By Greg Ransom

Written for Alexander Rosenberg and Larry Wright in partial fulfillment of the requirements for a Ph.D at UC-Riverside

ABSTRACT:   Carl Hempel and Thomas Kuhn suggest that the desiderata for theory choice shared by the members of a scientific community evolve with the advance of science.   Neither Hempel or Kuhn, however, provide a suitable mechanism for affecting that evolution.  This paper proposes a selective mechanism capable of underwriting the evolution of scientific desiderata for theory choice.  The proposed mechanism is termed membership selection which accounts for the evolution of group properties in terms of individual selection.  Using this new model, it is shown that Kuhn's own descriptive account of the scientific process provides a sufficient basis for a selective process in which the criteria for theory choice are selected for and against.  Changes in scientific desiderata need no longer go unexplained.  The evolution of the desiderata for theory choice can be accounted for as the product of a selective mechanism operating within the scientific process.     


KEY WORDS:  Scientific advance, theory choice, membership selection, desiderata, scientific community, shared values.

            Thomas Kuhn's descriptive account of science identifies a collection of desiderata which scientists have applied when evaluating scientific theories.  Broadly characterized, these desiderata have included accuracy, consistency, scope, simplicity, and fruitfulness, among others.  Kuhn makes two distinctions which allow for both the variability and the stages of change required for the operation of a selective mechanism.   First, Kuhn distinguishes between the uniquely expressed desiderata of each individual scientist and the generally shared desiderata of a scientific community.  Second, Kuhn distinguishes between periods of normal science when all scientists concur in their choice of a single over-arching theoretical paradigm and periods of revolutionary science when individual scientists will make conflicting choices about the conceptual framework in which to pursue scientific research.

            Let me elaborate somewhat on these related distinctions.  According to Kuhn's description of the scientific process, during periods of normal scientific research, a community of scientists broadly share all of the same general criteria for theory choice, including such standards for evaluation as accuracy, consistency, scope, etc.   However, each scientists within the community might weigh any one of these particular values more highly, or might order any number of them in one alternative rank or another.  That is, each scientist within the community might have his own unique weight for and individual configuration of criteria for theory choice, although all members within the community share the same basic set of desiderata.  I will call an individual's unique weighting for and configuration of desiderata an individual's desiderata array.

            These shared desiderata allow a scientific community to agree upon a single conceptual framework in which to conduct scientific research.  This conceptual framework consists of a disciplinary matrix or 'paradigm' made up of formal symbolic generalizations, preferred metaphors or models and exemplary problem solutions (Kuhn, 1974, p. 297).   Significantly, the conceptual framework of this normal science paradigm contains and produces only scientific theories which utilize commensurable theoretical terms.  In this way, a paradigm generates a linear context (Kuhn, 1970a, pp. 138-139) in which normal science puzzles and puzzle solutions are both grasped and solved.  Within this linear context of commensurable terms which characterize normal science, the shared desiderata of the community of scientists will provide a common standard sufficient to jointly determine for all members of the community when a particular normal science puzzle has been solved (Kuhn, 197Ob, p. 273; compare Kuhn, 1970a, p. 6; 1977, p. 273).

            However, during periods of revolutionary science, when the ongoing normal science paradigm begins to face intractable anomalies and alternative disciplinary matrices possessing incommensurable terms appear on the scene, there will no longer exist a linear context in which the same common theory choices will be made uniformly by all members of the scientific community.  Due to the variability of weighting and configuration between individual scientists within the shared list of desiderata held in common by the scientific community during the period of normal science, individual choices between rival paradigms during revolutionary science may differ among the various members of the scientific community.  In effect, the incommensurability of terms between rival paradigms creates a non-linear context in which the commonly shared criteria for theory choice held by the scientific community will no longer produce a linear algorithm which collectively determines a common conclusion within the scientific community about the adequacy of a scientific theory.  Rather, the non-linear context created by the incommensurability of paradigms offers an opportunity for the individual variability within the commonly shared list of criteria for theory choice to be expressed in alternative theory choices.

            We have here the basic building blocks required for a selective account of the evolution of the desiderata for theory choice.  What we have, first, is a source of diversity in the unique value configurations of individual scientists which is available as a target of selection and, second, a two-staged process which is available as the mechanism of that selection.  We have only now to show how that selective mechanism operates.  Fortunately, Kuhn has provided us in his descriptive account of science with the resources needed to fill out a selective model of the evolution of the desiderata of theory choice.  These resources include three primary elements.  First, a description of how values for theory choice are transmitted from one generation of scientists to another.  Second, a description of how scientists with a diversity in their configuration of desiderata for theory choice converge upon a single normal science paradigm after a period of scientific revolution.  And third, a description of how the transformation from one period of normal science to the next through a span of scientific revolution takes place as a consequence of the differential survival of only some but not all configurations of desiderata for theory choice.

            Although Kuhn has not recognized these selective elements for the true selective system which I believe they represent, he has consistently characterized the evolution of science as the product of a somewhat vaguely suggested selective process (see for example Kuhn, 1970a, p. 172).  Carl Hempel, on the other hand, has suggested that Kuhn's descriptive account of the scientific process might be supported by some sort of sociological functionalism (Hempel, 1979a, p. 300; Hempel, 1981, p. 402).   I will now show how the metaphorical allusions which Kuhn has used to suggest a selective process can be replaced by a formally modeled selective mechanism using only the descriptive elements provided in Kuhn's account of scientific change.  If my efforts are successful, then I will have shown how Kuhn's descriptive account of the scientific process can be supported without any need to appeal to the dubious assumptions of sociological functionalism.

            I will begin my account by introducing the concept of membership selection.   There is a common misconception that classical Darwinian natural selection, in which selection acts on individuals who compete directly against one another within a species, represents the only possible mechanism for modeling a selective system.  In fact, there are several different models representing various selective mechanisms in the biological world (Edelman, 1978; Piattelli-Palmarinin, 1986; and Darden and Cain, 1989).  We have clonal selection in immunology (Jerne, 1966), we have neural group selection in neuroscience (Edelman, 1987), and of course we also have sexual selection, group selection, and artificial selection in evolutionary biology which were recognized by Darwin (Ghiselin, 1974; Sober, 1984; Ruse, 1979) as alternatives to his theory of the natural selection of individuals.

            In classical natural selection there is selection over individuals for properties which are possessed only by individuals who compete directly with one another.  In classical group selection whole groups of individuals are selected over for a property of the group as a whole.  Significantly, as classical group selection is conventionally modeled, individual expression for that group property costs each individual adaptive advantage within the group (Williams, 1966; Sober, 1984).  By contrast, I propose to model a selective mechanism I will call membership selection.  As I will use the term, membership selection is a selective process which selects over individuals for a property of those individuals which either does or does not contribute to a group property, a property which cannot be exhibited alone by a single individual, but which can only be expressed as a group property.  Through this process which selects over individuals and for a property these individuals either do or do not contribute to the group, there will be selection for the group property which selected individuals exhibit.  

            The contrast between the conventional model of classical group selection and my model of membership selection should be clear.  In the process of membership selection, shared expression of a group property by the members within the group will cost them no selective disadvantage among themselves, but will contribute to each individuals selective advantage over individuals who do not participate in expressing a group property.  For example, in evolutionary biology, the circular herding of individual musk-oxen is a group property which gives each animal an adaptive advantage over other individual musk-oxen which do not herd in a circle.  Individual musk-oxen cannot express the property of herding in a circle alone as a single individual.  But when an individual's proclivity for herding in a circle is expressed in conjunction with other individuals who have inherited or acquired this same proclivity, that individual along with the other individuals who share that proclivity will be selected over those individuals who do not display this proclivity.

            It is crucial to recognize here that herding individuals will have an adaptive advantage over non-herding individuals as a result of the unique group properties of circular defense formations.  Herding individuals are more fit than non-herding individuals because the group property they produce through their own individual contributions -- an all horns out circular defense posture -- increases the survival chances of each against outside predators as compared to those individuals who face predators out alone with their back-side exposed.  The individuals who display this group property are not disadvantaged, however, vis-a-vis other members of the group.  This, of course, directly contrasts with the classical group selection model as it is conventionally modeled.

            Now that I have introduced the concept of membership selection, let me demonstrate how this mechanism operates within science, if Kuhn's  descriptions of scientific development are correct.   Recall that, according to Kuhn, a scientific community which shares a disciplinary matrix also shares a common set of scientific 'values' which includes such standards of judgment as accuracy, simplicity, scope, etc., which jointly commits the members of the community to the prevailing normal science paradigm.  Kuhn calls these common values a 'shared ideology' (Kuhn, 1970b:  241) because, although at a sufficiently high level of abstraction all of the members of the normal science community share the same basic set of criteria for theory choice, there remains within that common set a diversity between individuals in the unique expression that is given to those criteria by each individual.  That is, according to Kuhn, within this common set of desiderata each individual scientists belong to the community possesses his own unique weight for and configuration of criteria for theory choice.  For instance, one scientist might weigh simplicity more highly than another, or again, some other scientist might rank accuracy before scope whereas another might do just the opposite.

            However, at a sufficient level of abstraction, the aspects of judgment which determine individual theory choice are both similar enough and well or narrowly defined enough to unite the members behind a single paradigm, allowing them to cooperate in normal scientific research.  But not only does this 'shared ideology' commit the individual members of the scientific community to a single theoretical paradigm, it also provides them with an intersubjective standard which is sufficient for normal science puzzle solving.  Although the role these desiderata play in the recognition of normal science puzzle solutions remains somewhat unclear, Kuhn does suggest that the communities' common desiderata for theory choice act within a normal science paradigm as what he has called a 'shared algorithm' (Kuhn, 1970b, p. 241; Kuhn, 1977, pp. 329-331) which determines when a puzzle within that paradigm has been solved for all members of the scientific community.

            They do not, however, act as a shared algorithm for theory choice during periods of revolutionary science when individual members of the scientific community are choosing between rival paradigms at a time when the prevailing normal science paradigm is wracked by crisis due to the emergence of some intractable anomaly.  In fact, during periods of revolutionary science, when incommensurable theories vie to replace a crisis ridden normal science paradigm, Kuhn says that "two men fully committed to the same list of criteria for choice may nevertheless reach different conclusions."  (Kuhn, 1977, p. 324).  In other words, the common values which dictate individual theory choice, and at one time provide a shared algorithm for reaching unanimous theory choices within the normal science community, are no longer sufficient during periods of crisis science to dictate collective theory choices.

            According to Kuhn, new initiates into a normal science paradigm, i.e. non-scientists who are being trained within a disciplinary matrix, in large part acquire the commonly shared values of the community from "the study of examples of past applications rather than by learning rules about how they are to be applied"  (Kuhn, 1971, p. 146, compare Kuhn, 1970a, p. 10-11; 1974, p. 381-3-19; 1977:  327).  This primary method of acquisition allows for the persistence of relative diversity within the desiderata set of each individual within the community of scientists.

            Kuhn describes the transition from one normal science paradigm to the next as a time of revolution during which the once prevailing paradigm is exposed to rival paradigms after an initial period of crisis.  This rivalry between competing paradigms is short-lived in that a single paradigm soon emerges to replace its now moribund predecessor as the prevailing framework for normal scientific research (Kuhn, 1970a, p. 153; Kuhn, 1977, p. 291).  That is, on Kuhn's view, a period of revolutionary science quickly ends when one paradigm triumphs over its rivals through a process of convergence which results in the survival of a single paradigm. 

            Kuhn describes four steps by which a paradigm comes to prevail over its competitors during a period of revolutionary science.  First, a new paradigm is established by an initial group of supporters, who, although few in number, are successful in producing both theoretical and experimental results which increasingly satisfy any number of common desiderata for theory choice (Kuhn, 1977, p. 332; 1970a, p. 158-159). Second, individual scientists working within rival paradigms, including the once prevailing paradigm, are increasingly attracted to the success of the new paradigm as it increasingly matches up over time with the uniquely configured desiderata set of each individual (Kuhn, 1977, pp. 329, 332; Kuhn, 1970a, pp. 158-159).  Third, the new paradigm is able to attract fresh recruits form among the young and uneducated who wish to test themselves in the activity of normal science puzzle solving (Kuhn, 1970a, pp. 36, 151, 203; Kuhn, 1977, p. 290).  Fourth, scientists who remain attracted to rival paradigms, which continue to show no success at progressively fulfilling commonly held standards of evaluation or remain wracked by intractable anomalies, die out over a period of time (Kuhn, 1970a, p. 151).

            I will now show how the four steps of Kuhn's description of the transition from one paradigm to another through a period of revolution constitute membership selection for individual desiderata arrays.  The transition Kuhn describes is a process in which individual scientists are differentially attracted to rival paradigms according to the degree to which a paradigm makes a match to each individual's unique desiderata array.  As an individual's weighting for and configuration of desiderata are increasingly fulfilled, that individual becomes increasingly attracted to a particular paradigm.

            According to Kuhn, a single paradigm is able to attract most but not all members of the previous normal science research community.  A small number of the members of this once prevailing community remain committed to the old paradigm (or perhaps some other new paradigm) as a result of the individual variability in their own values for theory choice.  The number of those committed to these moribund paradigms dwindles to extinction as death takes its toll among the current population of scientists and the intractability of serious anomalies combined with a lack of success in normal puzzle solving makes recruitment among the young impossible.  By contrast, the new consensus paradigm which has had some success at normal science puzzle solving is able to recruit fresh recruits who soon become indoctrinated through tacit instruction with the same general set of desiderata priorities which are abstractly shared by others already working within the paradigm.  In other words, I am claiming that the differential survival of individual desiderata arrays is a product of differential selection affected through the mechanism of membership selection.

            In what manner does this constitute membership selection?  Recall that, at the conclusion of the process Kuhn describes, a single disciplinary matrix will have emerged alone as the prevailing paradigm which supports a community of scientists engaged in normal scientific research.  It can now easily be seen that in the transformation Kuhn describes from one paradigm to another through a period of scientific revolution there has been differential survival among divergent desiderata arrays.  Scientists displaying desiderata weightings and configurations which matched with and were thereby attracted to vanquished paradigms have been eliminated from the scientific community.  Scientists displaying desiderata arrays which matched with and were thereby attracted by the triumphant paradigm have been able to reproduce themselves in the desiderata arrays of tacitly trained new young recruits who come to share similar weighting for and configurations of standards for theory evaluation.  And of course, these generally similar desiderata provide the shared standards of evaluation which, according to Kuhn, makes normal science puzzle solving possible. 

            Note here first that the process Kuhn describes represents differential survival among individual desiderata arrays as a consequence of whether or not an individual either does or does not share certain general properties (within those arrays) with other member of a community.  Note here also that the general properties exhibited by a group in the shared desiderata arrays of individual members are crucial to the formation and success of that group.

            Why should this selective process be termed membership selection rather than Darwinian natural selection?  There are two important reasons.  First, because the differential selection of individual desiderata arrays leads to the reproduction of individual arrays exhibiting a general group property rather than the reproduction of individual arrays exhibiting the particular unique properties of any single individual.  Second, because the differential selection of individual desiderata is made possible only as a result of the formation of a group of scientists, all sharing common desiderata arrays at some level of abstraction, around a single paradigm.

            Let me elaborate somewhat on there two points.  The first point is a product of the means by which criteria for theory choice are learned by budding scientists.  Because new recruits learn their desiderata primarily through the study of examples rather than through explicitly articulated rules (Kuhn, 1971, p. 146.  Ronald Giere supports this interpretation, see Giere, 1988, p. 37) , each of their own unique weightings of and configurations for theory choice need only be in conformity with the general criteria which bind the community as a whole to the new paradigm rather than to rival paradigms.  Indeed, if we accept Kuhn's account of tacit value learning, the exact transfer of unique desiderata configuration from one individual to another should not be expected.  The second point results from the fact that the successful development of the paradigm, which is so crucial in the attraction of other scientists and new recruits, is in part dependent on a number of scientists -- all of whom share the same general criteria for theory choice -- working together to solve normal science puzzles.  Beyond the merely practical advantage of maximizing collective scientific output through shear numbers, this result follows from Kuhn's argument that, "No puzzle-solving enterprise can exist unless its practitioners share criteria which, for that group and for that time, determines when a particular puzzle has been solved" (Kuhn, 1977, p. 273, compare Kuhn, 1970b, p. 273).  In other words, the formation of a normal science puzzle solving tradition requires the common sharing of general desiderata as a group property exhibited through the individual desiderata arrays displayed by the member of that group.  It is the combined fact that a number of workers share a paradigm and a general pattern of desiderata which allows them to act successful working within that paradigm.  As Kuhn puts it, "The individual [scientist] can learn from his mistake only because the group whose practice embodies these rules [of puzzle solving] can isolate the individual's failure in applying them" (Kuhn, 1977, p. 278).

            To repeat, I call the process of the differential selection of individual desiderata arrays membership selection because of the fact that a common group property, i.e. shared general criteria for theory choice, is selected as a consequence of the process in which individual desiderata arrays are selected for and against because of whether or not they share a common property with other individual desiderata arrays, a commonly shared property which makes normal puzzle solving possible.  That is, individual desiderata arrays which match for, are attracted to, or, as we might say, are able to recognize (compare Edelman, 1974; 1975; 1978) the newly prevailing paradigm are selected for while the individual desiderata arrays which do not match for, are not attracted to, or, as we might say, are unable to recognize the newly prevailing paradigm are selected against.  Yet as we have seen, this process of selection proceeds only as a result of the fact that these individual desiderata do or do not share a common property with other individual desiderata which makes the success of a particular paradigm possible.

            The selection of the group property of possessing common general standards for judging the results of normal science puzzle solving is affected, then, through the differential selection of individual criteria which survive or don't survive as a result of whether or not they do in fact exhibit this general property which contributes to group success.  In other words, the selection of individuals and for the individual properties which individuals share with other individuals at a certain level of generality has produced selection for a group property which contributes to the success of the group.  This is the process which I have termed membership selection.

            Thus far I have shown how Kuhn has made use of a number of selectionist elements in his account of theory choice.  And I have shown how these elements provide a sufficient basis for an adaptive account of theory choice which is based on an actual selective mechanism and is not dependent on Kuhn's own unfulfilled Darwinian metaphor or the dubious sociological functionalism suggested by Carl Hempel.  To conclude this section I would now like to suggest that there is a deeper concern with Kuhn's account of science beyond the fact that he has given us an evolutionary account of science with no mechanism and only a metaphor to affect that evolution.

            Recall that, according to Kuhn, individual scientists make diverse and conflicting theory choices during periods of crisis within normal science.  That is to say, Kuhn sees individual theory choice during revolutionary science as a highly individualized affair in which each scientist has his own unique weight for and configuration of criteria for theory choice.  However, as we have seen, in the final analysis Kuhn wants to claim that theory choice is a 'community decision' which is only completed when a new normal science tradition has been fully re-established by community consensus (Kuhn, 1970a, pp. 145-164; 1971, pp. 145-146; 1977, p. 332).  How is it possible for the evolution of science to be driven both by the individual tastes of individual scientists and at the same time be determined by the character and structure of scientific communities?

             This apparent paradox raises something of a difficultly when one is attempting to give a proper characterization of Kuhn's account of the scientific process.  Kuhn has always claimed that there is a sociological base to his position (Kuhn; 1970b, pp. 241, 253).  This sociological base consists in taking the 'normal' scientific group as the unit of explanation rather than the 'normal' individual scientist (Kuhn, 1970b, pp. 240-1).  But just what Kuhn means by a 'community decision' regarding theory choice becomes highly problematic given his highly individualistic premises concerning the 'desiderata' of theory choice.  Indeed, the whole notion of a 'decision' seems paradigmatically to be the kind of things which an individuals makes, rather than something which a group makes.  Yet ultimately, what is most troubling about Kuhn's notion of 'community decision' is that it is nowhere precisely explained.  A successful model of scientific progress built upon Kuhn's account of the scientific process should be able to resolve this tension.

            To get a better flavor of this problem, let me quote from what Kuhn has to say about the character of his 'sociological explanation' of theory choice.  At one place Kuhn say, "The type of question I ask has therefore been:  how will a particular constellation of beliers, values, and imperatives affect group behavior?  My explanations follow from the answer" (Kuhn, 197b, p. 240).  In another he says, "Our concern will not then be with the arguments that in fact convert one or another individual, but rather with the sort of community that always sooner or later re-forms as a single group" (Kuhn, 1970a, p.153).   And yet again, in his Postscript--1969 Kuhn says "Communities .. are the units this book has presented as the producers and validators of scientific knowledge" (Kuhn, 1970a: 178).

            Kuhn's 'sociological explanation' of theory choice is most problematic to the point of seeming vacuity when he takes what I call a 'just stand back and watch' approach toward group choice (See for example Kuhn, 1970b, p. 237-8).  The idea is simply to wait for a consensus to appear within the scientific community.  But merely identifying the re-appearance of a consensus is no explanation for that consensus.  That is, this observation gives us no explanation which will account for why the diversity of individual theory choices at the time of crisis will eventually converge to a common choice at some later point.  Read uncharitably, it sounds as if Kuhn is saying that we can explain why a given group of scientists has selected one theory over another by pointing out that they have in fact done so, which, of course, is no explanation at all.  That is to say, viewed in simple enough terms Kuhn's account of theory choice can be made to sound vacuous.

            Sometimes Kuhn suggests that this consensus will appear when a new theory begins to 'bear fruit' -- when it begins to prove its promise by successfully developing and solving its own batch of puzzles.   It has been my contention in this paper that this must be Kuhn's underlying explanation for the formation of a group consensus or 'decision' on a new paradigm for scientific research.  Without it he is left with an unresolvable dilemma.  The dilemma Kuhn has set for himself is to account for both:  1) The initial diversity of theory choice among individuals at the onset of revolutionary science; and 2)  The eventual formation of a consensus of choice through time on one particular paradigm, using only his model of commonly held group 'values' (Kuhn, 1971, p. 146) for theory choice, which are uniquely and divergently expressed by each scientists.

            I am arguing, then, that if individuals do diverge during initial theory choice at the onset of a crisis within normal science,  and a proposed characterization of Kuhn does not offer some idea of what it is that will drive these divergent views toward a consensus in the choice of a theory, then we should have no expectation that these individuals will ever re-form as a group working within one particular paradigm or another.  I am arguing in effect that a proper account of Kuhn must provide some basis for the eventual re-convergence of the community of researchers upon a single disciplinary matrix in the face of a diversity of independent value configurations, configurations which allow individuals to make divergent theory choices and do not dictate the choice of any particular theory.

            In other words, a successful characterization of Kuhn's account of the scientific process must answer this question, "Why does the initial diversity of theory choices during a time of scientific crisis eventually collapse into a single consensus choice?"  It seems that the only basis for such a convergence would be the ability of a theory to increasingly fulfill its promise in terms solving puzzles, or meeting any of the other general criteria of choice held by the community.  Offering fulfillment of promise as the basis of convergence would meet the requirement that the collectively held although individually diverse 'values' should provide the only means for driving the community toward consensus.  It would do this by increasingly satisfying any number of the core criteria of the jointly shared 'values' of the group --  eventually fulfilling the individually diverse expression of these shared 'values' for each and every member of the group.   In this paper I have claimed that this is just what Kuhn is driving at when he speaks of a paradigm generating widespread conviction over times as it is developed (Kuhn, 1977, p. 332) and as it systematically winning over new converts one by one as differing criteria for theory choice become increasingly satisfied (Kuhn, 1970a, p. 158).

            More fundamentally, I have attempted to show that Kuhn's 'community choice' model of scientific change can be underwritten by an evolutionary mechanism I have called membership selection.  That is, I have suggested that the 'sociological base' of Kuhn's account of the scientific process can be grounded within a perfectly legitimate scientific framework which is modeled after some of the most successful and well established theories in science.


I would like to thank Larry Wright for conversations which contributed to the development of this paper. 

Postscript:  Only after this paper was written did I come across the important discussion of Arnold and Fristrup on the relations between group entities and individual fitness.  Anyone pursuing these issues should consult their paper, A. Arnold, and Fritstrup, K.:  1982.  ‘The Theory of Evolution by natural selection:  A hierarchical expansion’, _Paleobiology_, 8, 113-129.


Brandon, Robert and Burian, Richard, eds.: 1984, Genes, Organisms, Populations, MIT Press, Cambridge.
Darden, Lindley and Cain, Joseph.: 1989, 'Selection Type Theories', Phil. of Sci. 56, 106-129.
Darwin, Charles.: 1974, Metaphysics, Materialism & the Evolution of Mind, U. of Chicago Press, Chicago.
Edelman, Gerald,: 1974, 'The Problem of Molecular Recognition by a Selective System', in Ayala and Dobzhansky (eds.), Studies in the Philosophy of Biology, Macmillan, London.
Edelman, Gerald.: 1975, 'Molecular Recognition in the Immune and Nervous System', in Worden, Swazey, Adelman, (eds.), The Neurosciences:  Paths of Discovery  MIT Press, Cambridge.
Edelman, Gerald.: 1978, 'Group Selection and Phasic Reentrant Signaling:  A Theory of Higher Brain Function', in Edelman and Mountcastle, The Mindful Brain MIT Press, Cambridge.
Edelman, Gerald.: 1987, Neural Darwinism Basic Books, New York.
Ghiselin, Michael.: 1969, The Triumph of the Darwinian Method U. of Chicago Press, Chicago.
Ghiselin, Michael.: 1974, The Economy of Nature and the Evolution of Sex U. of California Press, Berkeley.
Giere, Ronald.: 1988, Explaining Science U. of Chicago Press, Chicago.
Hayek, Friedrich.: 1979, Law, Legislation and Liberty, Vol. III, The Political Order of a Free People U. of Chicago Press, Chicago.
Hempel, Carl.: 1979a, 'Scientific Rationality:  Normative vs. Descriptive Construals', in Berghel, Hubner, and Kohler, (eds.), Wittgenstein, The Vienna Circle and Critical Rationalism Holder-Pichler-Tempsky, Vienna.
Hempel, Carl.: 1981, 'Turns in the Evolution of the Problem of Induction', Synthese 46, 389-404.
Hempel, Carl.: 1983a, 'Kuhn and Salmon on Rationality and Theory Choice', J. of Phil. 80, 570-572.
Hempel, Carl.: 1983b, 'Valuation and Objectivity in Science', in Cohen and Lauden, (eds.), Philosophy and Psychoanalysis D. Reidel, Boston.
Hempel, Carl.: 1988, 'Limits of a Deductive Construal of the Function of Scientific Theories', in Ullmann-Margalit, (ed.), Science in Reflection, Kluwer Academic, Boston.
Hull, David.: 1988, Science as a Process U. of Chicago Press, Chicago.
Jerne, Niels.: 1966, 'The Natural Selection Theory of Antibody Formation:  Ten Years Later', in Cairns, Stent, and Watson, (eds.), Phage and the Origins of Molecular Biology Cold Spring Harbor Laboratory, Cold Spring Harbor.
Kuhn, Thomas.: 1962/1970a, The Structure of Scientific Revolutions U. of Chicago Press, Chicago.
Kuhn, Thomas.: 1970b, 'Reflections on my Critics', in Lakatos and Musgrave, (eds.), Criticism and the Growth of Knowledge U. of Cambridge Press, Cambridge.
Kuhn, Thomas.: 1971, 'Notes on Lakatos', in Buck and Cohen, (eds.), PSA 1970 D. Reidel, Boston.
Kuhn, Thomas.: 1974, 'Second Thoughts on Paradigms', in Kuhn, The Essential TensionU. of Chicago Press, Chicago.
Kuhn, Thomas.: 1977, 'Objectivity, Value Judgment, and Theory Choice', in Kuhn, The Essential Tension U. of Chicago Press, Chicago.
Kuhn, Thomas.: 1983, 'Rationality and Theory Choice', J. of Phil. 80, 563-570.
Penrose, Roger.: 1989, The Emperor's New Mind Oxford U. Press, Oxford.
Piattelli-Palmarini, Massimo.: 1986, 'The Rise of Selective Theories:  A Case Study and Some Lessons from Immunology', in Demopoulos and Marra, (eds.), Language Learning and Concept Acquisition Ablex, Norwood.
Ruse, Michael.: 1979, The Darwinian Revolutions U. of Chicago Press, Chicago.
Sober, Elliot.: 1984, The Nature of SelectionMIT Press, Cambridge.
Stone, Mark.: 1989, 'Chaos, Prediction and Laplacean Determinism', Am. Phil. Q. 26, 123-131.
Williams, George.: 1966, Adaptation and Natural Selection Princeton U. Press, Princeton.
Wilson, E. O.: 1975, Sociobiology Harvard U. Press, Cambridge.
Wynne-Edwards.: 1962, Animal Dispersion in Relation to Social Behavior Oliver & Boyd, Edinburgh.


See Kuhn, 1970a, p. 85; Kuhn, 1983; Hempel, 1981, 1983.  Compare Popper, 1965, p. ix. 

On the general nature of selective mechanisms, see Edelman, 1978; Piattelli-Palmarninin, 1986; and Darden and Cain, 1989.  See also the accounts of selection offered in Sober, 1984; and Hull, 1988.

  For Kuhn, criteria which are not shared across the community would be criteria "outside" science.  See Kuhn, 1977, p. 325.  Compare also Kuhn, 1970a, 152-153.

Elsewhere Kuhn includes the shared values of theory choice within his notion of a disciplinary matrix, see for example Kuhn, 1970a, p. 184-186.  However, for the purposes of this paper, I will limit the use of the term 'disciplinary matrix' to the three items listed in the text above.  I do this because it will become necessary to make a clear distinction between the nature and role of symbolic generalizations, model and exemplars, and nature and role of the shared criteria of theory choice.  Of course, Kuhn has himself emphasized the distinctive roles played by these two elements.  For example, Kuhn has written one essay concerned almost exlusively with the three items I am here calling a 'disciplinary matrix' (Kuhn, 1974) and another essay devoted almost exlusively to shared values (Kuhn, 1977).     

For more on the distinction between linear and non-linear patterns, see Stone, 1989; and Penrose, 1989.  Compare also Kuhn's discussion of algorithms for theory choice, e.g. Kuhn, 1970b, pp.  238-241; and  Kuhn, 1977, pp. 328-331.

Hempel has since repudiated this proposal, see Hempel, 1983, p. 95.            

This model of selection was inspired my own attempt to reformulate of Friedrich Hayek's model  of 'group' selection for liberal values (Hayek, 1979, pp. 155-176) into an individualistic framework.  It was also greatly influence by Edelman's characterization of the elements of a selective process (Edelman, 1978), which struck me at the time as deeply analogous to the elements described in Kuhn's account (Kuhn, 1977) of the scientific process.

See also the essays collected in Brandon and Burian, eds. on the unit of selection controversy.

For the distinction between selection of and selection for, see Sober, 1984.

This example is for illustrative purposes only.  It is not meant to capture the actual facts of evolution among current populations of musk oxen.  The herding behavior of musk oxen need not in fact have evolved as a result of a process of membership selection in order for this hypothetical example to show how a process of membership selection could be a viable mechanism of evolution.  E. O. Wilson's discussion of herding among the musk oxen (Wilson, 1975, p. 44-45) describes how females and the young are protected within the circular defenses of the bulls.  This could suggest that some form of kin selection might adequately explain this phenomena.  Wilson himself makes no such attempt.

The flying formations of birds may provide another example of membership selection which can be observed in the biological world.  Flying in a wedged formation is a group property exhibited by migratory geese.  Individual geese which lacked the instinct to fly in such a formation would be selected against because of the increased physical costs of flying to their destination alone.  Again, in this case, we have selection over individuals and for a property which is exhibited only by a group -- i.e. group flying formations which reduce wind drag.  The relative fitness of each individual goose is a product not only of its own properties and that of the external environment, but rather is also a product of group properties which the goose exhibits only when it shares a property which some but not all of its fellow geese.  This example may be complicated by the possibility that the non-lead geese in the group could be "free-riding" on the lead goose in the flying wedge.

Although the concept of membership selection is not to be found in Kuhn's work, in is worth mentioning that Kuhn has consistently been concerned with the nature of membership in a scientific community.  See, for example, Kuhn, 1970a:  169; 191; Kuhn, 1977:  291.

Kuhn emphasizes repeatedly that changes in paradigms represent a discovery of novelty, see for example, Kuhn, 1970a, p. 5, 24, 35, 64-65, 75, 96; compare Kuhn, 1970c, p. 288.  Divergent commitments therefore represent divergent recognitions of novelty.

In this regard it is perhaps significant that fruitfulness of new research findings is one of what Kuhn takes to be the core desiderata for theory choice.