is likely to act as both a cause of change and a source of variation that in intellectually isolated traditions such as a deme provides the raw material for selection. Drift will also often be due to what are sometimes called ‘external’ factors such as idiosyncratic educations, social or personal situations and psychologies, or historical circumstances. Their existence cannot be ignored, but are not relevant in themselves to theoretical change until the concepts they engender are exposed to the ‘meme-pool’ and the selective pressures of actual intellectual economies.
498 5. Social research There are similarities to this modeling process in methodologies already employed in the social sciences, notably in the health sciences. Social researchers Glaser and Strauss (Glaser 1978; Strauss 1986) developed a method known as ‘grounded theory research’, part of the so-called ‘discovery mode’ of social research, in which a hypothesis is tested for properties of a specific category. The researcher aims to have a theory to work by which is both modified by and influences the selection of the collection of qualitative data (Glaser 1978: p. 36; cf. Becker 1993). ‘Qualitative’ is, of course, a code word in social research for ‘subjective’, and the data collected represents attitudinal expressions from the subjects under study, delineated according to the researcher’s field of interest. It therefore lends itself to a social constructionist approach (Charmaz 1989). In grounded theory, though, there is a feedback loop to the researcher’s choice of categories, so that emerging issues will enable the researcher to redefine what data needs to be collected, and how it is to be interpreted. Out of a virtual infinity of categories, only a manageable set of categories are chosen for study, and this set is refined according to judgements based upon the data collected. Adherents of grounded theory, along with other phenomenological methods in social science, often have a flawed understanding of the possibility of completeness in their models, however, and it is instanced here only as a feedback model of social research, whereby the categories under investigation are refined as the study progresses. This is entirely a social and semantic approach to research. As with all research into subjective states, it is strongly open to preconceptional bias and ideological distortion, but it is widely used to guide research into social interactions, and cannot simply be dismissed as another sociological fashion. I have alluded to earlier approaches of this kind (Osgood and others 1967; Tulloch 1972), and something of this kind is a persistent theme in social, and therefore historical, research. I raise this method here as an example of the prima facie plausibility of the approach I am sketching – if this sort of approach underpins much actual social research, then it may be useful in studying the social aspects of science. However, studying the social aspects alone does not give a rounded picture of theory development. I reject the simple constructionist approach to science, as I would the simple social constructionist approach to epidemiology. No matter what the social influences on the course of science or disease may be, there are nevertheless asocial causal agents that initiate and continue these social movements. Having climbed these abstracted heights, the question will arise whether it can be applied.25 While the objections to Hull’s conception may not be fatal in principle, perhaps historians of science will be able to ignore this
499 and other approaches and continue as before. I feel that the techniques developed for both pheneticist and cladist analysis, of clusters in the phase space and Hennigian comb cladograms will provide an approach not previously available for the study of sociocultural phenomena, and change in science in particular, but without any reason to apply them, they will be ignored for the more familiar techniques of multifactorial analysis and narrative. One such technique that could be used is the Wagner Similarity method (cf. Wiley and others 1991). This method establishes an instance matrix of characters (presence and absence, but it could be any value 0 > x > 1) and calculates the sum of the modulus of differences to give the Wagner distance between taxa (or theories). From this a net diagram of relationships, or an unrooted tree, can be drawn up to give a sharp notion of the overall similarity of taxa or theories. The same data can be recast as a rooted cladogram if a sister-group can be selected. Software, both commercial and public domain, is available to perform these analyses and others using well-established algorithms. While this will not definitively establish ancestry of a conceptual lineage, it can be used to test hypotheses of ancestry, and to overcome the Whiggish tendency of historians to read preferred modern views back into a historical subject, such as Darwin’s reliance or not on now-discredited views such as embryological recapitulation (cf. Richards 1992).
Conclusion The concept of a meme is relative to the selection pressures operating upon it. There is nothing that ultimately ‘is’ a meme – neuronal connections, semantic entities, etc. A meme is something cultural that gets selected in a manner that is correlated with its spread through a community of cultural agents. In science, a meme is a theoretical position subject to the selection pressures of science: testing, further research, dissemination and use. Failure to distinguish this relativity of replicators to interactive success can give rise to the sorts of objection made to Hullian models of the kind expressed in the reactions to his 1988b (Mosterin 1988; Mundvar 1988; Smith 1988; Tennant 1988): that scientists are not the causal outcome of their theoretical commitments. When Popper wrote that our theories die in our stead, he was in a sense right and wrong. The professional career of a scientist can easily die if the scientist’s professional profile expresses the wrong memes, and those memes can have lower fitness in the sense that they have fewer instantiations in the next generation of scientists and so Popper was wrong that they die in the scientist’s stead. However, there is a decoupling from biological fitness of the fitness of concepts, and to that extent Popper was right. Theories are
500 not usually biological causes, and genes are not usually theoretical causes (scenarios of genetic engineering to one side). The model outlined here attempts only to show that memes may be, at least in science, particulate, and may be investigated as such and that this investigation can distinguish lineages of descendency. Hull’s generalised model of evolution clarifies matters in biological evolution, and mutatis mutandis in sociocultural evolution. Acknowledgements I have benefitted from discussions and comments from many individuals, without necessarily gaining their agreement or entirely learning their lessons. Special thanks are due to John D. Collier, David Hull, Kim Sterelny and HansCees Speel for comments on earlier drafts of this paper and the thesis, and encouragement. Hannelore Best provided me the information on grounded theory. I have also gained from the comments of the reviewers, who pointed out limitations and shortcomings that I trust I have at least acknowledged if not rectified. Notes 1
The major part of this paper was written as a masters thesis at the Department of Philosophy, Monash University, Wellington Road, Clayton 3168, Australia. 2 Hull 1980, cf. Richard Dawkins’ concept of vehicle/selector, Dawkins 1977. Sterelny 1995 provides a summary of the debate. 3 See Mayr 1982 for a discussion. 4 I am indebted to an anonymous reviewer for pointing out that the lack of particularity as such is not the problem with blending inheritance, and that particularity does not solve the problem of the decrease in variation. Darwin’s and his contemporaries’ difficulty was that the pangenetic particles had no alleles; there were no discrete states that persisted through reproduction. 5 Eg, Plotkin 1994: pp. 221. 6 The sources for the discussion that follows are van Fraassen (1970, 1980), Suppe (1977a, 1977b), Garfinkel (1981) and Dyke (1988), and broadly follows what is now known as the semantic conception of theories. See also Suppe 1989. There have been objections to the claims made for the Semantic Conception – e.g., Ereshefsky (1991) argues that it is no better able than the Received View to deal with the empirical interpretation of laws despite rejecting the simple observation term-theory dichotomy of the latter. However, for my purposes here, the conception of a theory as a family of models is all that is required in the first instance. 7 On the history of post-positivistic philosophy of science as a dialogue between the “Wittgensteinians” – Hanson, Kuhn, Feyerabend and Lakatos – and Popper (or more properly a monologue by Popperians against an unresponsive opposition), see Radnitzky (1982). 8 See Weber and others 1988 for discussions of thermodynamics in relation to evolutionary thought, especially the essays by Wiley, Collier and Olmstead. 9 The ‘space’ metaphor for concepts has quite a history in this century. van Fraassen (1970, 1980) and Lloyd (1989: pp. 33–41) following him use the thermodynamic term state spaces
501 while earlier Quine (1969b) had hinted at something similar with the term quality spaces, Garfinkel (1981) with his contrast spaces and Wittgenstein with his closely similar logical space (1922: 1.13, 2.11, 2.202, 3.4, 3.42, 4.463). We will retain the generic term phase space, cf. Prigogine and Stengers (1984) and Wiley (1988) (Hull occasionally refers to conceptual space – perhaps he has some similar model in mind). The phase space model I am proposing below bears a close relation to the semantic differential measurement method of Osgood and others (1967), in which conceptual structures, semantic connotations and individual attitudes can be mapped on n-dimensional semantic spaces, and the changes over time measured as a summation of the spatial differences between clusters. I have also made use of Tulloch’s (1972) issue space, developed to model political and market choices. In biological terms, Dennett’s 1995 discussion of the Library of Babel (and its subset, the Library of Mendel) and Design Space is directly related, and extends Dawkins’s (1986) notion of animal space and the parameter space in Raup’s model of shell shapes (Dawkins 1996). 10 Explanatory relativity is in effect identical to Dyke’s (1988) concept of explanatory closure – it delimits the alternatives to be considered live options. 11 Compare Kuhn’s use of the phrase “disciplinary matrix” – the set of all contrast spaces of a discipline is the disciplinary metric. 12 See Prigogine and Stengers (1985) for a discussion of “bifurcation points”. 13 This approach was independently developed by van Fraassen (1970, 1980). 14 In the sense of non-atomic – not simple. 15 In broader contexts than under discussion here, a professional would be a cultural entity. 16 This does not preclude evolution of a non-Darwinian (‘Lamarckian’) kind. I think that non-Darwinian evolution would entail that changes are ‘capricious’ – that is, that they would be neutral with respect to selection. 17 After writing this paper, I became aware of a similar methodology, with different theoretical underpinnings, in Lumsden and Wilson 1981: pp. 26–30. 18 This averaging is a simplification. Useful information would be obtained by measuring the conceptual “spread” of a program or tradition over time, and also the size of the population. Another useful side-effect is that one would also be able to determine who is (roughly) a member of a tradition or not by seeing whether (i) their “karyotype” is a closely similar set of loci (showing if they are within the domain of that tradition), and (ii) whether they have sufficiently close values at each locus to resemble the overall profile of the tradition, whether they represent an offshoot, or whether they are members of a competing tradition. The measures may seem to be arbitrary, and no doubt there will be considerable argument over whether the scales used are appropriate, but that is a feature of historical research now in any event. I do not claim this proposal will solve all problems of historiography. 19 Mutatis mutandis, or sociologist, or anthropologist, or philosopher. 20 I understand that Kuhn and Feyerabend felt their view committed theory change to radical meaning change. However, I think that what is common between, say, late Newtonian and early relativity theories is structural rather than extensional, and on that basis think that incommensurability is not plausible. However, and it is too big an issue to deal with now, this position may be seen as requiring some account of similarity of meaning. The best I can do is point to Hull’s use of Kitcher’s approach in his 1988c. It is, of course, an evolutionary account. 21 There is an obvious analogy to a debate that forms part of the subject of Hull’s 1988c: the taxonomic debate between so-called ‘transformed cladists’ and other cladists (Scott-Ram 1990; Panchen 1992). The former think that using phylogenetic hypotheses in classification is viciously circular and question-begging, while the latter are prepared to make use of existing phylogenetic reconstructions to test and revise classifications. 22 Ironically, the terms ‘uniformitarianism’ and ‘catastrophism’ were coined by a philosopher, Whewell, in an 1832 review of Lyell’s Principles of Geology. Mayr (1982: p. 375) says that the terms were actually misleading, since the issue was between a steady-state world or a directionalist theory that included both catastrophism and progressionism.
502 23 “It is abundantly clear that Darwin rejected questions of ultimate reality as unanswerable” (Ghiselin 1969: p. 159). 24 Although it is to be feared that, like ‘junk DNA’ that is reactivated when a regulator gene is moved or inserted, theodicy could again become an issue in science, at least at the policy level. 25 A point made in discussion by Kim Sterelny, Neil Thomason and by an anonymous reviewer.
References Axelrod, R.: 1981, ‘The Emergence of Cooperation Among Egoists’, American Political Science Review 75. Axelrod, R.: 1984, The Evolution of Cooperation, Basic Books. Becker, P.H.: 1993, ‘Common Pitfalls in Published Grounded Theory Research’, Qualitative Health Research 32. Callebaut, W. and Pinxten, R. (eds.): 1987, Evolutionary Epistemology, A Multiparadigm Approach, D. Reidel. Collier, J.: 1988, ‘Supervenience and Reduction in Biological Hierarchies’, Canadian Journal of Philosophy Supplementary 14, 209–234. Dawkins, R.: 1977, The Selfish Gene, Oxford University Press (1989 edition). Dawkins, R.: 1982, The Extended Phenotype: The Long Reach of the Gene, Oxford University Press, revised 1989. Dawkins, R.: 1986, The Blind Watchmaker, Longman Scientific and Technical. Dawkins, R.: 1995, River Out of Eden, Weidenfeld and Nicolson. Dawkins, R.: 1996, Climbing Mount Improbable, Viking Press. Dennett, D.C.: 1995, Darwin’s Dangerous Idea: Evolution and the Meanings of Life, Simon & Schuster. Desmond, A. and Moore, J. : 1991, Darwin, Michael Joseph. Dyke, C.: 1988, The Evolutionary Dynamics of Complex Systems: A Study in Biosocial Complexity, Oxford University Press. Eldredge, N.: 1985, Time Frames: The Rethinking of Darwinian Evolution and the Theory of Punctuated Equilibria, Heinemann. Includes Eldredge and Gould 1972. Eldredge, N.: 1989, Macroevolutionary Dynamics: Species, Niches, and Adaptive Peaks, McGraw-Hill. Eldredge, N.: 1995, Reinventing Darwin: The Great Evolutionary Debate, Weidenfeld and Nicholson. Eldredge, N. and Gould, S.: 1972, ‘Punctuated Equilibria: An Alternative to Phyletic Gradualism’, in T. Schopf (ed.), Models in Paleobiology, Freeman, Cooper & Co. (Also in Eldredge 1985) Ellis, B.: 1979, Rational Belief Systems, Rowman & Littlefield. Ereshefsky, M.: 1991, ‘The Semantic Approach to Evolutionary Theory’, Biology and Philosophy 6: 59–80. Fisher, R.A.: 1930, The Genetical Theory of Natural Selection, Clarendon Press. van Fraassen, B.: 1970, ‘On the Extension of Beth’s Semantics in Physical Theories’, Philosophy of Science 37: 325–339. van Fraassen, B.: 1980, The Scientific Image, Clarendon. Garfinkel, A.: 1981, Forms of Explanation: Rethinking the Questions in Social Theory, Yale University Press. Ghiselin, M.T.: 1969, The Triumph of the Darwinian Method, 1984 University of Chicago Press reprint with new preface. Glaser, B.G.: 1978, Theoretical Sensitivity, Sociology Press.
503 Gould, S.J.: 1983, Hen’s Teeth and Horse’s Toes: Further Reflections in Natural History, Norton. Hamilton, W.D.: 1964, ‘The Genetical Theory of Social Behaviour: I and II’, Journal of Theoretical Biology 7: 1–52. Hardin, R.: 1971, ‘Collective Action as an Agreeable n-Prisoners’ Dilemma’, Behavioral Science 16. Hardin, R.: 1982, Collective Action, John Hopkins University Press. Heyes, C.: 1988, ‘Are Scientists Agents in Scientific Change?’, Biology and Philosophy 3: 194–199. Hull, D.L.: 1980, ‘Individuality and Selection’, Annual Review of Ecology and Systematics 11: 311–332. Hull, D.L.: 1988a, ‘A Mechanism and its Metaphysics: An Evolutionary Account of the Social and Conceptual Development of Science’, Biology and Philosophy 3: 125–155. Hull, D.L.: 1988b, ‘A Period of Development: A Response’, Biology and Philosophy 3: 241– 263. Hull, D.L.: 1988c, Science as a Process: An Evolutionary Account of the Social and Conceptual Development of Science, University of Chicago Press. Kuhn, T.S.: 1977a, ‘Second Thoughts on Paradigms’, in Suppe 1977. Also in Kuhn 1977b. Kuhn, T.S.: 1977b, The Essential Tension: Selected Studies In Scientific Tradition and Change, University of Chicago Press. Lloyd, E.A.: 1989, The Structure and Confirmation of Evolutionary Theory, Greenwood Press. Lumsden, C.J. and Wilson, E.O.: 1981, Genes, Mind, and Culture: The Coevolutionary Process, Harvard University Press. Maynard Smith, J.: 1983, Evolution and the Theory of Games, Cambridge University Press. Mayr, E.: 1982, The Growth of Biological Thought: Diversity, Evolution and Inheritance, The Belknap Press of Harvard University Press. Mosterin, J.: 1988, ‘Ontological Queries and Evolutionary Processes. Comments on Hull’, Biology and Philosophy 3: 204–209. Mun´evar, G.: 1988, ‘Hull, Straight Biology, and Straight Epistemology’, Biology and Philosophy 3: 209–214. Nitecki, M.H. (ed.): 1988, Evolutionary Progress, University of Chicago Press. Osgood, C., Suci, G. and Tannenbaum, P.: 1967, The Measurement of Meaning, University of Illinois Press. Oyama, S.: 1985, The Ontogeny of Information: Developmental Systems and Evolution, Cambridge University Press. Panchen, A.L.: 1992, Classification, Evolution, and the Nature of Biology, Cambridge University Press. Plotkin, H.C.: 1994, The Nature of Knowledge: Concerning Adaptations, Instinct and the Evolution of Intelligence, Penguin; also published as Darwin Machines and the Nature of Knowledge, Harvard University Press. Prigogine, I. and Stengers, I.: 1985, Order Out of Chaos: Man’s New Dialogue with Nature, Fontana. Quine, W.V.: 1969a, Ontological Relativity and Other Essays, Columbia University Press. Quine, W.V.: 1969b, ‘Epistemology Naturalized’, in Quine 1969a. Quine, W.V.: 1969c, ‘’Natural Kinds’, in Quine 1969a. Radnitzky, G.: 1982, ‘Analytic Philosophy as the Confrontation between Wittgensteinians and Popper’, Scientific Philosophy Today, Essays in Honor of Mario Bunge, Reidel. Radnitzky, G. and Bartley, W.W. III (eds.): 1987, Evolutionary Epistemology, Rationality, and the Sociology of Knowledge, Open Court. Richards, R.J.: 1987, Darwin and the Emergence of Evolutionary Theories of Mind and Behavior, University of Chicago Press. Richards, R.J.: 1992, The Meaning of Evolution: The Morphological Construction and Ideological Reconstruction of Darwin’s Theory, University of Chicago Press.
504 Ruse, M.: 1989, The Darwinian Paradigm: Essays on its History, Philosophy, and Religious Implications, Routledge. Scott-Ram, N.R.: 1990, Transformed Cladistics, Taxonomy, and Evolution, Cambridge University press. Simpson, G.: 1944, Tempo and Mode in Evolution, Columbia University Press. Smith, C.U.M.: 1988, ‘Send Reinforcements We’re Going to Advance’, Biology and Philosophy 3: 214–217. Sober, E.: 1981, ‘Revisability, A Priori Truth, and Evolution’, Australasian Journal of Philosophy 59: 86–91. Sober, E.: 1984, The Nature of Selection, MIT Press (1985 reprint with amendments). Sterelny, K.: 1995: ‘Understanding Life: Recent Work in Philosophy of Biology’, British Journal of the Philosophy of Science 46: 155–183. Strauss, A.L.: 1987, Qualitative Analysis for Social Scientists, Cambridge University Press. Suppe, F. (ed.): 1977, The Structure of Scientific Theories, second edition, University of Illinois Press. (Includes Kuhn’s ‘Second Thoughts’) Suppe, F.: 1977a, ‘Introduction’, to Suppe (ed.) 1977. Corrected version of the 1973 introduction to the first edition. Suppe, F.: 1977b, ‘Afterword – 1977’, in Suppe (ed.) 1977. Suppe, F.: 1989, The Semantic Conception of Theories and Scientific Realism, University of Illinois Press. Tennant, N.: 1988, ‘Theories, Concepts and Rationality in an Evolutionary Account of Science’, Biology and Philosophy 3: 224–231. Toulmin, S.: 1970, ‘Does the Distinction Between Normal and Revolutionary Science Hold Water?’, in Lakatos and Musgrave 1970. Tulloch, G.: 1972, Towards a Mathematics of Politics, University Michigan Press. Weber, B.H, Depew, D.J. and Smith, J.D.: 1988, Entropy, Information and Evolution: New Perspectives on Physical and Biological Evolution, MIT Press. Wiley, E.O.: 1988, ‘Entropy and Evolution’, in Weber and others 1988. Wiley, E.O., Siegel-Causey, D., Brooks, D.R. and Funk, V.A.: 1991, The Compleat Cladist: A Primer of Phylogenetic Procedures, The University of Kansas, Museum of Natural History, Special Publication No. 19. Williams, G.C.: 1966, Adaptation and Natural Selection: A Critique of Some Current Evolutionary Thought, Princeton University Press. Williams, G.C.: 1992, Natural Selection: Domains, Levels, and Challenges, Oxford University Press. Wittgenstein, L.: 1922, Tractatus Logico-Philosophicus, 1961 translation by D.F. Pears and B.F. McGuiness, Routledge & Kegan Paul.