Developing Scaffolds in Evolution, Culture, and Cognition

Darwin’s principle of inheritance has until recently received less attention than his principle of fitness, or his turn toward “population thinking” and the reality of variation, or the deduction of the operation of natural selection from Malthus. For most of the twentieth century, the classical theory of the Mendelian gene filled in for Darwin’s lack of knowledge of the material basis of heredity and it became taken for granted that genes, as articulated by molecular genetics, are the units of heredity. The rise of molecular developmental biology, evo–devo, epigenetic inheritance, and various challenges to the modern evolutionary synthesis renewed emphasis on the important conceptual role of development in relation to heredity. One route to a richly interconnected view of “heredity–development” (Maienschein 1987), after nearly a century of specialization, is to revisit with fresh eyes the history of biology prior to the theory of the gene, as historians are increasingly doing (e.g., Müller-Wille and Rheinberger 2007).

Rereading Mendel, one is impressed with how much his theory focused on “the development of hybrids” (Griesemer 2007). Pre-twentieth-century problems of “generation” (Churchill 1987), and of generations (Parnes 2007), never really went away, I believe; they were simply ignored, suppressed, or transformed for practical, heuristic reasons of what could be done with the tools, ideas, and funding available (Griesemer 2007). Whatever the nature of the “re”-turn toward development, it is important that a philosophical account of heredity integrate development. The account of reproduction articulated in this essay is in counterpoint to gene and replicator concepts of heredity which hold development at arm’s length. My account shows how development is entwined with heredity, making conceptual room for the ongoing empirical revolutions in mechanistic biology of the last thirty years. I frame units of heredity problems in terms of what I call “reproducers.” My aim in this essay is to explore the character and status of hybrids formed in reproduction processes as a means of exploring the phenomenon of “scaffolded” reproduction. I propose to put these ideas to work examining empirical practices of hybridization in the biological sciences and to explore possible applications to cognition and culture. I shall not survey these concepts or practices but rather dwell on a few instructive examples.

“Hybrid,” in biology, generally refers to concrete, material individuals produced from sources of several provenances and which have characteristics that hinder conceptualizing and tracking stable units of investigation. This meaning is concordant with some historical analyses of the problems of hybrids and hybridization in the nineteenth century (Olby 1985, 1997; Müller-Wille 2007; Parnes 2007). On my view of the phenomena, hybrid states and hybridization practices are ubiquitous, including not only products of various organism mating systems but also agricultural and experimental construction of hybrids through breeding as well as cell and molecular manipulation. Equally widespread is the practice of constructing hybrid units of investigation by physically marking or labeling material parts of objects of interest with bits of other kinds of material, such as radioactive isotopes or fluorescent stains, in order to track them (Griesemer 2007).

Since I view tracking as a nearly universal practice in empirical biology, hybridization and hybrids are bound to be ubiquitous. Since distributed cognitive and cultural systems involve formations, transformations, reconfigurations, and dissolutions of groups of people and things, they may exhibit patterns and mechanisms interestingly similar to or different from biological hybrids. I view development as integral to reproduction and reproduction as fundamental to evolution, so the formation, organization, development, and distribution of hybrids in biological, cognitive, and cultural processes is central to delimiting the domain of evolutionary processes. The argument of this chapter is that understanding the “Darwinian domain” can be substantially advanced by interpreting hybrids and hybridization in terms of the concept of scaffolded reproduction.

The chapter will first describe my reproducer perspective and the notion of scaffolded development and then will link the perspective to allied views on the repeated assembly of core social configurations (Caporael 1997, 2001, this volume) and generative entrenchment (Wimsatt 1986, 2001, this volume; Schank and Wimsatt 1988). With the perspective formulated, I will then use its tools to characterize the development of hybrids in very general terms: first by distinguishing material and formal modes of hybrid reproduction and then by considering the interlocking empirical and theoretical projects of tracking hybrids through life cycles. I argue that adequate characterization of hybrids requires attention to a variety of spatial and temporal scales and choice of “central subjects” for models and life-cycle narratives particular to each case; as Caporael (2001, 255) says, what is “part” and what is “whole” depends on “the researcher’s point of view.” Choosing a scale and central subject for models and narratives is to choose a perspective and to create a need for robustness analysis at the level of perspectives. Finally, I turn to several apparently “borderline” cases of reproduction to illustrate how the conceptual issues turn as much on questions of empirical methodology for tracking hybrids as they do on philosophical problems of “units” of change.

Darwin and Malthus presupposed that the organisms to which their principles apply are “reproducers”—“more-makers” that participate in fundamental ways in the production, “plurifaction” (Gould 2002, 611), or “multiplication” (Maynard Smith 1986) of new “individuals.” Both were concerned with how their numbers lead inevitably to a struggle for existence and relative reproductive success. Darwin was also concerned with “the laws of life,” though his efforts were limited by his access to details of cellular and subcellular mechanisms. More basically, although many kinds of causal process “give rise to” outcomes we might recognize as “new individuals,” not all of them seem to qualify per se as products of biological reproduction. Factories that mass produce cars or rifles repeatedly assemble “new individuals” in processes that, as circumscribed elements of broader social networks, seem merely analogous or remotely related to reproduction. To imagine that particular artifacts, such as cars, participate in the production of new artifacts of that kind, we must consider the entire societal context in which car factory workers need and drive cars to get to work or workers elsewhere in a connected economy supply resources, maintain infrastructure, or participate in markets necessary to the successful operation of the factory. At the other end of the size scale, “Planets do not have children,” Gould quipped (2002, 608), “and therefore cannot function as Darwinian individuals.” It is tempting to identify “reproducers” with “individuals satisfying Darwin’s principles,” but that would beg the question of how reproduction bears on the scope of the Darwinian domain. In order to determine whether that domain extends to cognition or culture, for example, we need independent criteria governing the applicability of Darwin’s principles. This was implicit in the replicator/interactor function approach to units of evolution (Dawkins 1976, 1982; Hull 1980, 1988). That approach lacked robustness, however, because Dawkins used modern genes as the model for replicators, which conflated informational properties of genes with basic presuppositions about reproducers in general (Griesemer 2005).

Darwin closely attended to the variety of modes of reproduction among organisms. Barnacles, orchids, hawkweeds, passionflowers, and other plants were central subjects for his studies of the evolution of sexual modes of reproduction. But what is reproduction? What features of this biological process help delimit Darwin’s domain of individual organisms in populations and which might help establish the applicability of evolutionary principles beyond organisms? My view is that biological reproduction is a kind of “more-making,” but not merely “plurifaction.” Reproduction is a special form of plurifaction, which I have called “special progeneration”: material parts of parents become, or are materially continuous with, parts of offspring that confer “developmental capacities” on offspring (Griesemer 2000a). The parent–offspring relationship is not merely one of resemblance but also one of material overlap: offspring units are made from organization-preserving physical parts, either directly from the parents or indirectly through chains of material overlap. Although physical interaction without transfer of parts can propagate some aspects of form, as in crystal growth, and crystal growth and breakage simulates a simple, limiting case of reproduction (Dawkins 1982), propagation of developmental capacities of any complexity involves material overlap (Penrose 1959). Like resemblance, the material overlap relation is reflexive, symmetrical, and intransitive, so its distinctive features are prone to being overlooked when heredity is described in terms of parent–offspring resemblance.

Successful developers are not only born organized but are also often born into environments that “scaffold” them in ways that use order in the environment to organize aspects of the developing system.¹ Scaffolding refers to facilitation of a process that would otherwise be more difficult or costly without it, and which tends to be temporary—an element of a maintenance-, growth-, development-, or construction process that fades away, is removed, or becomes “invisible” even if it remains structurally integral to the product.² These properties—propagated developmental organization and scaffolding context—jointly ensure (or raise the probability) that offspring have or acquire the capacity to develop. Development, in a broad sense, is a matter of exercising capacities leading to acquiring the capacity to progenerate in this special way. Maynard Smith (1986, chapter 2; 2000) wrote of this special way in terms of a transmission of genetic information, which is a compatible view so long as genetic information is understood to confer, through embodiment in a mechanism, a certain capacity to develop on the offspring. Otherwise, information would be merely a sort of “archive” of parental “data” rather than a causal mechanism for the production of a new individual. Material overlap can increase the robustness and reliability of transmission of capacities, compared to reliance on an unstable and uncertain environment to deliver components in suitable temporal order and spatial configuration, because complex organization can be preserved and propagated in material propagules. In Simon’s “artificial sciences,” designed artifacts like watches are made by watchmakers who organize production environments in such a way that artifact parts can be reliably and repeatedly assembled without the parts’ carrying the developmental capacities characteristic of biological reproduction (Simon 1981). The propagated material self-organization of developing systems also provides an opportunity for development to evolve self-scaffolding, as relations with environments that scaffold development externally can become internalized in developmental relations among parts formed from hybridizations of system and environment.

A distinctive feature of my account of reproduction is that the developmental capacities are delivered to offspring by “material overlap”—by material continuity of organized material parts transferred from, or delivered into scaffolding contexts by, parents to offspring—rather than by “impression” of form or information on uninformed recipient matter.³ Eukaryotic chromosomes are a good example: a complex of DNA and histone proteins supercoiled in a highly complicated, organized manner, in the right inherited cellular environment, can serve as a key component of a generative process in an offspring cell. New chromosomes are not formed by impressing their structure on exogenous material individuals but rather by order-propagating division (progeneration). The relation of chromosome to nucleotides is neither that of sculptor nor sculpture to clay. Instead, chromosomes are machine parts which include templates for the assembly by recruitment of material (nucleotides, proteins) and incorporation as new chromosome parts (DNA or RNA strands).

Paley’s (1821) miracle of living things is that organisms are like factories that make more factories, not like simple human artifacts made by humans, even artifacts that influence the production of more artifacts. Stamping a design or structure on unformed material is insufficient. Genes organized in chromosomes represent a highly evolved grade of organization, in which developmental capacities are conferred by means of an evolved, entrenched “coding” relationship between nucleotides and amino acids. Although genetic coding is highly relevant to the developmental capacities of offspring in modern taxa, it is not the only mode of developmental specificity (Newman and Bhat 2009; Newman, this volume) nor the only means of robust transfer of developmental capacity—after all, a cellular environment must be transferred along with chromosomes for an offspring to develop or even for genes to be expressed, so cell division must be just as reliable in conferring a full complement of enzymes, membrane, and other cell constituents as mitosis and meiosis must be in delivering a full complement of genes (Jablonka and Lamb 1995, 2005).

The reproducer perspective supports a general account of units that recur in genealogical relationships generation to generation.⁴ Genes are but one example. The perspective draws upon and contributes to multilevel selection theory in part by providing a framework for interpreting evolutionary transitions as transitions to new levels of reproduction (Szathmáry and Maynard Smith 1997; Griesemer 2000b).

Most “ordinary” (i.e., familiar) cases of biological reproduction clearly involve material overlap. Humans reproduce by means of gametes: haploid cells of male and female parents—sperm and eggs—fuse to form offspring zygotes. There is material overlap between parent and gamete, gamete and zygote. Zygotes divide to form offspring cells made from zygote parts. Successive cell divisions produce material overlap relations between parent and offspring cells in each subsequent cell generation. Gametes overlap both parent and offspring. Gametes confer developmental capacities on zygotes by propagating highly organized parts rather than mere raw materials.⁵ Of course, a zygote in the wrong circumstances (such as one aborted by the mother or eaten by a predator) cannot exercise its developmental capacities properly or fully. The same holds for infants, children, and adult humans. Moreover, the developmental capacities of humans, and of most familiar reproducers, are organized so that development proceeds through a series of stages: the exercise of one set of capacities realizes another set of developmental capacities, and another, and so on until an event of “special progeneration” occurs: the spatial separation of a “propagule” or material part carrying developmental potential to (or as) a new individual organism.

It is also familiar that genomic DNA replicates semi-conservatively. Semi-conservation is a phenomenon of material overlap at the molecular level. Strands separate, serve as templates for the synthesis of new strands from free nucleotides, and form new double-helix molecules containing one strand from the parent molecule and one new strand. The resulting double helixes are called “daughters”: half their material and nearly all their organization is parental in origin and nature (Meselson and Stahl 1958). Although genomic DNA replication might have been fully conservative (Meselson and Stahl 1958), in which case there would not be material overlap of strands between parent and daughter molecules (Godfrey-Smith 2009), it isn’t. If we were to treat material overlap as definitional for reproduction, then discovering whether DNA replication does or doesn’t involve material overlap would be uninteresting: just a matter of whether we (I) classify it as reproduction or not. What makes it empirically and theoretically interesting is the empirical link to claims of reliability and robustness (Griesemer 2000a). If material overlap is an efficient and effective way of propagating and producing developmental order and organization, then it should be favored, entrenched, conserved in evolution or else its absence should require special explanation. When are environments reliable enough to deliver developmental capacities in the form of nutrition (material, energy), scaffolding (physical interaction through hybrid states to facilitate development), and prosthetics (organized parts that enhance or substitute for developed ones, e.g., hermit crab houses, nest sites, knowledge recorded in books, scientific instruments that enhance perception), so that developing systems may forego making or managing those processes on their own?

The biological literature is full of examples of life cycles more complex and varied than the ones mentioned above (Calow 1978; Bell 1982; Buss 1987). Single-celled life forms divide in two or bud small parts directly from the body of the parent to form offspring. Others alternate generations in such a way that offspring resemble parents very little but resemble their grandparents much more. Pea aphids develop inside mothers who may in turn be developing at the same time inside their mothers. Inside the cells of the aphids are developing communities of endosymbiotic bacteria (Moran 1992). In gametophytic plants, much of the life cycle is spent in a haploid cellular state, due to the ability of the haploid cells to multiply by mitosis rather than first fusing with another haploid cell to form a diploid zygote, while the diploid state that we usually associate with the “mature” organism (on analogy with humans) is reduced to a very temporary “sporophytic” phase. And so on. Godfrey-Smith (2009), focusing just on multicellular forms, calls it a “menagerie.”

In every one of these cases, there is a special form of material continuity between parents and offspring established either by direct material overlap or by a chain of materially overlapping steps and in which some of the overlapping parts confer the capacity to develop. In some “borderline” cases, however, it appears the chain of this kind of material continuity is broken at certain points in the life cycle and that only “information” flows across the break points. Retroviruses and prions, for example, seem to propagate “structure” or “information” but seem not to exhibit material overlap relations between parent and offspring (Godfrey-Smith 2009, chapter 4).

Systems of communication among cognizing organisms, for example, people using language to talk to one another, would appear not to have the structure of reproduction processes since linguistic or visual communication does not proceed by transfer of material parts between senders and recipients but rather by a sort of “action at a distance.”⁶ Xerographic photocopying seems to transmit information without material overlap of input and output objects. (I get my “original” back from the machine unaltered along with the copy.) In some cases, of course, communication is by material transfer, as in chemical communication by pheromones, though the material seems to function as an informational trigger rather than an organized developmental propagule. Systems of cultural transmission, likewise, would appear not to have the structure of reproduction processes since such systems are often built on communication systems, as when students learn socially, through visual and oral communication with a teacher, or by imitation on the part of the student from visual and aural cues from the teacher. In other cases, there is a material transfer, as in delivering a posted letter to one’s mailbox—at least, it is material transfer if possession of the letter counts. When I read it, the words don’t jump off the page and into my eyes, so no material overlap there, unless those photons bounced off the paper first count as part of the paper and then part of my eyes. In a sense, the “reproduction,” propagation, or transmission of information in these cases would appear to be through a different “channel” or by different, formal, means than the material channel of reproduction in ordinary biological cases (see Godfrey-Smith 2009, 79). An important motivation of the reproducer perspective was to explore differences between material and formal modes of “reproduction” in detail because sorting it out may provide valuable methodological lessons, heuristics, and tracking strategies for studying phenomena of cognitive or cultural, as well as biological, development.

Reproduction involves the exercise of developmental capacities in a suitable context, or differently put, developmental systems only carry capacities in suitable contexts. In most cases, development involves “outside help” in the form of scaffolds that establish conditions suitable for material organizations to acquire, embody, and realize developmental capacities. In many cases, a developing system is vulnerable in ways that “mature” systems are not (e.g., lacking a protective shell while molting or lacking an attentive predator response behavior while developing a nervous system), so scaffolding tends to facilitate development by lowering the cost or easing the task in ways developmentally appropriate to the system’s abilities (Bickhard 1992). Scaffolds can be artifacts, infrastructure, or other (biological, cognitive, or cultural) agents that temporarily facilitate development of the system (Wimsatt and Griesemer 2007). But what makes the difference between a material entity: (a) providing material “nutrition” for a developing system, (b) scaffolding a system’s development, and (c) serving as a parent, in the material overlap sense, of the system?⁷

It would appear that a scaffold per se does not contribute material parts to the developing system, so it cannot count as food and does not count as a parent, but only counts as a facilitating or catalyzing part of the system’s environment.⁸ That view depends, however, on where we place the system–environment boundary and how we narrate the changes in the system over its life trajectory. As I will argue, hybrids are forms that blur canonical system–environment boundaries and distinctions.⁹ The variety of modes of development of hybrids apparent from natural history blurs the lines in enough ways to complicate the characterization of scaffolding and its distinction from food and parents. In a sense, interactions that facilitate development always involve the formation of “hybrid” objects—physical systems incorporating interactants as parts of different provenance such as environmental factors vs. parents. Mother and infant form a dyad (Caporael 1997, 2007, this volume). The dyad is a hybrid object in the sense that the constituents have different provenance. Mothers and infants each scaffold the other (Caporael, this volume). Five members of a hunting party drawn from each of the families of a small village form a task or work group (Caporael op. cit.). The task group is also a hybrid object. The painter’s scaffold and the bank building “fuse” in a way that is relevant to their scaffolding interaction to form a hybrid “building” while the painter is painting. The scaffold becomes a prosthetic part of the building, allowing the painter to stand on the side of the hybrid building in ways that would have been more difficult otherwise (for human painters though not for spidermen). Depending on the extent to which we recognize hybridizations of such kinds as delimiting generations of hybrid individuals rather than mere steps in a sequence of transitional phases, hybrids are new individuals—offspring in a new generation ¹⁰—parented by the scaffold and the building, or by the scaffold, its assemblers and users, and the building (perhaps plus the building’s assemblers and users). And the hybrids sometimes beget the autonomous individuals we imagined entered the hybrid: mother and child emerge from their interactions, though each is developmentally transformed by them; the hunters go back to their families; painter’s scaffold and bank building each emerge from their interactions as well. A key point of contrast between hybrids produced by assembly and those produced by reproduction is that disassembly tends to yield the same (even if transformed) collections of individual parts that were assembled while progeneration rarely or never yields the same individuals that progenerated to produce the developing offspring.

Unlike the fusion of two gametes to form a zygote, in familiar examples of scaffolding we tend not to treat hybrids as individuals of a new “generation.” That difference in status marks practical conventions more than obligatory ontological distinctions, though the theoretical import is just as great. We also tend not to think of gametes as individuals, though obviously they are as much individuals (and cells) as is their zygotic fusion. The painter’s scaffold temporarily put up alongside a building doesn’t typically lead us to talk about the scaffold + building as a new individual for at least four reasons: (1) the scaffold participates in the “hybrid” for a very short time compared to the duration of the building,¹¹ (2) the scaffold + painter does not alter the structure of the building very much, nor the building the scaffold very much, (3) the alterations that do occur tend not to be very significant in conferring or altering the developmental capacities of the recipient, and (4) we classify the scaffold and building as kinds with different salience or valence in our narrative accounts of the painting process; since the hybrid is neither of these, it falls outside the classification scheme of painters, scaffolds, and buildings. Each of these reasons, however, is rather fragile and can fail, even in the case of a painter’s scaffolding of a bank building, if we view the building or the hybrid system as developing. If the new paint causes the bank to give a fresh impression that draws in many new customers, painting may facilitate all sorts of “developmental” changes in the bank building, such as the addition of a new wing, allowing the bank organization to grow and “bud” new branch offices that take up residence in different buildings. If the rough surface of the brick rubs through the metal joints of the scaffolding during high winds, causing its collapse and the death of the painter, the mishap may drive the painting company out of business due to the ensuing lawsuit. Clearly, the construction (as opposed to the painter’s) scaffold has everything to do with the assembly of a building in fundamental ways.

In other cases, the system interacts with features of the environment that persist on much longer time scales than the system, such as the sun interacting with organisms of comparatively shorter duration.¹² In both kinds of cases, the time scale in our descriptions of the durations of the parts and their interactions govern whether we view the interactants as forming a hybrid system or as a system–environment interaction.

In still other cases, system and scaffolding components dissipate on the time scale of single developmental stages (i.e., durations in which a set of developmental capacities are exercised so as to acquire a next set of developmental capacities in a sequence resulting in special progeneration). The description of material elements interacting and persisting on the same time scale as food, scaffolding, or parent seems more fluid and open to alternative choices and representational salience, depending on the research problem. As I will discuss below, in the case of HIV-1 retrovirus reproduction, the salient material parts of the retrovirus and host cell turn over on similar time scales, leading to challenging questions about the nature and status of hybrids as individuals formed from virus parts plus host parts.

Descriptions of processes in time take the form of narratives or chronologies. Process explanations often narrate the operations of mechanisms (Griesemer 2011a, b). In order to construct a narrative, we must choose “central subjects,” the subjects of which a narrative is the biography (Hull 1975). The choice of central subjects to represent in a narrative governs whether interactions in a hybrid count as system–environment interactions or internal “intra-actions” (Barad 2007) of parts of the hybrid system. In life-cycle narratives, we tend, moreover, to choose central subjects in such a way that we can classify them and their “parents” (in the material cause rather than scaffolding sense) as “of the same kind,” which usually entails a particular temporal scale of the representation of steps or stages in the life cycle. In the painter’s scaffolding case, scaffolding and building must be on a par in the classification scheme in order to pick out each as a “parent” and the “hybrid” as an offspring “of the same kind.” It is not hard to imagine a narrative, choice of central subject, temporal scale, and classification scheme in which the painter’s scaffolding case fits: a mechanical engineer or architect might well find each “parent” and the hybrid “offspring” of similar architectural interest, for example, on the time scale of the assembly, use, and disassembly of the scaffold, while the manager of the bank whose branch occupies a building and the manager of the painting company which owns a scaffold may each see the other’s structure as “environment” on a different time scale from their “system.” A tourist can find the hybrid system consisting of construction scaffolding and church building called “Sagrada Familia” in Barcelona, Spain, quite as fascinating as Gaudi’s masterpiece itself (see figure 1.1).

Figure 1.1

In short, evaluating the role of developmental scaffolding in reproduction depends on which material interactions count as “inside” the system and which ones cross system–environment boundaries. The difference turns, in part, on choices of which biological phenomena to track and which are taken to be salient to the projects, questions, and representations of the researchers describing the case, as well as the character, scale, and scope of the interactions of the physical systems involved.

The editors (Introduction, this volume) pointed to trait and demic (breeding group) models as a framework for articulating the reproducer perspective with Caporael’s notion of the repeated assembly of core configurations. In the next section, I discuss task and trait groups in order to compare their assembly, operation, and dissolution as hybrid individuals to breeding groups. Caporael argues that core configurations are modal group sizes that serve different kinds of activities or functions. I treat core configurations heuristically: assuming they are modal sizes/activity structures of hybrids in order to track “hard cases” of processes in biology, distributed cognition, and culture that may (or may not) turn out to have the structure of reproduction processes, as I have described them. The heuristic amounts to asking what about these modal sizes tends to support the propagation and realization of developmental capacities. I also consider Wimsatt’s (1986, 2001) view that nontrivial evolutionary processes involve developmental systems whose parts show differential generative entrenchment. Thus, the model is that reproduction supporting accumulation of adaptations leading to complex developmental organization involves material overlap of generatively entrenched parts that carry and confer developmental capacities into hybrids which typically find themselves scaffolded in suitable environmental circumstances. The entrenched features scaffold others within the hybrid (self-scaffolding) insofar as developmental dependency is itself a form of scaffolding relation: downstream features are easier to produce, relative to those same construction processes absent the entrenched part, because of entrenched upstream features.

Differences between repeated assembly and reproduction point to distinct dynamics. A craftsperson who repeatedly assembles artifacts may be an artifact’s way of making another artifact, but the causal mechanism and relations may be different in character than the ways in which reproducers make more reproducers. If assembly processes such as putting (solid-state) parts of a rifle together in a factory are not followed by a developmental process of realizing the capacity to reproduce, then assembly processes will have different dynamics than reproduction processes such as gamete cell fusions or assemblies that are followed by developmental processes leading to organism reproduction. These differences might be used to suggest empirical tracking strategies and tests of core configurations hypotheses by assuming that what is “core” turns on the responsive developmental character of the interacting parts and not only on those configurations that are repeatedly assembled. Core configurations, in other words, may have properties combining aspects of biological reproduction and assembly processes.

The theory that generative entrenchment is typical of nontrivial adaptive evolution suggests that to extend the Darwinian domain beyond biology, we should look for the propagation of developmental capacities not only in reproduction or reproduction-like processes but in the hierarchical organization of developmental dependencies. Hierarchical organization of this developmental sort is an indicator of self-scaffolding that should be harder to propagate through “environmental channels,” and there should be better and worse choices of scales of description of developmental mechanisms so as to pick out the salient capacities transferred in reproduction.

Individuals play central roles in reproduction and development processes, but they are also units of investigation. I will explore the linkage in terms of the role individuals play as subjects of causal narratives. Individuality is the ontological category of Darwinian units, but also the epistemic category of what biologists track and represent in their empirical work. I view “individual” as a technical term depending on empirical correlations that investigators and their research subjects codetermine. There is no general “solution” to the problem of individuality, either by discovering necessary and sufficient conditions that cover empirical cases of interest or by metaphysical fiat. Investigator interests vary in ways that make differently salient whatever conditions or properties the subjects contribute to different research projects, in the context of different specialties, for different audiences. Whether we rely on criteria at Gould’s (2002, 602) “vernacular” or Ghiselin’s (1997, 49) “metaphysical” end of the spectrum, criteria of individuality must support the empirical project of tracking organized parts through reproductive life cycles or end up on the trash heap of armchair metaphysics. In the case of genomic DNA replication, the entanglement of epistemic and ontological aspects are clarified if we attend to the choice of starting points for narrating double-helix life cycles and to the scale or “grain” of the narrative description. Different choices lead to different views of the status of hybrids: as mere transients or as salient individuals. These differences come to the fore if we focus on tracking material parts through the development of the hybrids rather than trying to read ontology off the theoretical models scientists offer.

Narrations of DNA replication usually begin with a description of structure at a time and then describe state changes until the starting state recurs with two individuals instantiating the state in place of a single initial one. The trajectory of replication, in other words, can be described in terms of a cycle of states. However, in a cycle, each state recurs, so any state could be chosen as the starting point of the narrative, treating the other states as intermediates. We usually think of DNA as a double helix because that is the structure we identify with the functional genes of the genomes in organisms. That is a misleading characterization of “the” structure of functional DNA, however, because DNA has multiple functions within living cells, most of which require that the structure change. Serving as a template for RNA transcription requires single-stranded DNA. Serving as a vehicle for carrying genetic (sequence) information from parent to offspring eukaryotes requires that DNA be double stranded, bound to histone proteins and supercoiled in chromosomes that can be reliably segregated by microtubules in cell division.

To consider DNA to have a double-helix structure is to think of it at a snapshot in its replication or expression/transcription cycle. During replication, its structure is more complex than a simple double helix since parts are unwound into single-stranded condition and enzymes bond new nucleotides to the growing strands. Meselson and Stahl’s schematic diagram (1958, figure 6) of semi-conservative DNA replication only showed images of “mature” or “developed” molecules in each generation or cycle of replication: original parent molecule, first-generation daughter molecules, second-generation daughter molecules. That is, they depict a single stage in the replication cycle to which DNA structure returns each generation. If we were to pick a state intermediate in each of Meselson and Stahl’s generations, for example replication forks, we would say that DNA has a “Y” structure and passes through a simpler double-stranded transient stage on the way to a new replication fork.

From the point of view of narrating the replication process, three points are relevant to my argument about hybrids. First, some starting points make for stories that are more or less salient to the speculations, arguments, or theories we want to make and also to the theoretical guidance we are likely to receive for empirical inquiry, because attention is drawn to the “central subjects” (Hull 1975) of the stories. We rarely narrate cyclical processes by picking as central subjects the most ephemeral states of the process, even though we could, since the process cycles. If our question is about the status of hybrids, however, then we ought to make hybrids our central subjects. The recognition that generations are not mere durations of time, but rather actors in biological processes bearing causal relations to one another was a critical turn in the history of biology because it helped focus on hybrids as central subjects in breeding and crossing histories (see Parnes 2007).

Second, choice of narrative starting point goes hand in hand with choice of grain or scale for the story. At one grain, subjects may look so transient that we hardly notice them; at others, the same subjects may be so large as to hardly fit in the frame, crowding everything else out. The story of DNA replication has everything to do with how replicating forks are formed and resolved, but if all we care about is how the relation between strands of a parent double helix compares to the relation between strands of an offspring double helix, our narrative might fade out where the parent strands start to separate and fade in where the offspring strands are just finishing zipping up. Why bother narrating DNA “zipping up” if replication mechanisms aren’t the focus of our narrative interest? It may be that the classical tradition of interpreting Darwin’s principles can work at fairly coarse grain, omitting details of hereditary mechanisms. However, if heredity doesn’t work the same way in some other domain, for example, the domain of cultural “reproducers,” or retroviruses, then the causal narrative needs a different grain so it can represent, display, and compare those scenes.

Third, choices of units of inquiry structure commitments to notions of “individuality.” Choice of DNA double helix as narrative starting point and coarse graining of the replication cycle obscures the individuality of molecular hybrids as superficial and transitory. The problem is that these perspectival choices guide and constrain the sorts of theoretical commitments that can readily be made in the conduct of empirical inquiry. Godfrey-Smith (2009, chapter 4) argues that to cover all cases, one must accept that some reproduction processes are formal rather than material. I accept his conceptual contrast of formal and material, but not his reasoning about which are the material cases.

My project is heuristic and empirical. I do think material overlap is basic to all cases of reproduction, so if there is less of it in the story of some central subjects than others, then my heuristic proposal is to look elsewhere for the material relations leading to adequate explanation of the developmental organization we observe or ask how subjects in some cases get away with so little. Perhaps substantial material organization or order is to be found in environmental scaffolding rather than “in” the system. In some cases, reproduction of a specified central subject that looks formal by Godfrey-Smith’s lights will turn out to be scaffolded reproduction of that subject. But in other, borderline cases, the central subject may be best described as a hybrid individual that self-scaffolds. Simple ontological choices—formal versus material—do no real work for either theoretical or empirical evaluation of cases. Our goal is to identify and understand distinctive similarities and differences in the dynamical behavior of reproduction processes. I propose that we reframe the seemingly ontological question, exploiting some of the epistemic considerations I have suggested, to focus on how the relevant theoretical concepts derive from empirical tracking commitments and practices.

A key challenge in any kind of empirical or theoretical work is to pay attention (Griesemer 2007). Engagement and attention are required in order to observe, intervene in, or understand phenomena on relevant time and size scales. The same holds for tracking symbolic expressions in equation- or computer code, mathematical derivations, or computer simulations. Science isn’t easy because the world around us does not come prepackaged in objects, properties, events, processes, or activities of shapes, sizes, or durations to which we can continuously attend and engage, or to the extent we often need in order to gain understanding and control. A critical problem for scientists is to align ontologies of units with units of investigation so that models and theories can not only represent but also guide inquiry. A practical way to do that is to grow the models and theories out of empirical engagements, rather than impose them from who knows where, so that tracking success will (hopefully) align with theoretical adequacy (Griesemer 2012). Put differently, empirical work should scaffold theoretical work (as Schank and colleagues, this volume, argue), not the other way around.

To follow a process continuously is usually impossible (see also Griesemer submitted). At the absolute limits of visual attention and stamina, you have to blink or your eyeballs will dry out. Instead, scientists learn to develop ways to mark processes in order to track them with less than continuous, full, personal engagement. Radioactive tracers, fluorescent stains, genetic markers, and tissue transplants all facilitate tracking biological processes and determining how physiological, molecular, and genetic outcomes result from known inputs. Marks can be “noticings” of distinctive features that persist in subjects in stable relations to other features, to be noticed again at a later time or place, allowing an inference of continuity between observed phases or stages without continuous engagement. Marks can also result from physical interventions that cause changes to features of subjects so as to be more “noticeable” or detectable to the intermittent observer or intermittently observant. Many of the devices, schemes, strategies, and technologies of science are designed for tracking work to facilitate empirical and theoretical engagement without requiring continuous attention.

Marking is a physical correspondent to the graining of descriptions, and a key reason that some grains or resolutions “work better” for some causal narratives than others is that the grain of description matches or complements the character and scale of the marking interactions. An important means by which grain of description and causal narrative become linked is that the physical marking or detecting used to track processes yields traces which can be collected as data and which also serve to initiate representations of the phenomena constructed through these marking interventions.

Tracking requires commitment, and commitments have implications for “the way things are,” which is to say that they are ontological commitments. To track is to make an ontological commitment—a pragmatic ontological commitment. Tracking is a subtle art, like dancing. You can’t dance well if all you do is literally follow your partner. You have to anticipate where your partner is likely to go in order to coordinate your movements with your partner’s and arrive at the right place at the right time. When scientists track, they dance with their subjects and coordinately generate phenomena. Failure to anticipate where your subject or partner is going leads to failure to direct your attention, your body, or your instruments sufficiently quickly, deftly, or appropriately to keep up with the action. The idea that ontological commitment is pragmatic is that it is manifest in tracking activity. Just as causal narratives call for choice of a starting point and grain of description, empirical inquiry requires commitment to a grain through the marking protocols needed for tracking, which in turn govern what “intermediaries” can be recognized along the way as individuals suited to serve as units of investigation or dismissed as mere transients. Which brings me to the problem of marking and tracking hybrids.

Typically, tracking an individual of interest means labeling a part and hoping the part retains its parthood relationship to the whole, so that tracking the part results in tracking the whole. The FedEx bar coded label goes on one flap of the shipping box, but if that flap gets torn off, you aren’t tracking the package anymore. In tracking reproduction processes, if there are fission or fusion events, tracking individuals by tracking their parts becomes problematic because parts change status with changes of location: by becoming parts of different individuals, by recombining or merging with parts of other individuals, by segmenting and ceasing to be individual component parts at all, or by virtue of the wholes of which they were parts merging so that the parts come to belong to new individuals. When hybrids form in fusion, assembly, diffusion, or progeneration processes, the criteria we use to signify and track individuality tend to change; hence the labeling scheme elevates or lowers the salience of previously labeled parts. While we may, up to a point, have tracked parts and interpreted individuality in terms of the sources of the parts’ production, after a hybridization event the source of production may be less salient than the part/whole relation the parts entered into in forming the hybrid, or the way in which parts produced from various sources became organized, interact, or function together.

In hybridization studies, the practice of labeling parts and then tracking the labels (or collecting their traces) requires that we shift our tracking practice during the activity so as to track newly emergent individuals of interest as parts and wholes change relations, come to be, or pass away. The shift, in operational terms, means relabeling entities as their relations or our ontological commitments change while we shift our attention. If our interest in DNA replication is the fate of strands rather than showing that replication is semi-conservative, then once new strands are formed in replication, the pair of strands should be relabeled to reflect their roles in a new individual of hybrid provenance. Relabeling will either amount to taking the existing label to signify a different individual after than before a stage of the process (mental relabeling) or adding or subtracting physical marks “on the fly” due to the reorganization of wholes whenever the originally labeled parts and labeling scheme are no longer a good choice. These conceptual shifts should then be aligned with symbolic representations, for example, in dynamical equations and diagrams by subscripting variable names or other visual indicators of new individuals being tracked, among generations. Mendel symbolized hybrid states, for example, by concatenating letters standing for pure states. If A and B stood for true-breeding traits, then AB stood for the hybrid. AB is a form of relabeling to mark the hybrids of intermediate generations, for example, the F1, between the pure-breeding parentals of generation P and the segregating grand-offspring of generation F2.

Relabeling techniques can be applied to tracking processes of hybridization, that is, the formation, assembly, development, and reproduction of hybrids, in order to address questions about their status as individuals and as units of investigation. Just as philosophers of evolutionary biology have addressed units of selection by considering questions of ownership and beneficiaries of adaptation (Lloyd 2012), I suggest we consider different ways of attributing and tracking individuality through reproduction by considering questions of ownership and scaffolding of developmental capacities: who are the component sources, the assemblers, the scaffolders, the carriers, and the realizers of developmental capacities (if any)? Attitudes toward these questions affect the salience of particular grains of description and use of particular tracking tools and techniques for causal narratives of reproduction processes. These choices bear on how seriously one takes formal vs. material modes of reproduction. At issue is the status of hybrids as new individuals or only as transitional states and how to track them.

The formal or material character of biological reproduction turns, I argue, on how one relies on hybrids as units of investigation and as individuals. Individuality, in other words, can be understood to name a relation between the attention, abilities, and interests of someone who tracks biological phenomena on the one hand and the properties, relations, behaviors, and activities attributed to the things tracked on the other hand, rather than as an intrinsic property of concrete particular subjects. This kind of “hybrid” onto-epistemic concept might be heuristically useful in searching for cases of cognitive or cultural development that extend the Darwinian domain. In order to have an exemplar for this kind of exploratory work, I will discuss the reproduction of HIV-1 retroviruses. Although stretching concepts in the direction of viruses would seem not to advance the cause of understanding cognition or culture, it does so by probing concepts beyond the zone of proximal intuition and expectation about the character and behavior of organisms. The question is whether HIV-1 replication is a case of: (1) repeated assembly of viruses by host cells, acting like weapons factories turning out “artifacts,” or (2) reproduction of viruses by viruses relying on substantial scaffolding by host cells, or (3) reproduction of viruses through hybridization of virus and host cell to produce “F1” hybrid virus-cell generations in between the parental generation of separate virion and T-cell and an “F2” generation of separate offspring virion and postinfection host cell.

The challenge is that there appears to be no material overlap between parent and offspring HIV viruses and that it’s a stretch to say that such viruses develop (Godfrey-Smith 2009, chapter 4). If they don’t develop, then it is unimportant on my view whether material overlap relations hold because the point of material overlap in reproduction is to confer developmental capacities through the transfer of organized material parts from parents to offspring. HIV-1 replication appears to be a case of assembly of components synthesized by a host cell which then persist in a virion “particle” in an equilibrium state until infection. Because the RNA genome of the virus is reverse transcribed to DNA by the host cell and integrated into the host genome, there appears to be no material overlap between parent and offspring virus: virus replication seems to be a case of repeated assembly and not a case of development, hence not a case of reproduction on my account. An alternative choice of central subject and narrative grain, however, supports either the scaffolded reproduction or alternation of reproductive generations narrative model.

Standard narratives of HIV-1 replication start with a mature virus or “virion” binding to a CD4 receptor and G-protein coreceptor on a helper T-cell (step 1, figure 1.2).¹³

Figure 1.2

HIV-1 reproduction in a human T-cell. Numbered steps are described in the text. Redrawn after Scherer et al. (2007), figure 1 (image copyright 2012, James Griesemer). PIC, pre-integration complex; RT, reverse transcriptase.

The virion envelope fuses (step 2) to become part of the host cell membrane along with the virion’s surface proteins (which had been part of the plasma membrane of the virus’s host cell parent from which it emerged after the grandparent virus’s reproduction). The nucleocapsid enters the host cell and is uncoated to expose the contents of the virus (step 3). The spatially distributed molecular virus parts are now like the ameboid cells of a slime mold, spreading through the “leaf litter” of the host cell’s cytoplasm, foraging for DNA nucleotide “food.” Some of the virion’s parts are left in the infected cell’s membrane (i.e., lipids the virion picked up from its parent host cell plus its viral envelope proteins). Others are degraded. The rest diffuse through the cytoplasm of the host cell. The virus’s single-stranded RNA genome travels as a package with a parent virus reverse transcriptase enzyme that inaccurately transcribes the virus genome to double-stranded DNA (step 4). Each step involves the material overlap from one stage to the next of virus parts that confer capacities (carried by enzymes that catalyze reactions) to move to the next step.

After reverse transcription, the parent viral RNA is degraded. The remaining “pre-integration complex” (PIC) of parent virus matrix protein, integrase enzyme, and reverse-transcribed virus DNA genome is transported to the host cell nucleus (steps 5, 6). Again, there is material overlap as the PIC shuttles to the cell nucleus. In each of these stages an organization of proteins and nucleic acids carries the developmental capacity to advance to the next stage into a particular host environment that scaffolds the development of offspring virus particles. The “provirus” is then integrated into the host cell DNA (step 7).

Later, the provirus, that is, double-stranded DNA integrated into the host cell genome, is transcribed to offspring virus single-stranded RNA by the host cell (step 8).¹⁴ Some copies end up as offspring virus RNA genomes. Some become messenger RNA (mRNA) coding for proteins that help transcribe the virus DNA (steps 8–10), and some are singly spliced to produce mRNA coding for polyproteins that become packaged into offspring virus capsids (steps 11, 12). Again, these are unproblematic cases of material overlap conferring developmental capacities to get to the next stage of virus production.

The offspring virus parts are, again, in a spatially distributed state, moving through the host cell to the cell surface (step 13). The offspring virus parts are recruited to a special location created by the distributed virus parts themselves: a so-called “raft” in the host cell plasma lipid bilayer membrane where transmembrane virus envelope proteins embed and recruit other virus parts.

There are two stages that might be considered “barrier steps” lacking material overlap between successive stages in the virus life cycle. The first is when the parent virus RNA genome is reverse transcribed to DNA (step 4). The DNA nucleotides were either acquired or manufactured by the infected host cell. The RNA strand came from the parent virus. If the “source of components” is the criterion of parts ownership, then the DNA strand assembled by the virus reverse transcriptase is part of the host cell. In that case, the synthesis involves a hybridization of virus and host at the nucleotide strand level and we would conclude that there is no material overlap between the two strands. However, the standard narrative of what happens next treats this RNA–DNA hybrid as an individual operated upon again by the parent virus reverse transcriptase: it copies the DNA strand to make double-stranded DNA and then degrades the RNA strand. In other words, the parent virus protein scaffolds the molecular hybrid offspring’s development from RNA–DNA to DNA–DNA. Here, if “source of components” continues to be the ownership criterion, then we have to say that this double-stranded DNA virus is made of entirely host material, even though the information—its sequence—came from the parent virus. Hence, Godfrey-Smith (2009) calls this “formal” reproduction.

If however we use a who-made-it “assembler” criterion of ownership, then we should say that the DNA strands are virus material because parent virus reverse transcriptase assembled the new strands using old strands as templates, as in “ordinary” DNA replication. On this interpretation, the RNA–DNA hybrid and the double-stranded DNA look like materially overlapping stages in the “maturation” of the parent virus genome on their final journey to the host cell nucleus, or an alternation of generations of hybrid offspring virus genome, scaffolded by parental virus enzymes, to develop from a single-stranded RNA to a double-stranded DNA state. Finally, if we consider covalent bonding of nucleotides into a polymer as a part/whole criterion of ownership, then the DNA strand takes ownership of the host-derived nucleotides as it grows, due to the scaffolding agency of the reverse transcriptase, though the strand’s organization is scaffolded by its template environment. To think about the individuality of the hybrids, consider a relabeling procedure which shows how tracking only source of components gets into trouble in analyzing the late steps in virus replication.

Suppose we think of strands rather than nucleotides as the salient developmental capacity-carrying parts of double helixes. After all, nucleotides don’t carry genetic information—sequences of nucleotides in strands do. Then the formation of the RNA–DNA hybrid would call for relabeling to show that both strands belong to a given double helix because covalent bonding of nucleotides into a strand marks a change of ownership of nucleotides at the same time hydrogen bonding of the nucleotides to the other strand in a templated strand assembly marks the internalization of the RNA scaffold (templating strand) as a part of a new, hybrid RNA–DNA individual. The Meselson–Stahl procedure for tracking the strands still works, but only insofar as strands are taken to be wholes and nucleotides their parts. If double helixes are the wholes and strands are the parts we want to track, then we have to relabel strands after they are synthesized, as members of a new “generation,” just as Mendel relabels traits as hybrid when factors are in the hybrid state: AB rather than A or B. With relabeling, we recognize a series of generations, each linked by material overlap: from single-stranded RNA to the hybrid RNA–DNA, because the RNA strand overlaps the hybrid after the first DNA strand is synthesized, and also from hybrid RNA–DNA to the double-stranded DNA, because the first DNA strand from the hybrid takes on the role of scaffold (template) for the synthesis of a new DNA strand and then becomes “internalized” as a strand part which overlaps the double-stranded DNA.

If RNA to DNA in step 4 isn’t a barrier to material overlap from parent to offspring viruses, in an alternation of RNA and DNA molecular-state “generations,” it’s no good trying to locate a barrier at the integration of provirus double-stranded DNA into the host cell genome either. The way the provirus DNA strand is integrated into the host genome doesn’t transfer any of the “genetic information” or developmental capacity from the virus part of the integrated genome to the host part of the genome although the provirus is certainly a part of the host cell genome and the system is, I think, best understood as a hybrid individual with host-parent and virus-parent mingled developmental capacities that are realized in concert when offspring viruses are produced. The standard narrative interprets the situation as virus co-opting the host cell’s transcription and translation machinery, as though the cell were a mere assembly factory, controlled by a diabolical agent, and no longer a developmental agent itself. Both formal and material accounts of reproduction would see the provirus as materially overlapping the host genome, but only the material account tracks where the developmental capacities are carried.

The “source of components” interpretation of the ownership of the parts completely breaks down when we get beyond the late steps to the escape of the assembled virion from the host cell. If the source of the original DNA nucleotide components interpretation is used to argue that there is a barrier to material overlap in reverse transcription (step 4), this interpretation also implies that the host cell owns the offspring virus RNA nucleotides after it has left the host to infect another cell. The cell membrane “skin” is not a relevant boundary/individuality marker of viruses or virus generations if “source of components” is the parts ownership criterion. Since tracing back enough virus generations leads to the discovery of a host-cell source of all of the component parts of the virus, it will turn out on that criterion that a virus is just means for a host cell to make another (hybrid) host cell. The standard narrative, however, was structured with the choice of a mature virion particle as central subject, initiating infection by binding to a host cell, as a means of infecting other host cells. Something has gone badly wrong if the criterion of individuality inverts what counts as subject and what counts as means during the narrative.

A plausible way to put the intent of the standard narrative right is to recognize the host cell as an elaborate scaffold for virus reproduction for which the shifting character of the scaffold’s role as a developmental agent is less important than its role in scaffolding virus development. That is the way scientists usually describe the process: the host cell is turned into a “factory” for virus assembly. An alternative I argued is also compatible with the material facts is to treat the hybrid RNA–DNA as an individual in a series of generations of individuals that confer developmental capacities to subsequent stages of virus development by means of the material overlap of parts. Just as fertilization–fusion events delimit organism generations for sexual multicellular organisms, fusion events producing molecular hybrids that carry developmental capacities can be interpreted as delimiting molecular generations of virus life cycles. The source of components criterion is the weakest of all the relevant part/whole criteria because it becomes irrelevant as soon as any causal interactions constituting molecular hybridization of the components take place. The relabeling idea tracks how investigators studying complex life cycles such as HIV-1 routinely shift their criteria of part/whole relations delimiting individuality as they track units of investigation. This is best seen by thinking about the checkpoints in the life cycle where relabeling would be required to track these changes.

A final point about HIV-1 replication is to note a key difference between repeated assembly processes and developmental ones. It is a stretch to say that retroviruses develop if development is a narrow concept limited, by definition, to multicellular organisms, for example, by limiting development to cellular differentiation. Enveloped viruses like HIV aren’t cells and they lack metabolisms in the usual sense of internal management of matter and energy transduction. However, there is a stage in the HIV replication cycle that biologists call “maturation,” and this fits pretty well the broad concept of development I propose here: the proteins packaged in the offspring virus particle that escapes the host cell (step 14) include a protease. In order for an HIV particle to become a mature, infectious virion capable of reproduction, the protease has to cleave the polyproteins packaged inside it to release the proteins that form the capsid and other structures (step 15). In short, without maturation, the virus particle will not be infectious (hence the early focus on protease inhibitors in HIV treatments). Infection is a capacity necessary for reproduction in these viruses, so acquiring infectiousness fits precisely my notion of development as the acquisition of the capacity to reproduce. Unlike the rifles produced in a rifle factory (see W. Wimsatt, this volume), virus particles transform themselves—they develop—in order to contribute to the production of new viruses while rifles do not do that—except possibly in the indirect sense that rifles may be used by people to force other people to build more rifles. If that occurs, then we should consider the dynamics of the social system and consider tracking rifle production as a form of socially scaffolded reproduction or perhaps consider rifle-people hybrids in an alternation of cultural generations model.

I conclude that HIV-1 replication can satisfy the material overlap and development conditions of my account of reproduction, contrary to the appearance that it involves barriers to the transfer of matter from parent to offspring viruses in such a way that reproduction must be considered formal. What the example shows is that formal mode reproduction amounts to the propagation of form via material means in which information dissipates on the time scale of its propagation if not maintained by material scaffolding. Dissipation in this case meant the active disassembly of RNA strands by ribonuclease activity of the parent virus and the transfer of the information or developmental capacity via hybrid molecular generations to a provirus that had disassembled its armor upon embedding in its host/hybridization partner. A reproducer perspective on HIV requires recognizing molecular hybrids as individuals in hybrid generations and whose assembly and transformations amount to development and reproduction in the senses I have defended. The proviso is that we must draw system–environment boundaries in particular ways that reflect the relevant research problem, which is a way of characterizing the Darwinian domain of “organisms” without being forced into the triviality of claiming that the domain is whatever Darwin’s principles apply to.

No “ordinary” conception of “organism” will suffice to characterize the domain since even enveloped retroviruses and proviruses consisting of double-stranded DNA in the right context belong to it. The more interesting question is how widely this view of “molecular” hybrid reproduction extends and whether including cases where some of the materiality lies outside the system in our initial descriptions should lead us to model them differently or to alter our initial descriptions and track them, by relabeling, as extraordinary hybrids. In the final section, I will very briefly raise the possibility that my concept of reproduction, together with the tools developed for thinking about reproducers in terms of tracking and relabeling hybrids, might give us some insight into units of distributed cognitive and cultural inheritance and development. I view this as a first step toward addressing the question of cultural evolution from a developmentalist, reproducer perspective.

In biology, we know that a wide range of interactions among organisms have indirect genetic effects, for example, epistasis between genomes (Wolf et al. 2000). Because organisms that interact in these ways are reproducers that form lineages, the presuppositions about reproduction that I have been exploring should apply. There is no particular conceptual barrier to extending the Darwinian domain of individual biological organisms, as interpreted from the reproducer perspective, to complex groups of interacting organisms.

In culture, interactions often are not like that, however. Interaction between organisms might be mediated by nonorganism, nonreproducer intermediaries, prostheses, or environmental features, for example a nest, a habitat, or a food resource. In human culture, humans make artifacts, we interact with these artifacts in development, and the artifacts themselves bear relations of similarity to one another that are tempting to describe as hereditary, such as series of designed products. In many cultural settings, acquisition of the cultural capacities needed to successfully reproduce biologically or to repeat the acquisition of a cultural capacity in a new individual depends on nonorganism components or intermediaries. In brief, cultural environments “scaffold” human cognitive and cultural development so that developmental capacities are realized in cognitive/cultural offspring, producing correlations between cultural parents and offspring that can be tracked and analyzed according to dual inheritance models (Boyd and Richerson 1985). Thus, dual inheritance models are to cultural inheritance as population genetic models are to biological inheritance. Scaffolding provides a salient concept for investigating mechanisms of cultural (and cognitive) development and inheritance.

Scaffolding brings heterogeneous entities together in performances of development. Heterogeneity is central to the scaffolding relation. (A builder’s scaffold could not work if it had the same structure or function as the building it scaffolded.) But heterogeneity poses problems for understanding development in relation to reproduction. How shall we conceptualize assembly processes for the purposes of modeling development in reproduction, or more generally in any kind of repeated assembly process, as cultural hybrids are formed and disposed? Tracking the impact of environmental scaffolding on a developing reproducer sounds a bit like what is called “multilevel selection 1” in the multilevel selection literature (Heisler and Damuth 1987): tracking groups through their contextual effects on “lower” level entities. Tracking the reproduction of hybrid complexes “at the level of the hybrids,” that is, tracking groups through their group-level effects, sounds a bit like “multilevel selection 2.” It is beyond the scope of this chapter to explore this analogy further, but the analogy suggests potential theoretical modeling strategies through contextual analysis.

I have tried to set up the problem of development in such a way that embodied and distributed cognition or cultural development need not look so very different from biological development: how can heterogeneous elements of distinct provenance or lineages be hybridized in such a way that development from the hybrids (as opposed to disassembly) may not only proceed but can also propagate developmental capacities to a next generation of new individuals, to repeat the assembly of successful reproducers, so as to produce observable patterns of variation among offspring, all without disrupting sequential expression of features differentially entrenched in a hierarchically organized, generative process? W. Wimsatt’s (this volume) notion of combinatorial entrenchment can help, but while the elements of a combinatorial algebra may be entrenched, in development we must also imagine combinations or their effects can be transformed in ways that prevent simple decomposition or disassembly back into starting configurations. W. Wimsatt (this volume) points out that with the standardization of parts in human cultural systems of manufacture, assemblies that figure in complex hierarchical artifacts tend to become “black-boxed,” which often means we must forego the option of disassembly for understanding, maintenance, or reuse of parts. Inability to disassemble, then, may be a key indicator of cultural development processes. One way hybridity enters the picture of development is through the interaction of developing entities with environments that scaffold their development, changing parts of the hybrid system in ways that make assembly contingently irreversible and disassembly into starting configurations correspondingly unlikely. As I have argued in the case of HIV, scaffolding creates hybrid developmental complexes, essentially making the development problem one of the development of hybrids and generative entrenchment of irreversible assembly processes.

There is nothing intrinsically biological about scaffolding or the general concept of development as I have formulated it, so it is an open empirical question whether developmental processes can be organized in cognitive or cultural systems in such a way that individuals in those domains function as reproducers. Structurally, a scaffold is an entity, typically exogenous to a system, unit, or object of interest, which interacts in a temporary association with the system to facilitate the development of an outcome or effect (positive or negative) which would otherwise be difficult or impossible to achieve. “Functionally, scaffolding is precisely the creation of … bracketed trajectories of potential development through artificially created nearby points of stability” (Bickhard 1992, 35). Processually, scaffolding is an extended interaction or series of interactions in development, through which a skill, capacity, function, property, or state is thereby acquired by a scaffolded process. However, while the interaction may be extended and temporary, scaffolds can persist on shorter, longer, or the same time scales as what they scaffold. When scaffolding persists on the same time scale as the system, then it is likely we will identify the assembly as a hybrid, but not necessarily as a new individual in a distinct generation. The dynamics and effects of scaffolding, therefore, can not only be quite diverse but can also have significant conceptual implications for the status of hybrids.

The challenge in exploring whether the Darwinian domain can be extended to distributed cognition or culture can be made a little more tractable if we begin with questions about whether and how such systems develop and reproduce rather than whether they evolve according to Darwinian principles. How should we assess hybrids that form in processes of cultural change which might constitute development, such as groups of interacting people, assemblies of people interacting with artifacts and infrastructure, and artifacts interacting with other artifacts? We commit ontologically to kinds of part/whole relations when we track configurations and organizations of these entities as they change, but a central question is whether we must relabel in order to track individuals of interest through cognitive or cultural processes. Task groups—small core configurations of people that may be repeatedly assembled within a social setting—provide instructive examples.

Task groups are one of Caporael’s core configurations, of modal size around five, having distributed cognition as their modal task (Caporael, this volume, table 2.1). Hunting parties, fire crews, and flight crews are some examples.¹⁵ A task group is similar to a trait group—one of two chief units that have been used to model group selection—except that it forms to perform a task and is size dependent, rather than in virtue of a common trait leading individuals to assemble into a group and share fates (Wilson 1980).¹⁶

For some task (or larger) groups, artifact production is the goal and completing the task means producing the final product (hand axes, an iPod component, a prey item, a sown crop, a harvested crop) by coordinating with other task groups into larger configurations (Caporael’s demes or communities and macrodemes). For other task groups, the “product” is a performance (teaching, learning, dancing, playing, navigating). Gerson (this volume) discusses “Saturday night musicians” coming together, with their instruments, in Chicago to perform as bands in nightclubs on Saturday nights, scaffolded by common knowledge and various institutions or conventionalized practices and procedures. Hutchins (1995) describes navigation of a ship through San Diego harbor in terms of a task group assembled, in certain parts of the ship, together with a collection of artifacts to perform navigation. Even when artifacts are the goal, we can analyze task groups (or groups of groups) or organizations in terms of process and treat the artifacts as “side effects” or analyze the artifacts (or series or sequences of artifacts) and treat the production process as “ways and means.” Or, we can treat artifacts and performances as parts of a process of “dwelling,” focusing on the life world in which people, artifacts, and performances hybridize and proceed together through a sequence of “generations” alternating among various hybrid forms (Ingold 2000).

Instead of the shared fate of trait groups suited to multilevel selection analysis (Wilson and Dugatkin 1997; Sober and Wilson 1998), the individuals of a task group have coordinated fates. Task group members may play complementary roles (teacher/student, lead/rear, guard/forward/center, holders of different pieces of knowledge or know-how) in virtue of their shared purpose rather than the same role in virtue of their shared trait and may not share fates as a result. Indeed, in task groups, division of labor, with potentially diverse fitness consequences, is nearly a requirement in order to pursue a shared task. The task group consisting of a five-fingered hand would not work well for many kinds of tasks if the fingers were all the same—shared morphology—rather than differentiated. In a hunting party, some task group members may serve as scouts or sentinels, others as attackers, flankers, or butchers. Their fates may correspondingly differ: sentinels fall to other predators, attackers suffer goring from prey, butchers suffer self-injury from their butchery. In a training task group, teachers teach and learners learn, but teachers also (may) learn how to teach and learners (may) learn how (or that) they teach their teachers, along with many other “side effects.”

Moreover, the individuals of a task group typically have multiple, fitness-affecting relationships beyond that group. People typically belong to many “reference” groups that are as often organized by task (e.g., business firm, volunteer fire department, chamber of commerce, government agency) as by trait (e.g., shared religion, political party, or fan club). D. S. Wilson’s trait group model (Wilson 1980; Wilson and Dugatkin 1997) supposes random mating within the meta-population regardless of trait. Task groups may be formed from members of families, trait groups, communities, firms, or other organizations. So, there can be multiple “reference groups” intersecting in task groups and complex patterns of relatedness and population subdivision (W. Wimsatt, this volume). The evolutionary consequences of selection among and within task groups will therefore be multivalent, affecting the intersecting reference groups in varying ways. Analysts will also tend to see the groups and members as units differently, depending on the ways in which group members are related and analysts attend to different relationships while tracking them.

Breeding groups (“demes” in the parlance of population geneticists), on the other hand, are delimited with respect to reproductive relationships as the organizing process. This makes demes both narrower and broader than trait groups and task groups (unless the trait is reproductive interfertility or the task is reproduction). They are narrower in the sense that breeding groups are groups organized for a single task—reproduction—and united (grouped, assembled) by a shared purpose or shared relation.¹⁷ They are broader in the sense that any activities that bring members of a breeding group together in condition to breed count as aspects of reproduction. Like trait groups, breeding groups can range fairly widely in size, though the mechanisms, conditions and material constraints of reproduction for a species will delimit an effective size.

Breeding and trait groups have been of interest in population biology, in part, because of well-known debates about units of selection (reviewed in Lloyd 2012). Interdeme (Wade 1976, 1977, 1978) and trait-group (Wilson 1980) selection, over and above selection on their member organisms (Williams 1966), were foci of controversies regarding group selection and multilevel selection in the 1980s and 1990s. Here, my interest is in the status of reproduction within and of groups.

Trait groups do not fully qualify as reproducers because trait groups disassemble to component organisms, which are incapable of carrying group developmental capacities other than via organism-level inheritance mechanisms after an episode of viability selection on the group. Differently put, there is no “population structure” in the (meta-)population of mating organisms in a trait group which mate at random while there may be population structure in a (meta-)population of trait groups.

In interdeme selection, groups are reproducers, producing offspring groups with which they materially overlap via group propagules that form a “propagule pool” of such groups rather than a “migrant pool” of organisms (see Wade 1978). A propagule in the group selection context is a group of organisms that constitutes a group-level individual that carries group developmental capacities and functions as an offspring group fissioned from the parent group or as a propagule fusing with others to form offspring hybrid groups from several to many parent populations (Wade 1976; Wade and Griesemer 1998; Griesemer and Wade 2000). Differential survival, development, and reproduction of groups forms the core reproduction process of interdemic group selection, though of course there cannot be any groups if there is not also organism reproduction, and selection processes at the two levels can be expected, in general, to interact.

The diverse contributions of members inherent to the nature of human task groups, in contrast to the defining similarity of members of trait groups, suggest the members are of different provenance. Task groups are hybrids. If task groups go through developmental processes of transformation that produce the capacity to form new task groups, then they are also reproducers. If, instead, the members of a task group disburse once their task is complete and return to their families, villages, or other-organized groups, then the task group behaves more like trait groups which can be repeatedly assembled from different participants, where the repetition is due to the shared trait of the assembling members.

Trait groups, breeding groups, and task groups (core configurations) form a triumvirate of distinct kinds of collective units that I suggest would serve much better than any single approach to units for tracking and modeling cognitive or cultural change through development and reproduction and, thus, for evolutionary analysis. Social organizations of humans have, as Caporael (this volume) points out, properties that mix characteristics of the three. In many cultural settings, task groups may also be related to breeding groups (e.g., families in bands or macrobands), individuals that assemble in a trait group may become a task group (see note 16). Breeding groups may subdivide into trait or breeding groups.

The concepts of scaffolding and the development of cognitive and cultural hybrids bring to attention the notion that these units may include artifacts as well as biological reproducers and that the status of scaffolds and hybrids may depend as much on which patterns are of interest on which time scales, as on any notion of a proper ontology for evolutionary units of culture or cognition to be derived from taking Darwin’s principles or the Darwinian domain of reproducers for granted.

I thank the Konrad Lorenz Institute for Evolution and Cognition Research staff and Executive Board, Gerd Müller, and Werner Callebaut for supporting the workshop and this volume; I also thank all the participants in the workshop, who contributed greatly to my understanding of the topic and expanded my horizons. I also thank Linnda Caporael and Bill Wimsatt for many fruitful conversations and for their feedback on my chapter as well as the Indiana University History and Philosophy of Science Department for convening an engaged audience. I also acknowledge the award of a Herbert A. Young Society dean’s fellowship for 2011–14 at the University of California, Davis.

1. A question of individuality can be expressed in terms of who owns this use: the environment or agent that scaffolds, the developing system that is scaffolded, or the hybrid formed by their interaction.

2. More generally, scaffolds persist on different time scales than what they scaffold. Infrastructure can persist on very long time scales relative to individuals who use it and thus create correlated environments for organisms of different generations.

3. My account is in sympathy with Oyama’s (1985) argument for the ontogeny of information.

4. This claim is intended as heuristic: to suppose genealogical relationships just are ones that involve material overlap presents theoretical and empirical, rather than conceptual, challenges to accommodate “information transmission” cases that appear not to involve material overlap, whether the inheritance is genetic, epigenetic, behavioral, or symbolic (Jablonka and Lamb 2005). Similarity relations are “genealogical” by analogy only. The repeated assembly of phenotypes in biological development, or artifacts in culture, by genealogically ordered genotypes or cultural agents, both have Weismannist structure (Griesemer and Wimsatt 1989). The relation of successive phenotypes or artifacts is not genealogical per se; both involve development. See Wimsatt and Griesemer (2007, figure 7.2).

5. It is not an objection to the view that some reproduction events fail to confer developmental capacities so that the offspring is sterile or dies before completing reproduction. We can distinguish the transfer of material parts conferring malfunctioning developmental capacities and unsuccessful reproduction from the transfer of functional parts and successful reproduction.

6. Light does have to reflect off the “sender” and be received by the “recipient.” Shall we interpret light as propagating developmental capacities by means of material overlap? There is perhaps a case to be made for interpreting Gibson’s (1979) notion of affordances of an optic array in material terms.

7. In many cases, parents scaffold their offspring in addition to providing initiating propagules of offspring, but here I distinguish these two connections between parents and offspring from one another. The “developmental process” interpretation of developmental systems theory also grapples with this problem of system–environment distinction (see Griffiths and Gray 1994). Here I address the issue in terms of the status of material objects formed in reproduction processes. It is beyond the scope of this chapter to relate the material perspective discussed here to process interpretations of developmental systems theory.

8. This of course is a notion of parenthood with respect to the material conditions of reproduction. There is a completely different and equally legitimate notion of parenthood that rejects material overlap genealogy as the basis of parenthood and instead identifies parenthood directly with the scaffolding function: there is no material overlap between such a parent and an adopted child. I have no objection to that notion of parenthood and indeed embrace it as quite relevant to human cultures. Cultural systems with sufficient scaffolding of fragile hybrids may be functionally equivalent to, and as robust as, systems using material overlap more substantially.

9. Haraway’s (1991) notion of cyborgs—organism–machine hybrids—pushes the blurring as far as it will go.

10. Parnes (2007) offers an important analysis of early-nineteenth-century changes in meaning of “generation” as critical to the Mendelian project later in the century and to the transition to a modern (twentieth-century) notion of heredity. See also Müller-Wille (2007).

11. The same goes for construction scaffolds, which remain only as long as the building is “under construction.” It’s a bad building that doesn’t outlast its scaffolding, though see below on Sagrada Familia.

12. Griffiths and Gray (1994) point out that the relevant interaction between organism and sunlight is not persistent but rather recurrent (as the Earth turns) and depends on the behavior of the organism as much as on the persistence of the sun, while organism–sun gravitational interactions are continuous rather than recurrent. Perhaps one can think of photon capture as a form of transfer of material parts; more often, these physical features are not treated as consumable resources like food but rather more like a substrate, platform, or enabling condition for organism–environment interaction.

14. Steve Lawrie (personal communication) points out that HIV infection may be co- or multiple infection and that viral DNA not integrated into the host genome can actually complete replication by complementation, so the “parentage” of the hybrid, alternating molecular generations described here is already a simplification (see Gelderblom et al. 2008).

15. Group sizes suited to particular tasks are no doubt culture, history, and technology dependent: a single farmer with a John Deere can plow a large field, though he or she is dependent on an entire society for the conditions to do so, or it can take a large number of workers with hand-tools they made themselves, or it can take one of the latter a much longer time.

16. Trait and task group formation can look the same if individuals aggregate because of a trait and then perform a task aggregatively, for example, if the trait shared is an individual goal state, such as foraging on a leaf in order to gain sustenance, with a shared fate as “side effect,” such as they all get eaten because a bird spots the aggregate as a blob of food. In human culture, people can form trait groups (e.g., by stepping into the same elevator) but then become a task group (if the elevator gets stuck) as Caporael (this volume) describes.

17. Note that Caporael (1997, 2001) treats dyads as essential for reproduction but demes (bands, communities, of modal size around thirty) as required for successful child rearing.

Barad, K. 2007. Meeting the Universe Halfway: Quantum Physics and the Entanglement of Matter and Meaning. Durham: Duke University Press.

Bell, G. 1982. The Masterpiece of Nature. Berkeley: University of California Press.

Bickhard, M. 1992. Scaffolding and Self-Scaffolding: Central Aspects of Development. In Children’s Development within Social Contexts: Research and Methodology, edited by L. Winegar and J. Valsiner, 33–52. Mahwah, NJ: Lawrence Erlbaum.

Boyd, R., and P. Richerson. 1985. Culture and the Evolutionary Process. Chicago: University of Chicago Press.

Buss, L. 1987. The Evolution of Individuality. Princeton: Princeton University Press.

Calow, P. 1978. Life Cycles: An Evolutionary Approach to the Physiology of Reproduction, Development, and Ageing. New York: Chapman and Hall.

Caporael, L. 1997. The evolution of truly social cognition: The core configurations model. Personality and Social Psychology Review 1:276–298.

Caporael, L. 2001. Parts and Wholes: The Evolutionary Importance of Groups. In Individual Self, Relational Self, and Collective Self: Partners, Opponents or Strangers? edited by C. Sdeikides and M. Brewer, 241–258. Philadelphia: Psychology Press.

Caporael, L. 2007. Evolutionary Theory for Social and Cultural Psychology. In Social Psychology: Handbook of Basic Principles, edited by E. Higgins and A. Kruglanski, 3–18. New York: Guilford Press.

Churchill, F. 1987. From heredity theory to Vererbung: The transmission problem, 1850–1915. Isis 78:337–364.

Dawkins, R. 1982. The Extended Phenotype: The Gene as the Unit of Selection. New York: Oxford University Press.

Gelderblom, H., D. Vatakis, S. Burke, S. Lawrie, G. Bristol, and D. Levy. 2008. Viral complementation allows HIV-1 replication without integration. Retrovirology 5:60.

Ghiselin, M. 1997. Metaphysics and the Origin of Species. Albany: State University of New York Press.

Gibson, J. 1979. The Ecological Approach to Visual Perception. Hillsdale, NJ: Lawrence Erlbaum.

Godfrey-Smith, P. 2009. Darwinian Populations and Natural Selection. New York: Oxford University Press.

Gould, S. 2002. The Structure of Evolutionary Theory. Cambridge, MA: Harvard University Press.

Griesemer, J. 2000a. Development, culture and the units of inheritance. Philosophy of Science 67 (Proceedings): S348–S368.

Griesemer, J. 2000b. The units of evolutionary transition. Selection 1:67–80.

Griesemer, J. 2005. The Informational Gene and the Substantial Body: On the Generalization of Evolutionary Theory by Abstraction. In Idealization XII: Correcting the Model, Idealization and Abstraction in the Sciences, edited by M. Jones and N. Cartwright, 59–115. Amsterdam: Rodopi.

Griesemer, J. 2007. Tracking Organic Processes: Representations and Research Styles in Classical Embryology and Genetics. In From Embryology to Evo–Devo: A History of Developmental Evolution, edited by M. Laubichler and J. Maienschein, 375–433. Cambridge, MA: MIT Press.

Griesemer, J. 2011a. Heuristic Reductionism and the Relative Significance of Epigenetic Inheritance in Evolution. In Epigenetics: Linking Genotype and Phenotype in Development and Evolution, edited by B. Hallgrímsson and B. Hall, 14–40. Berkeley: University of California Press.

Griesemer, J. 2011b. The Relative Significance of Epigenetic Inheritance in Evolution: Some Philosophical Considerations. In Transformations of Lamarckism: From Subtle Fluids to Molecular Biology, edited by S. Gissis and E. Jablonka, 331–344. Cambridge, MA: MIT Press.

Griesemer, J. 2012. Formalization and the meaning of “theory” in the inexact biological sciences. Biological Theory. Online First. doi:10.1007/s13752-012-0065-z.

Griesemer, J. Sharing Spaces, Crossing Boundaries. In Boundary Objects and Beyond: Working with Leigh Star, edited by G. Bowker, S. Timmermans, and A. Clarke. Cambridge, MA: MIT Press.

Griesemer, J., and M. Wade. 2000. Populational heritability: Extending Punnett square concepts to evolution at the metapopulation level. Biology and Philosophy 15:1–17.

Griesemer, J., and W. Wimsatt. 1989. Picturing Weismannism: A Case Study of Conceptual Evolution. In What the Philosophy of Biology Is, Essays for David Hull, edited by M. Ruse, 75–137. Dordrecht: Kluwer Academic.

Griffiths, P., and R. Gray. 1994. Developmental systems and evolutionary explanation. Journal of Philosophy 91:277–304.

Haraway, D. 1991. Simians, Cyborgs and Women: The Reinvention of Nature. New York: Routledge.

Heisler, I., and J. Damuth. 1987. A method for analyzing selection in hierarchically structured populations. American Naturalist 130:582–602.

Hull, D. 1975. Central subjects and historical narratives. History and Theory 14:253–274.

Hull, D. 1980. Individuality and selection. Annual Review of Ecology and Systematics 11:311–332.

Jablonka, E., and M. Lamb. 1995. Epigenetic Inheritance and Evolution. Oxford: Oxford University Press.

Jablonka, E., and M. Lamb. 2005. Evolution in Four Dimensions: Genetic, Epigenetic, Behavioral, and Symbolic Variations in the History of Life. Cambridge, MA: MIT Press.

Maienschein, J. 1987. Heredity/development in the United States, circa 1900. History and Philosophy of the Life Sciences 9:79–93.

Maynard Smith, J. 1986. The Problems of Biology. New York: Oxford University Press.

Maynard Smith, J. 2000. The concept of information in biology. Philosophy of Science 67:177–194.

Meselson, M., and F. Stahl. 1958. The replication of DNA in Escherichia coli. Proceedings of the National Academy of Sciences of the United States of America 44:671–682.

Moran, N. 1992. The evolution of aphid life cycles. Annual Review of Entomology 37:321–348.

Müller-Wille, S. 2007. Hybrids, pure cultures, and pure lines: From nineteenth-century biology to twentieth-centurey genetics. Studies in History and Philosophy of Biological and Biomedical Sciences 38:796–806.

Müller-Wille, S., and H.-J. Rheinberger. 2007. Heredity Produced: At the Crossroads of Biology, Politics, and Culture, 1500–1870. Cambridge, MA: MIT Press.

Newman, S., and R. Bhat. 2009. Dynamical patterning modules: A “pattern language” for development and evolution of multicellular form. International Journal of Developmental Biology 53:693–705.

Olby, R. 1985. Origins of Mendelism. 2nd ed. Chicago: University of Chicago Press.

Oyama, S. 1985. The Ontogeny of Information: Developmental Systems and Evolution. New York: Cambridge University Press.

Parnes, O. 2007. On the Shoulders of Generations: The New Epistemology of Heredity in the Nineteenth Century. In Heredity Produced: At the Crossroads of Biology, Politics, and Culture, 1500–1870, edited by S. Müller-Wille and H.-J. Rheinberger, 315–347. Cambridge, MA: MIT Press.

Paley, W. 1821. Natural Theology; or, Evidences of the Existence and Attributes of the Deity, Collected from the Appearances of Nature. London: Richardson & Co.

Penrose, L. 1959. Self-reproducing machines. Scientific American 200:105–114.

Schank, J. C., and W. C. Wimsatt. 1988. Generative Entrenchment and Evolution. In PSA 1986, vol. 2, edited by A. Fine and P. K. Machamer, 33–60. East Lansing, MI: The Philosophy of Science Association.

Scherer, L., J. Rossi, and M. Weinberg. 2007. Progress and prospects: RNA-based therapies for treatment of HIV infection. Gene Therapy 14:1057–1064.

Sober, E., and D. Wilson. 1998. Unto Others: The Evolution and Psychology of Unselfish Behavior. Cambridge, MA: Harvard University Press.

Szathmáry, E., and J. Maynard Smith. 1997. From replicators to reproducers: The first major transitions leading to life. Journal of Theoretical Biology 187:555–572.

Wade, M. 1976. Group selection among laboratory populations of Tribolium. Proceedings of the National Academy of Sciences of the United States of America 73:4604–4607.

Wade, M. 1977. An experimental study of group selection. Evolution; International Journal of Organic Evolution 31:134–153.

Wade, M. 1978. A critical review of the models of group selection. Quarterly Review of Biology 53:101–114.

Wade, M., and J. Griesemer. 1998. Populational heritability: Empirical studies of evolution in metapopulations. American Naturalist 151:135–147.

Williams, G. 1966. Adaptation and Natural Selection. Princeton: Princeton University Press.

Wilson, D. 1980. The Natural Selection of Populations and Communities. Menlo Park, CA: Benjamin/Cummings.

Wilson, D., and L. Dugatkin. 1997. Group selection and assortative interactions. American Naturalist 149: 336–351.

Wimsatt, W. 1986. Developmental Constraints, Generative Entrenchment, and the Innate–Acquired Distinction. In Integrating Scientific Disciplines, edited by W. Bechtel, 185–208. Dordrecht: Martinus-Nijhoff.

Wimsatt, W. 2001. Generative Entrenchment and the Developmental Systems Approach to Evolutionary Processes. In Cycles of Contingency: Developmental Systems and Evolution, edited by S. Oyama, P. Griffiths, and R. Gray, 219–237. Cambridge, MA: MIT Press.

Wimsatt, W., and J. Griesemer. 2007. Reproducing Entrenchments to Scaffold Culture: The Central Role of Development in Cultural Evolution. In Integrating Evolution and Development: From Theory to Practice, edited by R. Sansom and R. Brandon, 227–323. Cambridge, MA: MIT Press.

Wolf, J., E. Brodie, III, and M. Wade. 2000. Epistasis and the Evolutionary Process. New York: Oxford University Press.