PHILOSOPHY OF SCIENCE
Direttore Silvano T Università degli Studi di Sassari
Comitato scientifico Jesús Timoteo Á Universidad Complutense de Madrid
Dario A Libera Università Internazionale degli Studi Sociali “Guido Carli” (LUISS) di Roma
Roberto C Sapienza Università di Roma
Roberto G Università degli Studi di Cagliari
Amit H Indiana University
PHILOSOPHY OF SCIENCE
This edition series is based on two guidelines. Separation lines between disciplines exist for the pleasure of crossing them and that this intrusion demand is stronger than any imposed limitations to free interaction and dialogue between fields of knowledge. Yesterday such trespassing need was true for the theories of Copernico and Darwin. Today it is valid for cosmology, biology and physics, including computer science and high technology and the boundaries set between them. The other guideline is that the most interesting philosophy, as Ludovico Geymonat liked to say, is found hidden in the creases of science. We must look in various articulations and directions of science, across any artificial boundary between “science of nature” and “human science” to find adequate and trustworthy replies to questions philosophy meets in its path. In this general picture the single items discussed remain with many metaphorical question marks, to stimulate, as says Wittgenstein in Pensieri Diversi: “With my frequent punctuation I wish to slow down the reader’s rhythm. Because I would like to be read slowly”. These are not ‘disposable’ texts to view rapidly and in haste. We suggest you read further making yours a maxim said to be of Svetonio, an invitation to reflection: “Festina lente”.
Marta Bertolaso How science works Choosing levels of explanation in biological sciences Preface by Sandra D. Mitchell
Copyright © MMXIII ARACNE editrice S.r.l. www.aracneeditrice.it
[email protected] via Raffaele Garofalo, /A–B Roma ()
----
I diritti di traduzione, di memorizzazione elettronica, di riproduzione e di adattamento anche parziale, con qualsiasi mezzo, sono riservati per tutti i Paesi. Non sono assolutamente consentite le fotocopie senza il permesso scritto dell’Editore. I edizione: giugno
Contents
Preface by Sandra D. Mitchell
Chapter I Introduction .. Which science?, – .. Structure of the book, – .. Acknowledgments, .
Chapter II Cancer and cancer research .. Introduction, – .. Cancer as a complex phenomenon, – .. Pluralism of interpretative models, – .. A multi–level approach in cancer research, – References, .
Chapter III Are complex natural systems necessarily decomposable hierarchies? .. Introduction, – .. The hierarchical explanatory model of cancer, – .. Near–decomposability as a default state for cancer, – .. Formal requirements that are not accomplished in the model, – .. Why is the notion of differentiation crucial in solving the apparent paradoxes, – .. What does differentiation depend upon? , – .. Conclusions, – References, .
Chapter IV Why reductionism is not impossible, just (sometimes) impassible .. Introduction, – .. Partial reductions, – .. What do mechanisms explain?, – .. Cancer stem cells, – ... Problems with Cancer Stem Cells, – ... Proximate causes, – .. Impassible reductionism, – .. Reductions in scientific practice, – .. Conclusions, – References, .
Contents
Chapter V What does the context matter? .. Introduction, – .. Framing the picture, – .. Reducing biological explanations, – .. Explaining biological behaviour, – .. Explaining why a Tumour Cell behaves in that way, – .. Conclusions, – References, .
Chapter VI Pluralism out of unification .. Introduction, – .. Integrative pluralism, – .. Biological contingency and context dependency, – .. Epistemological consistency of integrative pluralism in biological sciences, – References, .
Chapter VII How does science work? .. Choosing the explanatory level, – .. Scientific Practice, .
Preface by S D. M
Complex biological systems, like the human body, function normally by the operation of multiple components each engaging multiple causes at different levels of organization in changing internal and external environments. Explaining scientifically how this works has generated a variety of theories, models and explanations. Some theories explain complex behavior by appeal only to properties of the simplest molecular components, others by appeal to system level properties and the interactions and processes that stabilize them. The first approach exhibits what philosophers of science have identified as reductive methodologies, the second to developmental, emergent, or systems methodologies. What all can agree upon is that the pluralism of biological explanations does not seem to be diminishing, resolving into a grand unified theory or reducing into some branch of chemistry or physics. And so the philosophical debate continues as well as to whether reductionism should, normatively, be the goal of science or not, whether unification or disunity should, normatively, characterize the relationship among different explanations or not. In my own work on these topics I have argued against reduction and unification as obligate goals and defended a position I call integrative pluralism instead (Mitchell , ). I appeal to scientific investigations of complex biological behavior in practice as well as provide arguments for contingency, emergence, partiality of scientific representation and pragmatic components of scientific explanation in developing and defending integrative pluralism. Bertolaso is concerned with an overlapping set of issues that arise from close attention to the details of cancer research. Understanding cancer, the targeted science for philosophical analysis in Marta Bertolaso’s book, adds another source of complexity
Preface
to the story. As Tolstoy brilliantly put it “Happy families are all alike; every unhappy family is unhappy in its own way.” While this does not map literally onto functional and dysfunctional behaviors (that is, a functioning complex biological system may have alternative ways to realize its “happy” state) it seems to me that prima facie there are many more ways to fail to function than to succeed. Thus explaining the origin, progress, and pathways of cancer presents even greater challenges for any simple explanatory strategy. As Bertolaso explains (Chapter ), the heterogeneity of cancer tumors eludes explanation by simple models of cancer progression. This heterogeneity has led to the development of explanations that appeal to higher levels of organization. With the details of these scientific developments at hand, Bertolaso interrogates both a mechanistic account of explanation (like that of Machamer, Darden and Craver ()) and Schaffner’s preferred causal model systems (a) to argue “that if reductionist-mechanistic explanations work it’s because of the non-reductionist dimension that characterizes the definition of their relata.” Throughout the book Bertolaso refers to what she calls the “double dimension” of describing (or defining) a biological behavior and causally explaining it. She argues that the two activities are not serially independent, but rather engaged in a dialectic of revision and refinement in the experimental practice of scientists. In cancer research, providing a definition of what kind of process it is has been intimately linked with different explanatory strategies. In examining the relationship between a reductionist approach (Somatic Mutation Theory) and a system-level approach (Tissue Organization Field Theory) in Chapter , Bertolaso draws out general philosophical issues which motivate the chapters in this book. What are the implications for reductive strategies of the hierarchical structure that is evident in cancer ontogenesis from gene, to cell, to tissue, to organ to organism? How does the role of context, often invoked to defend emergent properties and explanations in contrast to reductive ones play out in the case of cancer? By focusing on the scientifically and philosophical salient issue of which level of explanation or description is most appropriate to the study of complex biological behavior, and using developments in cancer
Preface
research as a source for answering this question, Bertolaso provides new entries into ongoing philosophical debates, as well as opening up new questions to engage the philosopher of biology. Sandra D. M Department of History and Philosophy of Science University of Pittsburgh
Chapter I
Introduction
“It is a common idea that some choices of level of explanation or causal description are more appropriate or perspicuous than others, although there is little consensus about what exactly this means” (Woodward , ). This is the introductory statement of a well–known paper of Jim Woodward that caught my attention in the last two years and has driven my research during my visits at the University of Pittsburgh. In this volume I collect some papers that answer that question in different ways. I have presented some of them in various conferences and I discussed these issues both in the States and in Europe. Colleagues and friends are now encouraging me to publish them all together, so that I have finally decided to share them with a wider group of readers. I have collected suggestions and have based some chapters on articles and a book already published. I have also integrated some chapters with footnotes whenever I have received comments on them in conversations with people I mention in the acknowledgments: they will be useful to further develop and articulate the discussion I am opening in these pages. I will be more than happy if others, who have already explored these issues, will be able to contribute to the discussion from these or different perspectives. This volume is a programmatic contribution towards a clarification of why science works and how it works. In these decades, there is a trend towards a new Philosophy of Science, which is much more focused on how science works in practice and interested in the epistemological implications of scientific explanations as a means to understand the natural world. I have already addressed this philosophical claim , and . Woodward J. (), Causation in biology: stability, specificity, and the choice of levels of explanation, Biology and Philosophy, : –. . Bertolaso M. (), Il Cancro come questione. Modelli interpretativi e presupposti epistemologici, FrancoAngeli, Milano.
How science works
I would like to provide in this volume some theoretical grounds for my (previous) viewpoint. I will be happy to receive comments and answers to the open questions I present at the end of the book. And to integrate and change my mind whenever different perspectives will prove to be more fruitful in accounting for the same basic scientific questions. .. Which science? When I was studying biology at the University of Milan I was told that the term biology was to be understood as the combination of bios, i.e. life, and logos, i.e. a principle of order and knowledge. Our training in ancient philosophy was such that none of us was actually surprised by this explanation and by the comments that some of our professors made when acknowledging that biology is not the only field of knowledge that deals with life. Life is a field for philosophical inquiry that grasps aspects of this concept that the formalisms and reductionist approach of scientific methodology can only partially address. Philosophical discussions often integrated our lessons and conversations at lunchtime and over summer school periods. This approach to the biological world changed when I started studying philosophy. The strong commitment to believe that the scientific method, and the peculiarity of its explanatory enterprise, provided the explanatory framework for biological phenomena, which I was encouraged to embrace, contrasted strongly with my experience in lab activity. There I realized that the first challenge was rather to design the right experiment depending on the scientific problem. The relevant issue was to ask the right questions and to set the adequate experimental control to generate significant data from the experimental tools available to us. In that period I realized that a double dimension with philosophical relevance was at stake: one related with experimental design and the other with the conceptualization and the explanatory relevance of the notion of life. The question that Woodward asked in his paper in gave me an interesting perspective to start from in trying to figure out what epistemological issues were at stake when the question was on some characteristic features of living systems and the adequate
. Introduction
explanatory framework to adopt to understand them. I am assuming that any biological question is characterized by the necessity to explain why something (usually identified as a system, i.e. an integrated functional unity of molecular parts) behaves in this way and not in another one. These kinds of dynamics, defined in terms of biological behaviours, are usually described in terms of inter–level regulatory or control processes. In biology it’s common to distinguish a double aspect of the experimental procedure: the definition of the systems (explananda) and the structuring of the explanation (explanans), which is typically causal. Such double dimension seemed to be overlooked in the philosophical discussion about how science works and how we know the world through science. I then decided to look at when and why research programs that addressed this kind of biological questions, related with an inter–level regulatory process, get stuck in their explanatory enterprise. In explaining a complex biological phenomenon that involves many and different levels of a biological organization, from genes to cells, up to tissues and organs’ functional organization, this inquiry has been particularly interesting. Philosophical questions arise, in my opinion, when we consider that the descriptive element is a first step, necessary although propaedeutic to the discovery of a causal relationship that describes the more specific behaviour of biological systems and subsystems. At the end the question is always on the living aspect of a biological system, i.e. on its peculiar way of being, or behaving. The focus is not on the parts and their causal interactions but on the peculiar dynamics that hold them. Training in biological science consists in developing skills in to bring these two dimensions — description of the system and articulation of causal relationships among its parts — into a unified experimental approach. The scientific outlook we have inherited from the modern era of the Renaissance in Europe is based on the idea that there are regularities and continuities of organizational principles in nature and that searching for such patterns of organization is particularly effective in exploring and understanding living systems. The notion of pattern effectively combines the dynamic aspect with the descriptive features of the systems that constitute the explananda as the object of scientific inquiry in biological sciences. In this volume I will explore some explanatory issues that emerge from my own expe-
How science works
rience in scientific work and my philosophical thought on it. I have to apologize for using some terms in a different way than what is sometimes considered ‘the standard one’ in American literature on these topics. However, I made sure that the meaning I am giving them is consistent and clarified in the papers. Others can discuss them within different philosophical traditions and consider these contributions or other perspectives they have been developing. .. Structure of the book In the next chapter, I introduce the case study. Cancer is one of the biological phenomena, to which I devoted my studies and experimental activity in the lab. The complexity of the neoplastic processes is one characteristic that appeared increasingly evident over the last few decades, both from clinical and molecular studies. Rapid advances in molecular biology have led to the acquisition of a considerable amount of data regarding the genes and proteins that are apparently involved in the progression of cancer, while the reductionist perspective — which has dominated cancer research over the last years and is characterized by the attempt to explain cancer in genetic terms and through mechanistic models of interactions among biological parts — has incorporated the data into ever more detailed and complex interpretive models of the origin of cancer. Nevertheless, an analysis of scientific literature highlights the lack of a formal definition of neoplastic pathology. Such analysis led me to explore the discussion triggered by the emergence of a number of paradoxes that have demonstrated the inadequacy of the reductionist models, particularly in explaining tumour latency and reversibility of the neoplastic phenotype. At the same time, a generalized tendency has emerged to consider cancer as a dynamic process the explication of which requires a systemic approach. In this framework, taking a certain distance from the reductionist view, often exemplified by the Somatic Mutation Theory (SMT), a new theory (Tissue Organization Field Theory, TOFT) and some new interpretive models show a clear movement toward the organic perspective. The antireductionism that characterizes these models is, to some extent, due to the historical opposition that TOFT’s authors elaborate against SMT in scientific
. Introduction
literature but also reflects the more general opposition between reductionism and holism at the heart of contemporary Philosophy of Science in general, and of Philosophy of Biology in particular. As hierarchical issues seem to be involved by different explanatory models of cancer, in the third chapter I confront the way they are formulated within cancer research with the well–known account of Herbert Simon on the hierarchical structure of complex and evolvable systems. The history of cancer research shows that one of the features of cancer’s complexity is that tumour heterogeneity compromises the hierarchical control of the organic system. Therefore, although over the past decades the dominant paradigm has been cell centred, more recently the dynamics of cancer development and tumour cells’ heterogeneity are captured by explanatory models of cancer referred to hierarchical structures. Considering that, starting from Simon’s relevant paper, hierarchical organization of evolving systems has been the subject of a consistent debate in philosophical literature as well: I then analyse to what extent neoplastic and metastatic phenomena meet the near decomposability feature of hierarchical organized systems proposed by Simon, and the implications of a hierarchical account of complex biological processes. I further discuss how we should understand biological interactions in order to make sense of hierarchical phenomenology of cancer and of historical evolution of its hierarchical explanatory models. In the forth chapter, I argue that attempts to explain higher–level properties in reductionist–mechanistic terms often fail because they are unable — they are impassible — to grasp the explanatory relevance of generalizations. The argument emerges from previous discussions about reductionism in biological sciences. I contend that requirements for reductions must be revised to explain how science works in practice. I consider examples from cancer research to outline a methodological and conceptual framework for our understanding of what a reduction is and how it works. Next I explore in detail the role that the context argument plays in the structure of biological explanations. I will argue that biological explanations have a peculiar structure, which is context–dependent, and that in particular the acknowledgment of type and token context–dependency contributes greatly to clarify some points of the debate about reductionism in biology. Firstly, I analyse the terms
How science works
of the debate on explanatory reductions focused on biological behaviours. Secondly, I investigate what kind of criticism explanatory models meet when committed to explain biological behaviours and some features of their structure. Thirdly, I discuss the epistemological role the context has in explanatory models of cancer, both reductionist and anti–reductionist. I argue that acknowledging the double dimension of context–dependency can prove fruitful to understand the structure of biological explanations and the debate on the shortcomings of reductionism. In the sixth chapter, I discuss the implications of the analysis of the scientific explanatory accounts of cancer at the beginning of the volume. The rapid evolution of such models, despite advances in cancer research, highlights the increasing effort to account for the complexity of cancer by means of different explanatory models. I then present some elements to show how our understanding of the epistemological issues that emerge in cancer research, requires an integrative approach to the neoplastic process, based on its specific dynamics. Some questions remain open suggesting directions for further research on the philosophical foundations of an integrative approach in biomedical science. In the last section of the book I sum up the main conclusions and leave open some considerations that are programmatic in character and meant to push the debate toward a deeper understanding of the issues related with the hierarchical account of biological systems’ behaviour, with the role that contingency plays in structuring levels of biological organization (ontological level) and in our understanding (epistemological level) of living systems, and with the relevance of the context argument in the explanatory scientific enterprise. .. Acknowledgments I wish to thank who has been encouraging me to explore these issues in important exchanges, allowing me to visit other universities on various occasions. First of all, I thank Alfredo Marcos and Sylvie Menard who, from a philosphical and scientific point of view respectively, followed my first steps in this field. Marco Buzzoni, Silvia Caianiello, Juan José Sanguineti, in Italy, and Sandra D. Mitchell, Jim Woodward,
. Introduction
Jim Lennox, Ken Schaffner in Pittsburgh, and Jean Gayon in Paris have also largely influenced my research and discussion of these topics. Their inspiring papers and litterature have been of great help to outline the main philosophical questions that constitute the framework of this book. This book also profited from many discussions with my co–fellows at Pitt last year, so that a special thank you goes to Kyle Stanford, Maria Kronfeldner and Collin Rice as well.I am also grateful to the anonymous reviewers of this volume who contributed to its final publication. Moreover, some studies by Marjorie Grene, David Hull, Francisco Ayala also helped me to broaden my reflection. They do point toward an interesting form of integration between biology and philosophy. This can stem only from the fact that what really drives research are not specific issues in the respective fields of specialization, but their common aspects, the conditions of possibility of both fields. This kind of convergence is what I am interested in and what I find particularly useful to explore in order to understand how science works and why it works. This will improve our awareness of the powerfulness of scientific enterprise and help assess its role in the wider human challenge to understand our world and our place in it. Finally, I wish to thank my family and friends without whose support this work would not have been possible. The discussions with some of them with different backgrounds — from engineering, to neuroscience, and from ethics to theology — had a great impact on my view of things and on the process of prioritizing the philosophical questions I have tried to address. This work has been developed also as part of the research project funded by the Ministry of Science and Innovation of the Spanish Government; project title: “Change: Semantics and Metaphysics” (ref. FFI--C-). We thank Serenella Perasso for the English revision and editing of this volume.
Chapter II
Cancer and cancer research .. Introduction “A question which has long concerned epistemologists and philosophers of science is whether science is a process which can be expected at some time to reach a terminus” (Duprè , ). This struggle seems to be ongoing within specific areas of science, like cancer research, where scientific debates involve philosophical issues, mainly in terms of the possibility to reduce biological phenomena to molecular terms. However, at the actual state of the art, the debate on to what extent reduction is legitimate –or even possible — in cancer research and in biology in general, is not just related to the question whether empirical science is devoted to explaining diseases in terms of a single objective: molecular parts — mainly genes. In this sense cancer research can eventually converge in a hard–wired explanation in terms of cellular parts and pathways involved in carcinogenesis. This view would imply that the crucial scientific challenge for cancer research is nothing less than developing a definitive theoretical and experimental approach to neoplastic phenomena. To the contrary, scientific literature on cancer is bursting with different explanatory models, all reflecting the common effort to couple findings with the complexity of neoplastic process. Moreover, beyond the trend to integrate data in a systemic approach, there is an interesting ongoing debate. Different theories about carcinogenesis are posing interesting questions on the epistemological presupposition and implication of a systemic approach to a phenomenon of such complexity. While the goal to describe cancer in unified molecular terms is arguably unattainable for cancer, even the possibility and consistence of a systemic approach needs to be clarified. In this introductory chapter, I briefly describe in which terms this issue has emerged in cancer research and highlight the epistemologi
How science works
cal issues related to the possibility of instantiating a systemic approach to account for such a complex phenomenon. New and specific concepts seem to be required to deal with cancer complexity and its specific dynamics. The aspect of reductionism we are confronted with is related more to the relating than to the logical deriving of one explanatory theory from another one (Nagel , Ayala ). To satisfy logic derivability implies that somehow the requirement for reduction has already been satisfied in the first place. As far as unification (of different theories) is envisaged, the hard work is related to the possibility to establish adequate definitions (Schaffner , ). Indirectly, this discussion will underline the inadequacy of reductionism in cancer research and the general need for a different epistemological approach to biological complexity. Intrinsic features of emergent properties seem to set the stage. .. Cancer as a complex phenomenon The world of cancer research, rather than optimistic, seems to be currently overwhelmed by the big questions raised by the clear inadequacy of those explanatory models that look at cancer as to a genetic disease or a disease to be defeated through target therapies. Such therapies aim to destroy tumour cells on the basis of biological markers, both at the level of genes or proteins. These models have their roots in a research program defined, in the last decades, as reductionist ( Jemal et al. ): i.e. as a research program that looks at genes as the privileged explanatory level of the processes and dynamics, characterizing a biological phenotype. Despite decades of great investment, the overall incidence and mortality due to neoplastic diseases has remained almost unchanged (Sporn ). These data raise up more than a suspicion that the current answers on the onset and progression of this disease might be inadequate. Surely these data, along with many others, suggest that the optimism that drove the claim that we would have eliminated the suffering and death caused by cancer by the Year (Cancer Trends ) was, at the very least, premature. Expectations continue to be fuelled by the promises that, by discovering new genes and proteins, we will find and adapt new molecular therapies to block or
. Cancer and cancer research
redirect the pathways involved in establishment and maintenance of the malignant phenotype of tumour cells (Tripathy , Deininger et al. ). Big Pharma has underwritten this research, stirring up a lot of expectations often shared by the general public (see on this issue also Law ). Nonetheless, even at molecular level, cancer appears more than ever like a disease with a high multiplicity of causal and contributing factors, so that it is clear that it can’t be encompassed only from the perspectives of genetic or genomic roles. Aside from the numerous factors that contribute to a perception of cancer as a complex disease, tumour heterogeneity at a molecular, clinical, and cellular level, seems to characterize the real complexity of cancer, highlighting its dynamic dimension. There are no molecular elements that identify cancer or specific steps in its evolution: genesis and progression of cancer are related to a large number of causes and mechanisms. Moreover, recent evidence is making it clear that any explanatory model of cancer has to account for another striking level of heterogeneity: the one that cancer cells show within the same tumour and throughout the tumour’s progression (Hanahan and Weinberg , ; Sottoriva et al. ). The increasing recognition of a specific dynamic component in the neoplastic phenomenon, and of cancer as a process rather than as an event, has certainly contributed to less traditional, yet interesting, points of view that led scientists to rethink the basic assumptions of the current research and practice of oncology, and to develop new and alternative approaches (Sporn, ). These views support perceptions of cancer, such as a case of “blocked ontogenesis” (Potter ), “maladjusted living entities” (Tarin ) or “as caricatures of their corresponding normal tissues” (Dalerba et al. ). From a methodological point of view, new models have been presented pointing out similarities of cancer with other complex biological systems and encouraging a systemic approach to deal with some neoplastic features, e.g. robustness (Kitano ). However, “[t]he emerging complexity of the entire ‘cancer system’ overwhelms us, leaving an enormous gap in our understanding and predictive power” (Hornberg et al. ; cited in Bizzarri et al. ). Cancer’s complexity and the multiple causalities involved in its original and final onset, account for the number of different definitions
How science works
of cancer flourishing in this theme’s literature. Such complexities are also often reflected in the increasing number of molecular elements that integrate the first, simple, explanatory models, so that a general trend towards an integrated and systemic view of the phenomenon is justified (Bertolaso a). Moreover, although the multiplicity of definitions of cancer given in this literature would suggest a complexity of this disease that might not be overcome, this output seems to contradict both common sense and clinical practice that continues to identify all cases of cancer with the same term, cancer, and as a sole class of phenomena. The epistemological analysis of the evolution of interpretative models encourages a deeper consideration and understanding of the meaning of this general trend to incorporate a systemic dimension in the explanatory arguments. Beyond different clinical manifestations and descriptive possibilities gained through new biotechnologies, a common denominator that specifies cancer’s behaviour does remain, and is supported by the analogy between some aspects of the neoplastic process and other complex biological processes, such as morphogenesis and development. .. Pluralism of interpretative models “In spite of advances since the s, the situation in biology remains characterized by tensions between two approaches that take different stands on the possibility of reducing biology to physics: a reductionist approach focused on individual molecules and a more systemic approach (. . . ) to study higher levels of organization” (Mazzocchi , ). The trend towards integration of numerous experimental data seems a hallmark of the history of research in the last years. However, this takes place on two constitutively different courses: the first one, i.e. reductionist, tries to put together the parts in the system in order to study the interactions and hopes to obtain a more unified vision of the phenomenon of cancer; a second one, i.e. system–based, reviews the assumptions on which research on cancer began, and consequently changes the perspective to approach the problem. Moving from the clonal genetic model, which attributed the origin of cancer to a single somatic mutation in a cell, over the last decades the neoplastic progression has been described as a progression of
. Cancer and cancer research
stages defined by mutations in oncogenes or tumour suppressor genes (Fearon and Volgestein ). The initial version was later integrated into the Epigenetic Progenitor model because of other non–genetic factors found to be equally involved in the process, like DNA methylation of the promoters of some genes or histone acetylation, which also have regulatory functions in gene expression during the process of cellular differentiation. Not only has understanding of the sequence of genetic and epigenetic events been growing but, interestingly, research has focused on the description of the relationship between different molecular components. Such relationships have been enriched with more and more details in an attempt to construct an integrated cell circuit (Hahn and Weinberg ) whose dynamics should explain the specific behaviour of cancer cells. While the molecular components and their interactions remain virtually unchanged, their functional activity changes in response to internal and external factors that ultimately involve multiple levels of DNA damage, such as alterations in gene expression by other mechanisms (Vogelstein and Kinzler , Jones and Baylin ). Despite this effort to recover a more systemic view, at least at the cellular level, the dominant reductionist paradigm seems to omit an important aspect of carcinogenesis for which there is still no satisfactory explanation. Among other features, there are the relevance of time in the final onset of cancer (e.g. metastasis development depends on organismic factors and not only genetic and molecular features of tumour cells); the instability of tumour cells’ phenotype; in many cases, the spontaneous regression of some tumours and, thus, mainly their contextual functional dependence (Baker and Kramer ). From an empirical point of view, then, the true challenge seems to be related with the understanding of the system’s behaviour, its stability or instability (Heng et al. ) more than mutation as stated by the Somatic Mutation Theory (SMT) of cancer. Studies have moved from genetic to genomic level. Patterns identified at this level are dependent on the genomic and micro environment: by changing the environmental context, a specific pattern may become so rare as to be considered “noise”, that is no longer essential for a process to happen and vice versa (Hillenmeyer et al. ). From a system point of view, karyotypic changes that can be identified in tumour cells represent, to some extent, stable functional states, even though gene mutations and
How science works
epigenetic changes can influence the karyotypic pattern. This might explain why the majority of patients show a wide range of mutations for the same tumour, while mutations do not appear to have a significant value from an epidemiological point of view (cfr. Sporn ). The contextual dimension specifies at least one aspect of the specific dynamic of the neoplastic process. Other authors, who explain the neoplastic phenotype in terms of system–level dynamics, also share the interest for the dynamic properties of the system (Huang and Ingber ; Ingber ). Their models seem to match much better with known facts, like the neoplastic phenotype can return and metastasis can actually be dormant for a long time rather than the reductionist perspective characterized by genetic determinism. The analysis of regulatory networks of genes seems to be a useful tool to explain changes from the cellular phenotype to the neoplastic one. These changes might occur through dynamic transitions of networks that can be described in terms of attractors and epigenetic landscape. This approach overcomes the limits of genetic determinism, emphasizing the relevance of non–genetic components in cellular mechanisms. The novelty of this approach is that it expands the systemic perspective not only in the space–time dimensions but by including different kinds of factors playing a role in the neoplastic onset. Furthermore, this kind of perspective is also highlighting the important role of the functional context of cells within a tissue, described in terms of cellular shape, form and tissue architecture. The failure of the reductionist approach to explain cancer merely in molecular terms, i.e. by looking for new key molecules responsible for the process or for some of its crucial steps, has emphasized the need to extend previous explanatory models. Such extension has been realized through new biological concepts, which involve a different dimension (i.e. time and functional states) with respect to molecular parts, and helped overcoming the tension among seemingly disparate properties of neoplastic phenomena, such as loss of functional specificity or changes in the properties of cellular and tissue plasticity and robustness. To account for this intrinsic dynamic aspect that characterizes tumours, some researchers began to take more seriously the similarities and analogies between the neoplastic phenotype and developmental and morphogenetic processes (Potter ; Rudden ). The need
. Cancer and cancer research
to explain the complexity of cancer through a systemic view has been framed within an organicistic perspective. Explaining the space and temporal properties of cancer is not only a problem of integrating all the available information to cope with biological complexity (Sonnenschein and Soto ). Instead a more comprehensive systemic account of the neoplastic phenomenon requires novel ways to think about networks where non–linearity is the rule, not the exception, as in the case of biological processes and functions. To think in terms of contextual conditions and their relationship with a specific phenotype a different epistemological perspective is needed. Such perspective presupposes the hierarchical organization of the living organism. This has been explicitly stated by the authors of the Tissue Organization Field Theory (TOFT) of cancer, who adopt an explicit emergence conceptual framework, whereby emergence refers to the fact that as we move ‘up’ along levels of complexity, we find new, unpredictable properties (Mayr ; Soto and Sonnenschein ). In contrast with the dominant reductionist paradigm, they shift the emphasis from the dynamics at the cellular, sub–cellular or molecular levels, to the organization of tissue. According to their view, carcinogenesis disrupts the three–dimensional and organizational structure connecting the stroma and the parenchyma, mediated by cell–cell interactions. In this perspective, therefore, carcinogens might not be directly responsible for neoplasia by inducing genetic mutations (Sonnenschein and Soto ). A chain of cellular miscommunication is a slow and subtle feedback chain of changes that generate even more changes. Hence, carcinogenesis and neoplasia would occur once the signals that maintain normal organization are disrupted: a developmental process gone astray (Soto et al. ). The organizational and historical dimensions of the tumour study caught up in this developmental approach, have been described in terms of morphogenetic fields that are compromised in the processes ultimately responsible for the neoplastic phenomenon (Maffini et al. ). The Dynamic Reciprocating Model explains a specific feature of this miscommunication. Tumorigenic context dependence has been demonstrated studying dynamic integration of tissue architecture and function at different levels of biological organization (e.g. cell membrane, cellular cytoskeleton, etc.) that ultimately drive or compromise tissue–specific gene expression as well (Bissell et al. ). Dynamics
How science works
underlying carcinogenesis are reciprocal (Xu et al. ) and regulatory inter–level mechanisms are involved, while new properties emerge at different scales that are real and autonomous. Also in the biology of cancer, then, there are good reasons for classifying objects in terms of properties that have some considerable stability over time (Duprè ). Moving from these premises, the evolution of cancer may be considered, in its first step, as a phenomenon related to the loosening of biological constraints (Bertolaso b). Such constraints are related with the organization of a biological system that guarantees its functional stability over time (Bertolaso ), account for the hierarchical control of the organisms, and the multi–scale phenomenology of cancer. The systemic approach required to study the specificity of the biological organization emerges at the crossroads of the peculiarity of this organization and our way of conceptualizing it through different methodological approaches. Loss of interdependency of the parts has its rationale in a different understanding of the nature of interactions between the parts that belong to the same level and the effects on higher order levels than the mechanistic one (see also Chapter ). .. A multi–level approach in cancer research Before confronting a multi–level approach and the importance of context, we will start by considering a series of questions that crop up in experimental research like oncology. Mainly, we must consider the following matters: — The impossibility of identifying a causal relationship between molecular parts, or a linear sequence of molecular alterations in the neoplastic phenotype. — The evidence that the neoplastic process involves different levels of biological organization. — The evidence that the manifestation of the phenotype is strictly dependent on the contextual conditions in space and over time. Furthermore, these assumptions justify that the interpretative models of the neoplastic process have undergone an integration or
. Cancer and cancer research
re–definition in terms of system biology, involving a hierarchical approach to the problem. The dogma of molecular biology fails when faced with a phenotype that is not completely determined by its genotype and with the fact that genetic factors do not have the full responsibility for a heritable illness. The optimism of molecular biology was quickly tempered by obstacles in explaining many rich features, like functional redundancy, metabolic robustness, multi–functional proteins, etc., that started emerging as molecular biological research advanced. One same protein could be demonstrated to carry out multiple roles in the same organism in different cell types (Gilbert and Sarkar ) or a molecular signal could induce the expression of different genes, or even induce disparate cellular differentiation programs in different cell lines (Brisken et al. ). Phenomena like morphogenesis, immunological and oncological processes present some of the best–studied examples of this multiplicity. The question therefore shifts from the existence and nature of biological laws to the type and nature of pattern recognition and regularity. My suggestion is that emphasis on the biological complexity of cancer and other similar phenomena, is not only connected with developmental dimension of the organism and interactivity among parts, but with the epistemological status of the emergent properties and their features: the dimensions involved in the inter–level regulation characteristic of any biological process. The question about the functional dynamics that govern each level, the functional definition of its explanatory terms and their intrinsic conceptual relationship therefore become an interesting issue to explore. I believe that discussion about emergent properties and their relationship with proximate and evolutionary explanations in biology (Mayr , Hull , Emmeche ) is better understood from the perspective of the explanatory role that the context plays in biological explanations. In cancer research, for example, both these dimensions — proximal and evolutionary — have a common explanatory background in the specificity of the microenvironment where the organizational disruption of the tissues is observed and new functional properties appear. Finally, by addressing biological system behaviour we are implicitly assuming that living beings respond to the same inputs at different
How science works
levels, that molecules can enter into different pathways with different functions, and that pleiotropic responses in nature are the norm and not the exception. The relationships between biological factors at various levels are not independent of, nor indifferent to, each other. The behaviour of complex systems is dependent on the context, as defined by internal and external conditions that can affect the system. It is not surprising that any attempt of molecular studies to grapple with the biological complexity of cancer, as organicistic explanatory theories do, actually refer to a multi–level organization of the biological system to explain tumour heterogeneity and its temporal dynamics as well. Currently a number of new studies indicate that the developmental limitations of tissue–specific stem cells are regulated by the microenvironment and that host cells, under specific conditions such as tissue injury or infection, are able to provide specific signals that counteract these restrictions (Nelson et al. ; Mueller and Fusenig ). Different studies focus on the fate and function of stem cells, which are governed by a combination of intrinsic determinants and signals from the local microenvironment or niche, i.e. the germinal compartment in different tissues, able to assure cellular turnover. The Epithelial Mesenchymal Transition, a program of differentiation and organization of cells mainly characterized by loss of cell adhesion, and increased cell mobility, has also been included among the mechanisms which could account for the tumour cell invasiveness (Kalluri and Weinberg ). However, the crucial question is still about “[t]he distinction between explaining how something does what it does and explaining what it does” (Dupré , ). That is why TOFT, by assuming the emergence approach as default, can identify a first level of analysis where the phenomenon appears, and address it by means of biological concepts such as morphogenetic field. In our analysis, moving from evidence of emergent properties and their intrinsic feature, we find a reciprocal dynamic mixed in a hierarchical organization in which the notion of functional field and architectural structure as causal dimension of the neoplastic phenotype give a sense of the inadequacy of ontological reductionist stances. This entanglement can never be totally spread out in lower–level explanations, because in the ultimate through–and–through lower–level explanation one might never know what higher–level phenomenon it explained.
. Cancer and cancer research
From an epistemological perspective, such description shows the limitation of reductionism applied to the complexity of biology and suggests the opportunity to keep at every level what is most meaningful. “A more holistic, hierarchical approach to carcinogenesis therefore yields any observations that are difficult to explain from a purely reductionist perspective” (Root–Bernstein , ). Each level of the biological organization has unique features through which the structure and information of one level are re–interpreted at a superior level. To ignore this, means to lose sight of the basic principles of hierarchic structure and to depart from the explanation of their natural origin, which is not molecular but systemic, i.e. organic. “Scientific knowledge is organized in levels, not because reduction in principle is impossible, but because nature is organized in levels, and the pattern at each level is most clearly discerned by abstracting from the detail of the level far below. (. . . ) And nature is organized in levels because hierarchic structures (. . . ) provide the most viable form for any system of even moderate complexity” (Pattee , –). References A F.J. (), Biology as an autonomous science. American Scientist, : –. B S.G., K B.S. (), Paradoxes in carcinogenesis: new opportunities for research directions. BMC Cancer, : –. B M. (a), Towards an Integrated View of the Neoplastic Phenomena in Cancer Research. History and Philosophy of the Life Sciences, : –. ——— (b), The neoplastic process and the problems with the attribution of function. Rivista di Biologia / Biology Forum, : –. ——— (), Breaking down levels of biological organization. Theoretical Biology Forum, accepted. B M.J., R D.C., R A., W V.M., P O.W. (), The organizing principle: microenvironmental influences in the normal and malignant breast. Differentiation, : – . B C., S M., L H.F., W R. (), The signaling domain of the erythropoietin receptor rescues prolactin recep-
How science works
tor–mutant mammary epithelium. Proceeding of the National Academy of Sciencies of the USA, : –. Cancer Trends Progress report (), http://progressreport.cancer.gov (last access: ), D P., C R.W., C M.F. (), Cancer stem cells: models and concepts. Annual Review of Medicine, : –. D M., B E., D B.J. (), The development of imatinib as a therapeutic agent for chronic myeloid leukemia. Blood, : –. D J. (), The disunity of science. Mind, : –. ——— (), The disorder of things: metaphysical foundations of the disunity of science. Harvard University Press, Cambridge. ——— (), It is not possible to reduce biological explanations to explanations in chemistry and / or physics. In Contemporary debate in philosophy of biology. Edited by Ayala J and Arp R. Wiley–Blackwell, Oxford. E C. (), Aspects of complexity in life and science. Philosophica, : –. F E.R., V B. (), Genetic model for coloreotal tumorigenesis. Cell, : –. G S.F., S S. (), Embracing Complexity: organicism for the st century. Developmental Dynamics, : –. H D., W R.A. (), The Hallmarks of cancer. Cell, : –. ——— (), Hallmarks of cancer: the next generation. Cell, :–. H W.C., W R.A. (), Modelling the molecular circuitry of cancer. Nature Reviews Cancer, : –. H H.H., B S.W., S J.B., Y K.J., L G., Y C.J. (), Genetic and epigenetic heterogeneity in cancer: a genome–centric perspective. Journal of Cellular Physiology, : – H M.E., F E., W J., P S.E., H S., L W., P M., S. O R.P., T M., K D., A R.B., D R.W., N C., G G. (), The chemical genomic portrait of yeast: uncovering a phenotype for all genes. Science, : –. H J.J., B F.J., W H.W., L J. (), Cancer: a systems biology disease. BioSystems, : –.
. Cancer and cancer research
H S., I D.E. (), Shape–dependent control of cell growth, differentiation, and apoptosis: switching between attractors in cell regulatory networks. Experimental Cell Research, : –. H D.L. (), The Metaphysics of evolution. Baltimore: John Hopkins University Press. I D.E. (), Can cancer be reversed by engineering the tumor microenvironment? Seminars in Cancer Biology, : –. J A., S R., W E., M T., X J., S C., T M.J. (), Cancer statistics, . CA: A Cancer Journal for Clinicians, :–. J P.A., B S.B. (), The epigenomics of cancer. Cell, : –. K R., W R.A. (), The basics of epithelial–mesenchimal transition. Journal of Clinical Investigation, : –. K H. (), Cancer as a robust system: implication for anticancer therapy. Nature Reviews Cancer, : –. L J. (), Big Pharma, Einaudi, Torino. M M.V., C J.M., S A.M., S C. (), Stromal regulation of neoplastic development: age–dependent normalization of neoplastic mammary cells by mammary stroma. American Journal of Pathology, :–. M E. (), The Growth of Biological Thought: Diversity, Evolution, and Inheritance. Belk– nap Press, Cambridge. M F. (), Complementary in biology EMBO Report, : –. M M.M., F N.E. (), Friends or foes — bipolar effects of tumour stroma in cancer. Nature Reviews. Cancer, : –. N E. (), The Structure of Science. Harcourt, Brace and World, New York. Nelson W.G., De Weese T.L., De Marso A.M. (), The diet, prostate inflammation, and the development of prostate cancer. Cancer Metastasis Reviews, : –. P H.H. (), Hierarchy Theory — The Challenge of Complex Systems. George Braziller, New York. P V.R. (), Phenotypic diversity in experimental hepatomas: the concept of partially blocked ontogeny. The th Walter Hubert Lecture, British Journal of Cancer, : –. R–B R.S. (), Complementarity and contradiction in can-
How science works
cer research: the role of hierarchies in carcinogenesis. Anticancer Research, : –. R R.W. (), Cancer biology. Oxford University Press, New York. S K.F. (), Reductionism in Biology: prospects and problems. Choen RS (ed.),. PSA, Dordrecht: Reindel , pp. –. ——— (), Theories, models and equations in systems biology. In Systems Biology: Philosophical foundations. Boogerd FC, Bruggeman FJ, Hofmeyer J–HS, Westerhoff HV (ed.), Elsevier. S C., S A.M. (), Why system biology and cancer? Seminars in Cancer Biology, :–. ——— (), The Society of Cells: Cancer and Control of Cell Proliferation. Springer–Verlag Inc, New York. ——— (), “Are Times a ‘Changin’ in Carcinogenesis?”, Endocrinology, : –. S A.M., M M.V., S C. (), Neoplasia as development gone awry: the role of endocrine disruptors. International Journal of Andrology, : – . S A., V J.J., B T., MW S.K., N L., M J.P., S P.M., V L. (), Cancer stem cell tumor model reveals invasive morphology and increased phenotypical heterogeneity. Cencer Research, :–. S MB (), Dichotomies in cancer research: some suggestions for a new synthesis. National Clinical Practice Oncology, : –. T D. (, ), New insights into the pathogenesis of breast cancer metastasis. Breast Disease, : –. T D. (), Targeted therapies in breast cancer. The Breast Journal, : S–. X R., B A., B M.J. (), Tissue architecture and function: dynamic reciprocity via extra — and intra — cellular matrices. Cancer Metastasis Reviews, : –. V B., K K.W. (), Cancer genes and the pathways they control. Nature Medicine,: –.
Chapter III
Are complex natural systems necessarily decomposable hierarchies? .. Introduction Cancer is commonly recognized and defined as a complex disease. Besides its epidemiological and clinical manifestations, many different genes and molecular pathways are involved and different levels of the biological organization are compromised during its development, from the genetic to the cellular and tissue level. However, the history of cancer research shows that one of the features the complexity of cancer is related with is tumour cells’ heterogeneity, which is common in different kinds of cancer and within the same cancer as well. Morphological and architectural differences between tumours and healthy tissue are used in the histological analysis by optical microscopy, and so far represent the essential parameters in clinical practice to define the anatomical origin of a tumour. As we have seen in Chapter , reference to a hierarchical organization has been often presented in literature to account for tumour heterogeneity and its complexity in recentyears so that tumours have been compared to abnormal organs that act as caricatures of their corresponding normal tissues (Dalerba et al. ). Cancer is mainly understood in terms of disruption of a hierarchical organization of tissues and cells. Moreover, at a microscopic level tumours do not appear merely as aggregates of malignant cells, but as maladjusted living entities that recruit host stromal cells such as fibroblasts or endothelial cells, with which they interact (Tarin ). The result is often the formation of an expanded and distorted but recognizable structure of the histology of the organ from which tumours are derived that invades normal tissue. This evidence adds another feature of tumour heterogeneity: “An important source of intratumoral heterogeneity
How science works
has been revealed by the discovery that populations of cells within a tumor, like those in the corresponding normal tissues, are organized hierarchically” (Bonnet ; Al–Hajj , cit. in ). Over the past decades, the dominant paradigm has been cell centred and cancer has been explained in terms of cells that proliferate and eventually metastasize. Cancer cells compromise the normal hierarchical organization of a tissue or of an organ and are considered to be, to some extent, autonomously able to move, invade and grow within different environments. To cope with these dynamic features of cancer and its heterogeneity a Hierarchical Model has been proposed (Vescovi ). The dynamics of cancer development, the heterogeneity of cancer cell composition, and the hierarchical organization that mimics normal tissue structure, in short, cancer complexity, can be best understood in an integrated framework of a cell differentiation. Cancer disrupts the hierarchical functional organization of the organism and, at the same time, cells in a tumour mass mimic a hierarchical organization although aberrantly (Bertolaso c). Hierarchical structures of natural systems have been used, also in philosophical literature, to characterize their complexity, with near decomposability as one of their main features (Simon, ). When talking about the nearly–decomposability” of hierarchical systems, Simon refers to systems “in which the interactions among the subsystems are weak, but not negligible” (Simon , ) (see Section .). In this chapter I propose to analyse to what extent the neoplastic and metastatic phenomena meet the near decomposability feature of hierarchical organized systems proposed by Simon and what implications this has for a hierarchical account of complex biological systems. I argue that this kind of hierarchical structure alone is insufficient to capture the complex features of cancer, as it encompasses only some mechanistic steps of the neoplastic process. In exploring the explanatory adequacy of the hierarchical structure new dimensions of cancer complexity are exposed. “Are complex natural systems necessarily decomposable hierarchies?” . The biology of cancer gives us insight to claim that this is not the case, and the reason is mainly related with the assumption that in a nearly decomposable system its . With this question Bechtel and Richardson conclude their considerations about Hierarchies and Organization (Bechtel and Richardson ).
. Are complex natural systems necessarily decomposable hierarchies?
components might exist as stable units in their own right, within a biological hierarchical organization. The argument outline will compare the organizational properties of cancer and of hierarchical systems. The chapter is organized as follows. In Section ., I briefly present the Hierarchical explanatory Model of cancer. In Section ., applying Simon’s notion of near–decomposability I will describe in formal terms the character of cancer and metastasis. In Section ., I present empirical evidence that seems to contradict the formal features of cancer as a near decomposable hierarchy. In Section ., I will discuss why the notion of differentiation is crucial in solving the apparent paradoxes. I will argue that an emphasis on differentiation in explaining the neoplastic process might highlight some crucial features of the hierarchical organization and that those features are required for the maintenance and stability of the system. I will argue that a hierarchical explanation of cancer cell behaviour is just a heuristic tool, useful to some extent but not able to capture the real specificity of the neoplastic phenotype and thus to explain its characteristic features and development. Finally, in Section ., I clarify which kind of interactions have to be taken into account when explaining a biological phenomenon supporting the argument with evidences from other, organism centred, explanatory models of cancer. .. The hierarchical explanatory model of cancer The dominant cell centred perspective, that has dominated cancer research over the last years, considers carcinogenesis as a progressive deregulation of gene functions, whether caused by mutation, deletion, amplification, translocation, or some other mechanism, leading first to clonal expansion. Thus it was assumed that both the origin and progression of the tumour are mediated by a sequence of several molecular causes, ultimately responsible for the neoplastic phenotype (Hanahan and Weinberg ). Clonal heterogeneity of initiated and progressing cells would arise subsequently, preceding local tissue invasion and metastasis. A Multistep Model (Kinzler and Vogelstein , ) was further proposed, which had the advantage of clarifying some genetic mechanisms in the initiation and progression of tumours. Within this explana-
How science works
tory framework, where tissues are envisaged as a complex network of parts that are quasi independent of each other and where the properties of the whole are defined by the intrinsic properties of parts and their interactions among them, the reconstruction of a linear series of steps can explain the development of cancer. Moreover, this model provided premises for successive studies on the neoplastic process in which the tissue and epigenetic components play a fundamental role, supporting a view of carcinogenesis as characterized by errors in proliferation, cellular death or differentiation (Lloyd et al. ): processes that, in these models, rely on intrinsic features of the cell. However, more recent data showed that the majority of tumours contain a heterogeneous tumour cell population and that therefore cancer cannot be considered as a simple clonal expansion of transformed cells. This raised the question of in which terms cancer can be considered a dis–organization in a complex three–dimensional structure in which cells become functionally heterogeneous as a consequence of differentiation. A systemic model and a hierarchical account came to be perceived as necessary to account for cancer, as no particular and specific molecular parts seemed to underlay it. Plasticity and epigenetics have been included in the model to explain cancer development, as they imply temporal factors and appear to be able to account for the heterogeneity of phenotypes among cancer cells and tumours. However, to make sense of the phenotypic heterogeneity resulting from differentiation, the temporal and phenotypic hierarchy among tumour cells implies, almost obligatorily, the idea that tumour initiation is related to tumour progenitor cells (Feinberg et al. ). The hypothesis of the model proposed by Feinberg is that cancer arises from Cancer Stem Cells (CSC), that is cells that do not undergo normal differentiation due to aberrant epigenetic marks such as methylation, hypo–acetylation of histones (a modification of these proteins that interferes with chromatin structure and thus DNA . At this stage, we make use of the “systemic” adjective only to refer to those explanatory models that are presented through systemic graphics and networks. Systemic perspectives, which support an explicitly organicist view of the disease, are also present in literature. Such is the case of Sonnenschein and Soto (Sonnenschein and Soto ) who, moving from a holistic perspective of the disease, present cancer as “the result of the disruption in the tissue’s architecture” or a “systems biology disease” (Hornberg ) rather than a genetic or cellular one. I will refer to them by the end of the chapter.
. Are complex natural systems necessarily decomposable hierarchies?
expression), hypo–methylation of specific genes involved in regulation of promoter regions and tumour silencing bound to oncosuppressor genes (Feinberg and Tycko, ). The fact that epigenetic changes are found so early in carcinogenesis, and even in normal tissues before tumours arise, indicates that early epigenetic changes in stem cells might provide a unifying view of cancer aetiology: it is in fact at this level that the process of differentiation is mediated through epigenetic regulatory mechanisms. Such alterations are inherently polyclonal and can mimic, in some cases, the effect of genetic mutations as well. The existence of CSC has then led researchers to focus their efforts on the identification of epigenetically compromised progenitor stem cells. Supported by such an epigenetic framework, “[t]he Hierarchical Model of cancer implies that only a small subpopulation of tumour stem cells can proliferate extensively and sustain the growth and progression of a neoplastic clone” (Vescovi et al. , ). According to this model, cancer originates from cancer stem cells that perform characteristic features of their normal counterpart, such as the processes of auto–renewal and differentiation [Lobo et al. ]. Thus cancer would be linked first to abnormal regulation and hierarchical organization of cell differentiation. .. Near–decomposability as a default state for cancer An explanatory account of complex systems in terms of hierarchies goes back to Simon’s article on “The Architecture of Complexity” (). In its simplest definition a hierarchic system is “a system that is composed of interrelated subsystems, each of the latter being, in turn, hierarchic in structure until we reach some lowest level of elementary subsystem” (Simon , ) . This definition seems to meet the organizational properties attributed to cancer cells in the Hierarchical Model of cancer. In that framework, tissue cancer is considered to be a . Things we consider to be elementary particles aren’t usually elementary particles. Simon already makes this remark. The quote follows like this: “In most systems in nature, it is somewhat arbitrary as to where we leave off the partitioning, and what subsystems we take as elementary” (Simon ). Equally, a cell is not elementary but built of all kinds of other crap. However, we assume it to be elementary enough for the purpose of the present discussion.
How science works
phenomenon with a modular compositional structure, for which near decomposability is revealed by the almost autonomous properties of the parts, the cancer cells. Nearly decomposable systems are ones where: a) in the short term the components of a subsystem are independent of components of other subsystems; b) in the long term each subsystem depends, in an aggregate way, on the other subsystems. The dependency or independency of components is specified in terms of behaviour, i.e. of some dynamic properties (Simon , ). The room’s temperature example, used by Simon, fits quite well with the hierarchical structure cancer seem to derive from. Although each part of a room starts at a different temperature, the room reaches temperature equilibrium quickly; the rooms in the building will also all eventually reach equilibrium, but in a much longer time than individual rooms. Analogies can be run with some cancer cells properties, e.g. genomic instability. This feature, which is supported by sequence analysis of tumour cell genomes (Yachida et al. ), is an on–going process during tumour progression, as it is commonly recognized as a hallmark of neoplastic cells, and as a cause of (or one of the causes of ) the striking intra–tumoral genetic heterogeneity often described. This genetic diversity may “enable functional specialization, producing subpopulations of cancer cells that contribute to distinct, complementary capabilities, which then accrue to the common benefit of overall tumor growth” (Hanahan and Weinberg , ). Hierarchic systems are, in this view, nearly decomposable: only aggregative properties of their parts enter into the description of the interactions of those parts. Let us just assume that the hierarchical system we are looking at is in its simplest formulation the one described by Simon as near decomposable and defined by Bechtel and Richardson () as a Component System (CS) where the behaviour of the parts is intrinsically determined. Therefore, “[t]he organization of the whole is critical for the functioning of the system as a whole, but provides only secondary constraints on the functioning of the constituent” (Bechtel and Richardson , ). Coherently, cancer can thus be portrayed as “a cell–autonomous process” intrinsic to the cancer cell that takes place mainly at the genetic level (Hanahan and Weinberg ) and tumour invasion is controlled by coordinate processes that enable cancer cells to dissociate and migrate from the primary tumour to
. Are complex natural systems necessarily decomposable hierarchies?
the microenvironment of a distant tissue (Chaffer and Weinberg ). Metastasis of carcinoma therefore implies tumour cells survival in blood circulation, proliferation at secondary sites and the possibility for these cells to avoid immune clearance (Takanori et al. ). The hierarchical feature that characterize tumour cells organization and behaviour follows, in these models, a strict bottom–up determinism. The way of discussion of some biological features of the tumour cells reflect it: “the cell type from which a particular cancer arises plays a significant role in determining the likelihood that a given tumor will eventually metastasize, independently of genetic variation among individuals” (Gupta , ). It is generally the case, in fact, that the same set of genetic lesions that are present in the genomes of primary tumour cells are found as well in the genomes of their derived metastases ( Jones et al. ). Within the cellular and genetic deterministic framework these evidences support the belief, even if the driving force for metastasis are not specific genetic alterations that were acquired late during the multistep formation of a primary tumour, that there are intrinsic properties to the cancer (stem) cell that account for the final onset of cancer. This approach, from an epistemological point of view, mirrors the classic ontological reductionism whose notion of hierarchies takes into account just the molecular level. Inter–level dynamics here are explained by means of lower level properties and parts. Their point is that tumour cells do perform their metastatic features in their own right. That is why a conclusion like this has been driven: “Together, these diverse lines of evidence suggest, but hardly rigorously prove, that the dissemination of cancer cells from a primary tumor occurs as an almost inadvertent side–effect of primary tumor formation rather than a trait that is actively selected during this multistep process” (Weinberg , ). A decomposable system, in fact, is modular in character, with each component operating primarily according to its own intrinsically determined principles. Parts can be studied and investigated separately. This implies that the components have discrete intrinsic functions intelligible in isolation, even if such functions do not independently replicate those of the system as a whole. Following the Hierarchical Model, cancer cells are sub–systems initially composed in a tissue and subsequently able to behave in a peculiar, homogeneous way dependent upon their intrinsic properties. A mod-
How science works
ule in fact is mainly conceived as a “component that operates in an integrated and relatively autonomous manner in the production of the properties of the system of which it is a part” (Mitchell , ). The question thus becomes whether these assumptions are realistic. The answer, in fact, can influence the explanatory role of this hierarchical approach for carcinogenesis. However, as noted, only a posteriori can it be stated that the assumption is realistic, by seeing how closely we can approximate system behaviour by assuming it. Moreover, “the failure of decomposition is often more enlightening than its success: it leads to the discovery of additional important influences on behaviour” (Bechtel and Richardson , ): although we may be led to erroneous explanations, it should be the only way to begin. There are two different points to be addressed here: adequacy of the model and necessity to pass through it to come to a fully explanatory model. .. Formal requirements that are not accomplished in the model In a nearly decomposable system, the short–run behaviour of each of the component subsystems is nearly independent of the short–run behaviour of the other components. In the long run, the behaviour of any one of the components depends in only an aggregate way on the behaviour of the other components (cfr. Section .). A nearly decomposable system is a CS to the extent that the causal interactions within (or intra) subsystems are more important in determining component properties than are causal interactions between (or inter) subsystems (cfr. Simon , in Bechtel and Richardson , ). In the Hierarchical Model of the neoplastic process, CSC cancer (stem) cells might just lose inter–interactions, showing their intrinsic property to proliferate and differentiate. This would be enough to explain how cancer cells present the capability to resemble a hierarchical organization, although aberrant, once those weaker constraints, i.e. inter subsystems, have been lost. Some problems, however, arise when trying to fully match these requirements with some other empirical evidence about metastasis. A perspective that looks at cancer as a cell–autonomous phenomenon (Hanahan and Weinberg ), in fact, seems to review its position when analysing some features of the metastatic process. In a recent
. Are complex natural systems necessarily decomposable hierarchies?
review Hanahan and Weinberg explain that, considering the dependence of the metastatic process on the micro–environment, we have to recognize that “the phenotypes of high–grade malignancy do not arise in a strictly cell–autonomous manner, and that their manifestation cannot be understood solely through analyses of tumour cell genomes” (Hanahan and Weinberg ). Moreover, colonization itself is unlikely to depend exclusively on cell–autonomous processes (ibidem). Contextual factors play an equally crucial role. Moreover, tumours tend toward greater dedifferentiation as they become more aggressive, to a point where it is difficult or impossible to determine their tissue of origin based solely on histological examination (Milovic et al. ; Dowell ; Pavlidis et al. ). If metastases are just cancer cells that have moved from their original place, having lost aggregative constraints, they should maintain modular properties, mainly stability. Instead they display a less stable phenotypic state, whose output strictly depends on the microenvironment. So that another modular feature fails, i.e. context–insensitivity. Further evidences make this point even more clear. They are mainly related with the possibility of spontaneous regression of tumours: cell and tissue differentiation and apoptosis are central to the regression process, which ends up showing mature and differentiated cells (Brodeur ). Moreover, when early embryos are transplanted into ectopic places, they acquire properties of malignant neoplasms called teratocarcinomas. When teratocarcinoma cells were injected into a blastocyst, they generated normal tissues and organs. Thus, embryonic cells produced neoplasms when misplaced in adult tissues and reverted to normalcy when placed into an early embryo (Mintz and Ilmensee ). Thus, it does not seem plausible that it is “the differentiation program of the normal cell of origin to exert strong influence on whether or not metastasis will eventually occur” (Weinberg , ). This differentiation program hypothesis suggests that once a mutation causing specific features of malignancy takes place (i.e. invasiveness, differentiation, metastatic property), such features are necessarily transmitted to the progeny and are thus present in all the cells of the neoplasia. Intratumoral heterogeneity is due to subsystems of cells that are genetically identical although phenotypically distinct: “Cancer cells that are less and more differentiated can share a common set of genetic alterations” (Chaffer and Weinberg , ). The differentiation pro-
How science works
gram of the normal cell of origin represented a strong determinant of eventual metastatic spread. Hence, the nature of the normal cell of origin, which serves as the progenitor of all the neoplastic cells within a tumour, sets the stage for whether its descendants, years or decades later, will or will not show metastatic tendencies. However, this view of cancer considers the aberrant differentiation not the explanans of the final neoplastic phenotype but an explananda. In other words, it explains how we do see a differentiated phenotype among cells produced by a tumour cell, and so why differentiation has been an argument for the Hierarchical Model, but not how it could be considered an argument when explaining the whole phenomenon, especially if you keep considering it as an intrinsic property of tumour cells. What does it mean, then, that malignant cancer cells retain some sort of memory of the cell type from which they originated (Gupta ) during later stages of tumour progression? Evidence of the origin of cancer cells from an original cancer stem cell seem to call for another kind of explanatory tool that is no further related with the functional properties that identify the parts (tumour cells), but with some other organizational features that characterize the system they belonged to. In other words, integrating explanatory models of cancer by means of the concept of differentiation shifts the real claim — and it is not a trivial one — to a background condition related to the real hierarchical organization of biological systems: differentiation depends upon the organismic context and differentiated cells maintain a sort of memory of their historical and genealogical derivation from the original cellular parts that constituted the organisms in its first developmental steps. Epigenetics explain how this is possible and how evolving hierarchies get and maintain stability by changing. The scheme of self–renewing progenitor stem cells and fully differentiated end–stage cells recapitulated in many carcinomas and other tumour types (Ailles and Weissman ) might not work as an explanatory model of the neoplastic process itself, although it is an effective tool to present and explain some features of the neoplastic phenotype . Coherently, the epithelial mesenchymal transition (EMT) — i.e. . Let us clarify here that authors, like Feinberg or Baylin, who have an argument in the CSC theory to support the polyconal origin and some features of cancer heterogeneity, would not derive such strong conclusions from their evidence and presuppositions.
. Are complex natural systems necessarily decomposable hierarchies?
tumour cells’ capability to invade other tissues — turns out to be again a necessary but not sufficient condition for metastasis, while a model of cell cooperation has been proposed to this aim (Takanori ), in which different kinds of cells contribute to the migration of metastatic cells. Matrix degradation, local invasion of tissues and blood or lymphatic vessels and other processes required to complete the metastatic process do not seem to belong to the same cell. Evidence that many invasive and metastatic carcinomas have not undergone a complete transition to a mesenchymal phenotype or even lack signs of EMT, and that invasive carcinomas do not invade adjacent connective tissue as individual mesenchymal–like cells, suggests that these cells invade cooperatively, as multicellular aggregates (Friedl ). Accordingly, the cell cooperation theory has been experimentally proven to play an important role in cancer metastasis (Lyons ). The changes in cell adhesion and migration during tumour invasion are thus to be considered reminiscent of a developmental process termed epithelial–mesenchymal–transition (Thiery and Sleeman ), another proof of differentiation through derivation but not that differentiation, nor EMT, derives from intrinsic cellular properties. .. Why is the notion of differentiation crucial in solving the apparent paradoxes As differentiation implies the genealogical origin of the cells, the hierarchical control in the organism is based on differentiation and not vice versa. This also implies that the Hierarchical Model of cancer works within its own constraints, those of looking at cancer as a cell centred phenomenon, and thus can be useful to understand some features of the neoplastic process when it can be viewed and described in that way, but it cannot provide a full explanation of the whole process itself. Interestingly, some aspects of cancer are aggregative, but even more interesting is that other features are not. The pathological features of tumour cells and metastasis in this model show characteristics of nearly decomposability, while the features that do no fit into that model are better understood from the perspective of the real hierarchical system such cells belong to and its peculiar property and that resist such nearly decomposability. Epistemological problems arise when we consider
How science works
dynamic processes that intrinsically constitute the system as explanatory of constructive and disruptive organizations. The logical limits of the bottom–up hierarchical model and of a nearly decomposable system can be overcome just restricting its explanatory role to a specific and concrete feature of the phenomenon or widening the perspective on the latter by means of different systemic assumptions. In fact, cancer cells do work in an aggregative way to some extent: they depend upon inputs of other components; they influence other components by their outputs; they have specific, intrinsic, properties like proliferation. Nevertheless, they perform so by loosing the capability to integrate contextual signals in a specific way, i.e. characteristic of the biological level of organization (an intermediate stable state). However, they maintain some memory of the properties they acquired during the process of physiologic differentiation. Cells’ origin and differentiation process leave a hallmark that, under certain contextual conditions of proliferation and differentiation, will resurface although in an aberrant way. Metastases, in fact, are able to reconstruct in some cases features that are typical of the organ of origin. A real systemic perspective implies a dependence of identity of biological entities on context (Duprè ). What picture comes up from this puzzling evidence? Cancer itself seem to be a nearly–decomposable system, where the modular feature is reduced to the nature of its proliferative process that is in fact shared, at least in principle, by all the cancer cells. The organismic level of analysis of the neoplastic process, instead, shows features that are much more similar to an integrated system. “This is [a type] of system in which the component subsystems have evolved together, and are not even obviously separable, in which case it may be conceptually difficult to decide what are really relevant component subsystems” (Levins , , cit. in Bechtel and Richardson , ). Systemic organization is significantly involved in determining the functions of its constituents and provides primary constraints on their functioning, which is far from being intrinsically determined (Bechtel and Richardson ) . . As Silvia Caianiello pointed out, this follows a concept of modularity and of the hierarchical account of Simon that can be discussed and reframed in a different way taking into account the recent discussion of modularity in Evo–Devo and Systems Biology (cfr. Gass and Hall ). Alfredo Marcos also suggested that some ambiguity that arises from the nearly–decomposable account of Simon is due to the fact that the nearly decomposability
. Are complex natural systems necessarily decomposable hierarchies?
Whom does, then, differentiation depend upon? The following arguments will focus on this dimension of the property of differentiation, without however discussing in detail the kinds of differentiation implied, which, however relevant, is beyond the aim of this chapter. The Hierarchical Model itself has its roots in experiments that demonstrated that fusion of cancer cells with normal cells lead to a reversion from the neoplastic phenotype, and prompted the definition of oncogenes (Harris ). These results suggested that the mechanisms of tumour transformation represent a much more complex actual processing than the genetic dominance doctrine would ever predict. Oncogenes predispose to cancer, but other events that may occur stochastically are required to bring about the neoplastic transformation. A different view, implying a new epistemological framework for explaining cancer progression, considered tumours as “a disease of cell differentiation rather than multiplication” (Harris ). This new theory of cancer shifted the focus about the onset of cancer from the cellular proliferative process towards the differentiation process. (Mechler et al. , Mechler et al.). Although genes remained the key characters and their mutations the crucial element in carcinogenesis, other mechanisms — besides cellular proliferation — played also a role in the onset and progression of cancer. This is why, also from an historical point of view, we end up with different answers to the Simon question depending from what perspective we look at carcinogenesis. According to this latter theory, named Cancer Cell Theory, an impediment in the key steps of the normal cellular differentiation is responsible for the onset of cancer. From this point of view cellular differentiation determines tissue specificity and, in doing so, can suppress or enhance cell multiplication, but only under precise conditions and inducing — in extreme conditions — cell death (Harris ). These conditions are precisely the ones assumed by the Hierarchical Model, where epistemic constraints come finally under the spotlight. In fact, it doesn’t seem reasonable that evolution selects genes dediconcept suggests that hierarchical systems are decomposable while also denying this. In this paper, following Bechtel and Richardson’s discussion I have emphasized the first aspect (decomposability).
How science works
cated specifically to the formation of tumours. On the contrary, it is reasonable to think that the same genes that suppress the proliferation of normal cells during their differentiation suppress tumour cell multiplication. Thus, the process of differentiation is crucial in maintaining the normal phenotype and, if compromised, in causing the neoplastic phenotype. Researchers took a major step toward establishing tumour cell hierarchy as a fundamental concept in cancer biology by defining the dynamic element that characterizes carcinogenesis and the neoplastic process in terms of cell differentiation. Tumour heterogeneity stopped being a clutter but was instead recognized as a characteristic of the neoplastic mass. To cope with evidence related with tumour heterogeneity and processes involved in growth and development, other approaches and experimental evidence have connected cancer to a different epistemological perspective. The view has been changing from considering tumours as resulting from dynamic changes in the genome, to a perspective that considers cancer as a problem of tissue organization: a problem of a specific level of stability of the hierarchical system rather than of its parts. Cancer would thus not be just a simple clonal expansion of transformed cells, but presents a three–dimensional complexity in which cells become functionally heterogeneous as a result of a process similar to development or to an aberrant organogenesis. This perspective is also consistent with the evidence that some cells within a tumour exhibit phenotypes that apparently correspond to different stages of development, thus prompting the suggestion that, even though originally arising from a single clone, most tumours eventually contain a heterogeneous population or partially differentiated cells, exactly reflecting the composition of normal organs (Pierce ). As already mentioned, the normal diagnostic procedures seem to support this thesis, which involves looking at cancer as a disorganization of the hierarchical structure of the normal tissue and structure of an organ. A more comprehensive view of cancer can rightly look at it as a development that is gone awry (Soto et al. ). Adopting emergentism as default, multicellular organisms are viewed as entities that simultaneously coordinate and control the proliferation of different cell types, ontogenetically and historically linked to the whole of which they are part (Soto and Sonnenschein ).
. Are complex natural systems necessarily decomposable hierarchies?
Tumour would then be the result of a chain of miscommunications, with slow and almost imperceptible stochastic changes within a continuum of feedback that produces further changes. Eventually, control of the signals is weakened, so that the cells express only the phenotype appropriate to their positional context. This permits the cells to express, in their three dimensional environment, their innate capacity for proliferation and deregulated differentiation, and to present, once the regulating signalling has been disrupted, a heterogeneous phenotype. About years ago, Potter (Potter , ) somehow anticipated this idea. He envisioned the neoplastic growth as a problem in intercellular communication and differentiation, and championed the concept that carcinogenesis is a “blocked ontogeny” (Potter ). The basic idea was that cancer cells have lost a feedback control mechanism in proliferation, so that their ability to divide becomes unrestricted. In addition, cancer cells acquire a variety of new properties that render them destructive to the organism as a whole. The idea that cancer might be a problem of cellular communication, has also been stated by Biava (). Studying the relationship between carcinogens, mutagens and teratogenesis, his attention focused on data that showed how cancer–causing agents, when administered during pregnancy, had different effects depending on the period in which they were given. He highlighted again that differentiation is a systemic property and not an intrinsic one of the constituent parts. In this context, it is impossible to isolate gene or cell activity from tissue interactions. In this section we have seen that there are important questions posed by the process of cell differentiation and by the multi–level phenomenology of the hierarchical control that characterizes a biological system. At this point, the question is how can we describe the dynamics that hold the functioning of biological systems that arise through differentiation. .. What does differentiation depend upon? Simon argues that hierarchies arise because the forces governing interactions between objects typically do not form an equally distributed continuum (Simon, ). The strongest forces govern interactions at the lowest level and give rise to reasonably stable units at a middle
How science works
level, so that forces at higher levels determine the relationships between subunits. But, for Simon, it is the equilibrium of forces at the lowest level that determine stable systems. However, as it has been highlighted, “[t]his may or nor may be true” (Bechtel and Richardson , ). Abovementioned data and others reporting examples of reversibility of the neoplastic phenotype during tumour progression, are compelling (Rapp et al. ). Such evidence suggests that contextual factors do play a role in cell division and differentiation, normal and tumoral. Cancer has been therefore also defined as a disease related with a disruption of the hierarchical control. Metastases just apparently are parts that are able to grow/proliferate autonomously, merely overcoming hierarchical and tissue (inter) constraints. One major difference between normal tissues and tumours, in fact, is the malfunction of the differentiation process in tumours, so that undifferentiated and mitotically–active cells eventually accumulate. However, these points are not fully captured by the bottom–up hierarchical account of cancer. An organicist perspective is required to recover the adequate point of view that makes sense of cancer as a disruption of the hierarchical organization of the organism. When explaining the aberrant behaviour of tumour cells, differentiation can be considered as the sum of all the processes by which cells of a developing organism reach their specific function. However, the organization of somatic cells in the tissues is a sine qua non condition of life in multicellular organisms. Such functional condition is reached through differentiation that is, nevertheless, highly controlled by the properties of the tissue once constituted. There is a synchronic and diachronic feedback on functional organization of the cells therefore. This combination of synchronic and diachronic dynamics means that, once established, the upper level is equally responsible for the stable state of intermediate levels of a hierarchical organization (Bertolaso c). The peculiarity of the reference system is not linked here to a cellular system that can give rise to a hierarchical structure characterized by phenotypic heterogeneity, as in the previous interpretative models of cancer, but it is the constituent hierarchical organization of an organism itself that makes sense of the phenotypic heterogeneity of cancer, once the pathological condition for carcinogenesis is settled. Thus there is a reciprocal relationship between the components of the system and their functioning. Looking at cancer in terms of
. Are complex natural systems necessarily decomposable hierarchies?
near–decomposable systems, it seems possible to characterize it in terms of a parameter that is a “measure of the relative magnitudes of intra — and — inter–systemic interactions for these subsystems” Wimsatt (, ). That is why a decomposable approach still has, at some level, a heuristic potential. A bottom–up notion of hierarchy implies that a high level entity is the boss of its lower level parts but that is not really the case: there are no bosses and the relationships are far more complicated. Simon explicitly avoids addressing this feature when talking about formal hierarchies, and from this point of view he is probably wrong in putting biological and physical systems in the same box . Are complex natural systems necessarily decomposable hierarchies? That is, can natural systems be understood in terms of parts whose behaviour depends in only an aggregate way on the behaviour of the other components? From this answer depends, in part, the possibility to explain them, as a whole, just through linear, feed–forward mechanistic explanations. The answer to these questions also give us the possibility to better explore in which sense we should understand that in a nearly decomposable system interactions among the subsystems are not negligible, although weak. Thus the question becomes, weak to what respect? What follows the analysis of the biological features of cancer is that the answer to the main question seems to be: natural systems are not necessarily decomposable. Moreover, when possible, it is just in heuristic terms, because of the failure of the assumption that components arise independently, existing as stable units in their own right, as systemic account of the neoplastic process suggest. Talking about systems and parts, we are incorporating a hierarchy of levels. This implies a horizontal process, i.e. interactions between units at one level, but . I have to acknowledge here that Simon clarifies that when talking about social hierarchies, he does not refer to a control hierarchy, although social hierarchies are non–nested while biological hierarchies are. Substantially, for those who are not familiar with Simon’s paper, he says that formal organizations do not exhaust social hierarchies, although they are prominent in social systems. In such formal organizations “the formal authority relation connects each member of the organization with one immediate superior and with a small number of subordinates” (ibidem, ), so that at a first evaluation social systems can be seen as formal hierarchies, i.e. organizations that “have a clearly visible parts–within–parts structure” (Simon , ). To what extent this is explanatory of social dynamics or not is therefore worthwhile to be further explored.
How science works
also a vertical question, when we are looking at units interacting in larger units. It has been stated that, if the degree of interactions among components when they come to form a whole does not obliterate the components as autonomous entities, the result is a decomposable part–whole hierarchy. But this seems to be the case of cancer and not of a normal hierarchical organized biological system in which the contextual features take over the control of the parts, i.e. of their functional properties. Such relationships — that are strictly related with the notion of differentiation through generation — play an equally important role, or even more important, in maintaining parts’ functional stability as well. Formation of stable combinations that would meet the demands at the higher levels does not seem to be sufficient to account for the real existence of higher stable state (cfr. Bertolaso b). And if these higher stable states are not real, why cancer? .. Conclusions Simon’s paper offers an interesting framework in order to understand different aspects of the hierarchical constitution of biological systems. His account of nearly–decomposability equally forces us to consider apparently opposing features of tumour cells, like their relative autonomy in the process of tumour progression. Although some aspects of the discussion I have started in this chapter will be clarified in the next ones, we can say that a bottom–up hierarchical account of cancer clearly shows to what extent a nearly–decomposability of a hierarchical organized system can be . This seems to overlap with what Simon says, i.e. that stability depends on strong links within subassemblies, but if we accept that this view is explanatory of cancer we are, to some extent, accepting a sort of ontological status of cancer itself. Contradictions thus arise when we have to explain its pathological aspect. We can only assume that such perspective can be used for heuristic purposes and within a wider conceptual framework (I will discuss this point in more detail in Chapter and ). . This is pragmatically assumed by Simon (): “given the properties of the parts and the laws of their interaction, it is not a trivial matter to infer the properties of the whole” (Simon , ) so that genesis of legality and regularities at different levels through organizational principles that are beyond the hierarchical structure is the last issue to be explained. The question about emergence is not completely solved within Simon’s view, but is at the end the real issue for our understanding of cancer.
. Are complex natural systems necessarily decomposable hierarchies?
assumed as fully explanatory of all the behavioural properties of the system’s component. When such accounts work, is through an abstraction of some properties in order to focus on some other that constitute the explananda of the research project (e.g. the persistence of tumour cell properties in the offspring of a CSC in the hierarchical model of cancer). Moreover, the biology of cancer pushes us to move from an analysis of hierarchies in terms of aggregative interactions to a view of biological dynamics in terms of integrative systems. This leads to an understanding of interactions in a more comprehensive way, i.e. able to account for the relative autonomy of the component of the system at different levels of analysis. If in Simon’s account a complex system is made up of a large number of parts that interact in a non–simple way, the challenge is to clarify what that ‘non–simple way’ means. Moreover it is commonly accepted, as in Simon’s account, that the whole is more than the sum of the parts, is a claim that should be firstly assumed in a pragmatic sense, i.e. that given the properties of the parts and the laws of their interaction, it is not a trivial matter to infer the properties of the whole. The question is to clarify how science grasps this non–triviality in its ordinary activity. In Chapter I will clarify how I understand this pragmatic sense through a discussion of how reductions work in biological sciences. What we can conclude at the moment is that in understanding cancer, one may start at the cellular level, or at the level of tissue interactions. However, one will be forced to eventually move up and down levels of organization, since multiple levels are participating in the phenomenon. Following Bechtel and Richardson’s distinction between aggregative and integrative systems, it seems clear that a bottom–up hierarchical structure alone is insufficient to capture the complex features of cancer. So that, from this point of view, in understanding tumour cells behaviour, nearly decomposability is useful as an heuristic tool and offers the advantage to focus on some dynamic features of these parts. On the other hand, the biology of cancer gives us insight to claim that the reason of this insufficiency is mainly related with the assumption that in a nearly decomposable system its components might exist as stable units in their own right, within a biological hierarchical organization. Mainly the context sensitivity of the components’ behaviours will require further clarification and discussion for its epistemological implications as well (see Chapter ).
How science works
References A L.E., W I.L. (), Cancer stem cells in solid tumors. Current Opinion in Biotechnology, :–. A–H M., W M.S., B–H A., M S.J., C M.F. (), Prospective identification of tumorigenic breast cancer cells. Proceeding of the National Academy of Sciencies of the USA, : –. B W., R C. (), Discovering Complexity: Decomposition and Localization as Strategies in Scientific Research. Cambridge, Massachussetts, London: MIT Press. B M. (a), Sulla ‘irriducibilità’ della prospettiva sistemica in biologia. In Urbani Ulivi L. Strutture di mondo. Il pensiero sistemico come specchio di una realtà complessa. Il Mulino, Bologna. ——— (b), Breaking down levels of biological organization, Theoretical Biology Forum (in press),. B P.M. (), Complessità e biologia. Il cancro come patologia della comunicazione. Mondadori, Milano. B B.M., W M.S. (), Cancer stem cells: a step toward the cure. Journal of Clinical Oncology, : –. B D., D J.E. (), Human acute myeloid leukaemia is organized as a hierarchy that originates from a primitive hematopoietic cell. Nature Medicine, : – B G.M. (), Neuroblastoma: biological insights into a clinical enigma. Nature Reviews Cancer, : –. C C.L., W R.A. (), A Perspective on Cancer Cell Metastasis. Science, , . D P., C R.W., C M.F. (), Cancer stem cells: models and concepts. Annual Review of Medicine, : –. D J. (), It is not possible to reduce biological explanations to explanations in chemistry and / or physics. In Contemporary debate in philosophy of biology. Edited by J. Ayala and R. Arp. Wiley–Blackwell, Oxford. D J.E. (), Cancer from an unknown primary site. The American Journal of the Medical Sciences, : . F A.P., O R., H S. (), The Epigenetic Progenitor Origin of Human Cancer. Nature Reviews Genetics, : –.
. Are complex natural systems necessarily decomposable hierarchies?
F A.P., T B. (), The history of cancer epigenetics. Nature Reviews Cancer, : –. F P., W K. (), Tumour–cell invasion and migration: diversity and escape mechanisms. Nature Reviews Cancer, : –. G G., H B.K. (), Collectivity in Context: Modularity, Cell Sociology, and the Neural Crest. Biological Theory, (),:–),. G P.B., M S., Y J., H K., W R.A. (), The evolving portrait of cancer metastasis. Cold Spring Harbor Symposia on Quantatitative Biology, : –. H D., W R.A. (), The Hallmarks of cancer. Cell, : –. ——— (), Hallmarks of cancer: the next generation. Cell, :–. H H. (), Tumour suppression: putting on the brakes. Nature, : . ———, M O.J., K G., W P., T T. (), Suppression of malignancy by cell fusion. Nature, : –. H J.J., B F.J., W H.W., L J. (), Cancer: a systems biology disease. BioSystems, : –. J S., C W.D., P G., D F., B N., A T., T A., N M.A., S C., V V.E., K K.W., V B., W J., M S.D. (), Comparative lesion sequencing provides insights into tumor evolution. Proceeding of the National Academy of Sciencies of the USA, : –. K K.W., V B. (), Lessons from hereditary colorectal cancer. Cell, : –. ——— (), Cancer–susceptibility genes. Gatekeepers and caretakers. Nature, : –. L R. (), Complex Systems. In Towards a Theoretical Biology, Waddington CH, ed, –. Edinburgh: University Press. L A.C., O F., S S., B C.F., MM M., L H. (), Coopera– ting oncogenes converge to regulate cydin/cdk complexes. Genes Day, : –. L N.A., S Y., Q D., C M.F. (), The biology of cancer stem cells. Annual Review of Cell Developmental Biology, : –. L J.G., L E., M A.M., M M.R. (), Clonal
How science works
diversity in carcinomas: its implications for tumour progression and the contribution made to it by epithelial–mesenchymal transitions. Clinical and Experimental Metastasis, :–. M B.M., M. G W., G W.J. (), Molecular cloning of lethal(),giant larvae: A recessive oncogene of Drosophila melanogaster. EMBO Journal, :–. M B.M., T I., S M., O M., K A., M R., P U. (), Molecular basis for the regulation of cell fate by the lethal(),giant larvae tumour suppressor gene of Drosophila melanogaster. Ciba Foundation Symposium, : –. M M., P I., J S. (), Tumor markers in metastatic disease from cancer of unknown primary origin. Medical Science Monitor, : MT. M B., I K. (), Normal genetically mosaic mice produced from malignant teratocarcinoma cells. Proceeding of the National Academy of Sciencies of the USA, : –. M S.D. (), Modularity—More than a buzzword? Biological Theory, : –. P N., B E., H J., G F.A. (), Diagnostic and therapeutic management of cancer of an unknown primary. European Journal of Cancer, : . P G.B., N P.K., M–H A., W J.M. (), Ultrastructural comparison of differentiation of stem cells of murine adenocarcinomas of colon and breast with their normal counterparts. Journal of the National Cancer Institute, : –. P V.R. (), Biochemical perspective in cancer research. Cancer Research, : –. ——— (), Phenotypic diversity in experimental hepatomas: the concept of partially blocked ontogeny. The th Walter Hubert Lecture, British Journal of Cancer, : –. R U.R., C F., S R. (), Oncogene–induced plasticity and cancer stem cells. Cell Cycle, : –. S H.A. (), The Architecture of Complexity. Proceedings of the American Philosophical Society, :–. S C., S A.M. (), The Society of Cells: Cancer and Control of Cell Pro– liferation. Springer–Verlag Inc, New York
. Are complex natural systems necessarily decomposable hierarchies?
S A.M., M M.V., S C. (), Neoplasia as development gone awry: the role of endocrine disruptors. International Journal of Andrology, : – . S A.M., S C. (), Emergentism by default: A view from the bench, Synthese, : –. ——— (), The tissue organization field theory of cancer: A testable replacement for the somatic mutation theory. Bioessays, :–. T T, S I, G– H (), Epithelial–Mesenchymal Transition and Cell Cooperativity in Metastasis. Cancer Research, : –. T D. (, ), New insights into the pathogenesis of breast cancer metastasis. Breast Disease, : –. T J.P., S J.P. (), Complex networks orchestrate epithelial–mesenchymal transitions, Nature Reviews Molecular Cell Biology, :–. V A.L., G R., R B.A. (), Brain tumour stem cells. Nature Reviews Cancer, :–. W R.A. (), Mechanisms of malignant progression. Carcinogenesis, : –. W W.C. (), Complexity and organization. In K Schaffner and RS Cohen, eds. PSA : Proceedings of the Biennial Meeting of the Philosophy of Science Association, : –. Dordrecht: D. Reidel. Y S., J S., B I., A T., L R., F B., K M., H R.H., E J.R., N M.A., V V.E., K K.W., V B., I–D C.A. (),. Distant metastasis occurs late during the genetic evolution of pancreatic cancer. Nature, : –.
Chapter IV
Why reductionism is not impossible, just (sometimes) impassible Alice says: “I simply must get through!” The Doorknob replies: “Sorry, you’re much too big. Simply impassible.” Alice insists: “You mean impossible?” but the Doorknob concludes: “No, impassible. Nothing’s impossible.” L Cl, Alice’s Adventures in Wonderland,
.. Introduction The inter–level regulatory features of higher–level properties– often addressed in terms of emergent properties– pose major difficulties for reductions in biological sciences (Kim , Boogerd et al. , Schaffner b). However, in this chapter, instead of challenging the idea that emergent properties contrast the possibility of reduction, and trying to show whether a reductionist–mechanistic account explains them, I will focus on the conditions of validity for any reduction that addresses higher–level biological dynamics. I will start by examining some limits of reductionist mechanistic explanations of biological behaviours that imply inter–level regulatory features and then discuss some epistemological implications of their emergent features. Cancer research and cancer biology offer us adequate case studies. We find in literature that cancer is generated at different levels of the biological organization of a multi–cell organism. Physical forces or chemical factors that alter the structure of tissue are equally relevant in carcinogenesis (Ingber ) and tissue dynamics are crucial for the final onset of cancer, mainly for its metastatic properties (Nelson and Bissel ). Various papers discuss how the complexity of cancer is actually related with its dynamics, which compromise the functional stability of the hierarchical organization of the body at different scales (e.g. Rubin ,
How science works
Heng et al. , Bizzarri et al. ). Explanatory models of cancer deal with inter–level regulatory features of a complex biological system. These models are typically inter–level models that explain how emergent properties are lost and work through partial decomposition of the phenomenon, which is eventually described in molecular terms. In the first part of the chapter, I take advantage of the analysis already performed by Schaffner on the possibility of working on ‘creeping reductionism’ to explain complex biological behaviours (Schaffner a). I then consider what mechanistic explanations try to explain, and what they (apparently) leave out. I suggest that requirements for reductions imply a conceptual shift in the definition of parts that integrate explanatory accounts. Schaffner’s notion of PCMS (Preferred Causal Model Systems) seems to rely on these requirements. My contribution is to make them explicit and show how they entail a non–reductionist relationship among the relata of the explanatory argument. In the second part of the chapter, I will give some examples of how this conceptual shift occurs in practice, and in what sense a reductionist mechanistic approach is unable to capture the explanatory relevance of higher–level features in cancer biology. Examples come from scientific literature, supported by an extensive literature on the limits of mere mechanistic account of cancer’s complex dynamics (Bizzarri et al. , Bertolaso a, Moss , Mukherjee ). In the third part, I discuss in more detail the process of reduction in biological sciences and the conditions of validity, or requirements of their explanatory accounts. Beyond the detailed analysis by Schaffner of the formal structure of models, it will be argued that the reductionist presumption to explain thoroughly biological systems by appealing only to the properties of their structure cannot be sustained and that emergence shall not be considered ‘innocuous’ even in epistemological terms. My discussion aims to clarify what features of scientific explanations are at stake when a reduction is performed in biological sciences. Discussing the relation between reductionism and emergence in terms of ‘impassibility’ instead of ‘impossibility’ helps distinguish methodological issues from the epistemological explanatory enterprise of . The background of the discussion about the relata of an explanation can be found in Silberstein .
. Why reductionism is not impossible
reduction in biological sciences. To be impassible marks how we can and cannot in general explain nature according to what the world allows rather than what logic requires . .. Partial reductions Philosophical discussion about reductionism in biology is historically related to the possibility of theory reduction. The backbone of this tradition traces back to the work done by Nagel and revised by Schaffner. He argues that the thesis that reduction of one theory or one branch of science by another, considered more fundamental, is possible, is no longer applicable when trying to capture what is going on in biological scientific practice, though it may make sense in physics (Schaffner , b). Schaffner integrated Nagel’s original account by introducing a different analysis of biological theory as a collection of overlapping causal and inter–level models. In , he distinguished the Nagel type of Generalized Reduction–Replacement (GRR) model from a causal/mechanical approach (CM) that works through partial reductions, i.e. reductions that (paradoxically, in his view) are typically multi–level in both the reduced and the reducing sciences. In partial reductions, the explanandum is a macro property or an end state, and the explanans is identified with parts of the organism or a process of interest. Assumptions that permit relations of the macro and micro descriptions are thus specified and called connectability assumptions (CAs), or bridge laws and reduction functions. CAs can be causal sequences or identities, related with some regulatory aspect of the system of interest. These reductions — defined by Schaffner ‘creeping reductions’ — turn out to be partial because scientific explanations always imply or include higher–level features even accounting for them in terms of . Some discussions of the limits of mere logical accounts to explain and understand the world and its laws are also present in Mitchell , . . This denomination creates a contrast with the original account of reductionism (sweeping reductionism), which implies an ontological claim and intra–theoretic reductions: “I call ‘sweeping reductionism’ where we have a sort of ‘Theory of Everything’ and there is nothing but those basic elements — for example, a very powerful biological theory that explains all of psychology and psychiatry. The second kind is ‘creeping reductionism’ where bit by bit we get fragmentary explanations using inter–level mechanisms” (Schaffner a).
How science works
molecular features and mechanisms. Creeping, partial, reductions are those that mix higher entities and predicates with relatively lower–level entities and predicates. An explanation is truthfully reductive in so far as it appeals to entities that are parts, but it is also non reductive because: ) it does not explain all the cases of the scientific problem under consideration; ) it is plausible that key entities have not yet been discovered to account for the higher level behaviour of interest; ) it refers to middle–level entities (mainly collective systems, my emphasis); ) it is a causally qualitative model (not quantitative and only roughly comparative); ) the set up of the models takes place at a higher (aggregative) level than molecular (my emphasis). Given these features of partial reductions, two elements formally characterize such causal mechanical explanation models in Schaffener’s account: Field Elements (FE), that refer to plausible explanatory candidates (generalizations, mechanisms, kind of experiments, etc.), and Preferred (Causal) Model Systems (PCMS) that have to be understood as “causal system[s] representing a temporal process” (Schaffner , ; see also Schaffner ). PCMS implicitly or explicitly involve laws and generalizations that are relevant for the scientific problem’s explanation. Such generalizations are causal and qualitative, describing parts of mechanisms in a process. The issue at stake in the enterprise of a reduction is to assess whether it is possible to explain higher–level properties in terms of interactions of parts located at a different, lower, level of the biological organization. Asserting that a creeping reduction is not committed to a ‘nothing but’ account of the (biological) world while relating explanations to inter–level mechanisms, Schaffner leaves open the debate in which sense and to what extent mechanistic explanations really fit with the inter–level regulation of biological systems. Similarly, acknowledging emergent properties in epistemological and compositional terms . In Schaffner’s account of partial reduction, emergence should be understood as a “failure of any possible explanation of a whole in terms of its parts and their relations
. Why reductionism is not impossible
does not resolve the question to what extent reductionist–mechanistic explanations can explain higher–level properties without any residue, i.e. without need of further explanatory clarification. This concern emerges in Schaffner’s papers when claiming that the power and limits of reductive explanations at some point have to face the discussion about the epistemological role of generalization and laws in biological theories (Schaffner , ; see also Schaffner , Ch. ). If generalization is relevant in explanatory accounts, merely mechanistic explanations might fail to capture relevant features of biological phenomena that are necessary to develop a fully explanatory account of the object of inquiry in the terms sketched by CM explanations. I agree with this remark, and the analysis of some issues in cancer biology, along with the epistemological role of PCMS, will provide further material to its discussion. However, as we will see, the outcome of this analysis will compel us to consider a ‘non innocuous emergence’, in which not only the parts and their interactions still have to be specified to account for the higher–level properties, but also the definition of parts and interactions depends on the properties of the system to be investigated. Another way to put it is that the need to bring emergence into the picture is not related to a general claim about the ‘in principle’ impossibility of reductionism in science but rather to the need to investigate the specific requirements for any substancial reduction when regulatory features of biological behaviour have to be explained. These requirements — or condition of validity of explanatory models — challenge the mereological explanations, typical of mechanistic accounts, as they overlook the explanatory process of conceptual generalization that characterize the definition of the parts and upon which their explanatory accounts rely. This kind of reductionism is therefore ‘impassible’ to the explanatory challenge that inter–level regulatory features pose and does not overcome the limits of theory reduction. (and expressed only in the parts’ language)” (Schaffner , ). The assumption is that the thesis that parts do not tell you what the whole will do without a specification of the interrelations among parts themselves is commonly accepted. This is what Schaffner calls “innocuous emergence” (ibidem). This allows adopting, as first step for an epistemological analysis of how reductions work in science, a pragmatic approach that acknowledges our current impossibility to derive higher–level behaviours from the knowledge of parts and of their relationships.
How science works
.. What do mechanisms explain? Discussion about mechanisms is centred on what mechanisms are and how mechanisms explain. What they explain is often overlooked. In scientific practice they are, nevertheless, used to account for (as a feature of ) an emergent property, which I am mentioning here in its wider meaning, i.e. a higher–level property that is usually conceived as a behaviour or a stable functional — or end — state of the system of interest. From a scientific point of view, mechanisms do explain robust phenomena. From a philosophical point of view, the aim is to understand how different levels of biological organization are inter–regulated and where the nomological dimension that any regulatory process entails is to be searched. Regulatory and inter–level features of biological dynamics are the real challenge of any explanatory account of biological processes and phenomena. Moreover, statements about the (causal) relevance of higher (emergent) properties in biological systems are still often seen in contradiction with any attempt of reduction in biological sciences, because of the kind of mereological–mechanistic frame adopted by mechanistic accounts . The definition of mechanism proposed in the important paper published in is clearly open to this criticism: mechanism are a collection of “entities and activities organized in the production of regular changes from start or set up conditions to finish or termination conditions” (Machamer, Darden, and Craver , , below: MDC). This account originally offered a general characterization of ‘mechanism’ that attempts to capture how scientists refer to mechanisms and to show the ways in which mechanisms are involved in the explanation of phenomena. Examples taken from different domains in life sciences, from molecular biology to physiology and neurobiology, have been used to support mechanistic accounts. However, to bring the abovementioned dynamic feature of biological systems into the picture, the traditional MDC model has been revised. Richardson and Stephan enrich the definition of mechanistic explanation with Kauffman’s insight (Kauffman ). Mechanisms . This is partially captured by Schaffner when considering that, in strong mechanistic accounts the proper concept of mechanisms is considered as an alternative to rule–based approaches to explanation and reduction (this point is made by Schaffner in , Ch. and in , ).
. Why reductionism is not impossible
can be seen as “an articulation of parts explanation”, so that a mechanism is “an explanation of systemic behavior in terms of the behaviors of the constituent parts within the systemic context” (Richardson & Stephan , ). The reference to Kauffman’s work brings into the picture the regulatory dimension that a mechanistic explanation is meant to capture. For Kauffman a decomposition of the system into parts is conditional on what was seen as the ‘goal’ of the system, what the system is doing. The same attempt, as reviewed by these authors (Richardson and Stephan , ; Bechtel and Richardson ), characterizes Glennan’s account of the mechanism of behaviour as “a complex system that produces that behavior by the interaction of a number of parts, where the interactions between parts can be characterized by direct, invariant, change–relating generalizations” (, S). Nevertheless, the apparent circularity of the former definition in appealing both to the parts’ and wholes behaviour and the meaning of ‘generalizations’ in the latter still need to be clarified. We can overcome the difficulty of resolving the circularity of a definition of the same term (behaviour) used to describe the relata assuming that the behaviour attributed to the parts and the whole are different in kind. Parts’ behaviour is functional to the whole (explanans), while the whole’s behaviour is constitutive, i.e. the higher–level property that defines the object of inquiry (explananda). This perspective is unknown to mechanistic reductionism. Unlike the latter, it aims at a philosophical account of how science works and does not appeal to a ‘nothing but’ or ‘lowest level’ claim as the relevant explanation, but denies in practice that something really new happens in nature: in fact, any occurrence at higher levels can always be explained in mechanistic terms. Instead of a ‘nothing but’ or ‘lowest level’, we are confronted with a ‘nothing new’ issue. As Dupré puts it, the difference between explaining how something does what it does, and explaining what it does through a mere specification of the mechanisms at work, is still an unresolved challenge for reductionist explanation (Duprè ). Therefore, it does not seem that the integration of the traditional MDC account through the emphasis on parts’ and wholes’ behaviour and context is neutral in respect to the explanatory role of mechanisms themselves. The CM account addresses the same problem, i.e. what generalizations are necessary to account for the inter–level regulation of biological systems. The discussion on the partiality of CM models
How science works
and of their causal account can be finally related to the kind of generalizations required for biological processes and their pathological features. At stake is the ‘novelty’ of higher–level properties and, from a methodological point of view, the identification of a level of analysis at which parts, their interactions and their contextual dependence can be integrated into an explanatory model so that higher–level features can be eventually described in terms of mechanisms. Let’s see how this can be applied to a concrete cancer research program and how mechanistic limits and possibilities emerge in scientific literature. .. Cancer stem cells After the discovery of the first genes related to cancer, a linear sequence of genetic mutations was proposed as an explanatory account of the neoplastic features of cancer cells. The features to be explained, commonly known as the ‘hallmarks of cancer’ (Hanahan and Weinberg ), were the following: a) cells become capable of unregulated replication; b) they show independence from growth factors while inducing blood vessel formation; c) cells become able to avoid apoptosis (resistance to apoptosis) and capable of tissue invasion and metastasis . Different molecular factors causally linked to cancer were therefore incorporated into an increasingly branched explanatory model that resulted in what has been called the “integrated circuit of the cell” (Hahn and Weinberg ). An interesting turn occurred, when scientists tried to fit these features (related to the behaviour of cancer cells) with two other features (related to the overall organization of cancer cells) common to all cancers: the heterogeneity of cancer cells within a tumour mass and among different tumours as well. In the following discussion I will focus on the first kind of heterogeneity, i.e. why not all the cells coming from cancer cells present the same neoplastic features: some of them retain the above–mentioned “hallmarks”, while others don’t. . The process of new blood vessel formation (angiogenesis) is secondary in my view, as it does not add significant features to the other already mentioned. Moreover, in scientific literature there are often examples that this capability is not found in all exemplars or types of tumours. For further details, scientific and epistemological, see also Bertolaso b.
. Why reductionism is not impossible
This feature of biological complexity of cancer has prompted researchers to integrate explanatory models of cancer progression. The overall structure of the model has moved to a different level of biological organization rather than the genetic one. To explain how different properties (the hallmarks of cancer) at different times fit into a causal nexus, no new genes but cancer cells became relevant integrating popular explanatory models. On the contrary, when the search was limited to new mechanisms, explanatory models appeared an even more complicated spider–web of parts. From an historical point of view, a glimpse of a more plausible explanation to tumour cells’ heterogeneity developed: epigenetic changes (e.g. non genetic regulatory factors of protein production) appeared to be relevant for the functioning of genes involved in the neoplastic process (Greger et al. ). In contrast to the widely accepted model of cancer as a monoclonal disorder arising from an initiating mutation, a polyclonal model entered the scene with the Epigenetic Model, consistent with the biological features of epigenetic alterations that are, in fact, inherently polyclonal (Feinberg et al. ). Given the role of epigenetics in cell differentiation, this model implied the idea of a temporal and phenotypic hierarchy among tumour cells and tumour initiation and progression that was supported by the “tumor progenitor cells” notion. In this way the above–mentioned difficulties related to the molecular and cellular heterogeneity of cancers were overcome by fixing the initial conditions at a higher level of biological complexity, through the concept of Cancer Stem Cells (CSCs). A typical model successfully developed within this framework is the Epigenetic Model of cancer. According to this model, cancer arises in three steps. An epigenetic alteration of stem/progenitor cells within a given tissue. . A mutation in genes, among those known as relevant in cancer origin and progression . Genetic and epigenetic instability arise, eventually leading to increased tumour evolution. The bulk of the explanatory burden was therefore on progenitor cells: “Note that many of the properties of advanced tumors (invasion, metastasis and drug resistance) are inherent properties of the progenitor cells that give rise to the primary tumor and do not require other mutations (highlighting the importance of epigenetic factors in tumor progression)” (Feinberg et al. , ).
How science works
The relevance that the epigenetic landscape assumed in this model (which works as a PCMS) is also highlighted by a further assumption shared by these authors: alterations in the stem/progenitor cell “can be due to events within the stem cells themselves, the influence of the stromal compartment, or environmental damage or injury” (ibidem). In the process of creeping reduction, the PCMS was adequately chosen for explaining the polyclonal origin of cancers. The relevance of one causal factor with respect to another was also commonly justified in terms of the temporal priority of one element relative to another (Feinberg ) so that the PCMS was able to capture the temporal dimension typical of any biological process. The explanatory scheme that emerges is a serial sequence of mechanically described events (figure .). Instead the other feature that cancer heterogeneity implies, the one of involving different stages of differentiation, has been mainly explained through a model that integrates the epigenetic one, and that is commonly known as the Hierarchical Model of cancer (Vescovi ). This model aims to couple, in a more detailed way, the timing sequence of neoplastic progression with the concept of cell differentiation and tumour heterogeneity. The Hierarchical Model states that only a small subpopulation of tumour cells can proliferate extensively and sustain the growth and progression of a neoplastic clone. The resulting (dis)organization of differentiated units provides therefore a better explanatory system to account for cancer heterogeneity and its temporal dynamics. The PCMS identified through the notion of CSC should also explain the different rates that cancer cells present in producing new cancers. Actual examples of cancers that arise from stem cells prompted interesting research programs that are trying to clarify whether all cancers can be eventually related to some original molecular alterations in stem cells. At the same time, although this seems to be true for a number of cases, it does not seem to cover all of them. However, the working concept of CSC is useful and explains some relevant features of the development of cancer in general. Experimental tests have been run accordingly, mainly based on the possibility to identify CSC through the (rate of the) capability to generate new tumours once transplanted in new organisms (Gupta et al. , Vermeulen et al. ). In these models, it is still possible to fit parts and events into a linear causal sequence, but at the cost of assuming that the crucial
. Why reductionism is not impossible
carcinogenic factor resides in stem cells: the inter–level character of the model involves not only genes, epigenetic factors, and cells, but also the additional condition of stemness of those cells. It is namely . I.e. the capability of a cell to either produce cells that still maintain a multi–potent capability to differentiate or to differentiate into a specialized cell within a tissue.
a)
b)
c) Figure .: In this article Feinberg elaborated the original clonal model (a) into an epigenetic one (b) in which, in any case the level of explanation was changed, as graphically described in (c), making sense of some features that depend on the context and of the causal relationship of epigenetic factors with it. The difference between a) and c) is not just in the complexification; the PCMS is changed through the concept of Progenitor Cell and the explananda focused on the different tumour properties of these Progenitor Cells. (Reprinted by permission from Macmillan Publishers Ltd: «Nature Reviews Genetics», Feinberg A.P., Ohlsson R. and Henikoff S. The epigenetic progenitor origin of human cancer, vol. : –, Copyright ).
How science works
the level of generalization of the biological concept (stemness) that accounts for the specific behaviour of the parts (tumour cells). When this aspect is overlooked, problems arise (in the next chapter a clarification of the epistemological implications of this point is presented). Now it will be easier to understand the relationship between the two points left open in Schaffner’s CM account: a) what determines the causal model system (PCMS) to be ‘preferred’ and b) what are the requirements that reductions (also mechanistic–reductions) rely upon and that justify their characterization in terms of creeping reductions through the PCMS notion. Strong mechanistic accounts by definition involve a regular set of changes, which also implies a continuum of events in which the entities and their relationships are maintained. Intrinsic to the latter condition is also the requirement of (temporal) directionality of causal accounts as generally understood in scientific explanations. However, one of the major features of biological processes is precisely the temporal disengagement of causes and effects or a midstream interruption of the cause while the effect still holds. This is one of the intrinsic features characterizing the idea of inter–level regulation. Problems arise when this intrinsic feature is ignored and explanations are extended beyond the limits posed by the explananda (in the abovementioned cases: the polyclonal nature of tumours and thus tumour cells’ heterogeneity). ... Problems with Cancer Stem Cells The CSC Model, or PCMS in our case, is what accomplishes the partial reduction in Schaffner’s terms. General evidence that cells are involved in the development of cancer, that epigenetic changes are relevant in this process, and that the timing of the process itself is not (only) driven by genetic factors, is satisfactorily captured by explaining these features in terms of progenitor cells and by widening the characterization of these cells through the notion of stem cells. The model is simplified and idealized, and uses causal language such as ‘results in different rates of cancer spreading’. The PCMS also clearly has inter–level characteristics, showing cells, parts of cells (genes), regulatory devices (epigenetic alterations), as well as connections among them (clonal derivation) that ultimately result in the polyclonal feature that characterizes tumour masses.
. Why reductionism is not impossible
However, when the parts (cells) and their interactions are supposed to explain other features that characterize the PCMS as defined in the Hierarchical and CSC model, i.e. the cellular aberrant differentiation, contradictions emerge. The kind of extension that has been attempted is as follows: CSCs “undergo aberrant differentiation” (Dalerba et al. , ) or “tumors are hierarchically organized tissues with CSCs at the top of the hierarchy driving tumor growth and progression” (de Sousa et al. , ). The insights gained with the CSC and with the Hierarchical Model of cancer are lost when the overall structural–functional organization of the organism and its relevance in framing the mechanistic account is overlooked. In other terms, the role of the context is not just a generic epistemological assumption for mechanisms to work, but its specific features –that are different depending on the PCMS — do play a role in the adequacy of those mechanisms as well. Let’s go into some details to show how this works. When treating the biological units in isolation, we face new problems related to the real existence of these units and to their explanatory potential in all solid kinds of tumours (Gupta et al. ). This is because differentiation and multi–potentiality depend biologically on growth factors, tissue structure, and physical forces that do not belong to the cellular level by definition. The biological concept of stemness is heavily determined by the cell’s position in the tissues. Although some features of their organizational behaviour (dynamic properties, i.e. how cell change over time through, for example, epigenetic changes) can be adequately captured in mechanistic terms, other features (like stemness) are relevant for the whole explanatory structure of the model and have a nomological relevance. They define the state of those parts that can explain some features of the biological process; thus the biological level at which the phenomenon needs to be explained – in this case, tumour heterogeneity – is not neutral with respect to the identification of the mechanisms that are explored afterwards or to the identification of the explanans themselves . . Somebody might ask why heterogeneity could not be simply explained by multiple mechanisms. A simple answer might be that mechanisms can eventually explain this kind of heterogeneity but their identification crucially depends on the PCMS. Different examples are available in literature, e.g. stochastic models of cancer, the hierarchical model, a population model, etc. (for a review of some of these models see Vermeulen ).
How science works
Feinberg seems to acknowledge this aspect, but other authors don’t. The priority of the linear, although branched, explanatory sequence of events over the biological (functional) properties of the parts and of them as a whole makes it difficult to go forward through functional explanatory accounts, i.e. attributing new functions to the same parts in order to account for the overall behaviour of the systems they belong to. When CSC are meant to be causally explanatory of both tumour initiation and tumour growth at the same time and in the same sense, the production of differentiated non–tumorigenic offspring by those cells (Vermeulen et al. ) might seem contradictory (and in fact, given the properties of a stem cell, logically speaking, it is) (figure .). The proliferation of CSCs should proceed in either direction at the same rate, making logically untenable the notion of the CSC that is linked, instead, to the notion of oriented division and functional behaviour. A qualitative feature of the explanans (stemness of progenitor cancer cells) is relevant to the adequateness of the mechanic explanation of how not all tumour cells retain neoplastic properties, but it hardly accounts for a next–step–level process like cell differentiation.
Figure .: The stemness character of Progenitor Cells allows a hierarchical structure of cell differentiation. However the tumorigenic character of the PC might imply a bidirectional inter–convertibility between CSAs and non–CSCs. The assumption is that tumour and stemness properties belong to the same cell in the same way. (Reprinted by permission from Macmillan Publishers Ltd: «Nature Medicine», Gupta P.B., Chaffer C.L. and Weinberg R.A. Cancer stem cells: mirage or reality?, vol. : –, Copyright ).
. Why reductionism is not impossible
From a complementary perspective, we could say that emergent properties require that any mechanisms be framed in the adequate PCMS, and that no auto–referential explanatory power is conceded to the mechanism as such. Because of its mechanistic structure, these explanations have to consider the difference of the epistemological role of the parts’ and of the system’s behaviour. This distinction resolves the apparent circularity and ambiguity of the abovementioned more dynamic definitions of mechanisms (Section .). The case study will now help us to understand better the problems that simple mechanistic reductions face, and to adopt the PCMS view as a possible solution to these problems. ... Proximate causes Both the discussion of what mechanisms explain (Section .) and the examples we had from cancer research (Section .), shifted the focus on how different levels of biological organization are regulated. The role of PCMS and the concern about the ambiguity that MDC’s account of mechanistic explanations seems to entail can be better understood. The resulting picture is consistent with Craver’s characterization of mechanistic explanations as multilevel causal explanations that “explain by showing how an event fits into a causal nexus” (Craver , ), which hints at the possibility that multilevel explanations are constitutive and need not refer exclusively to proximate causes (Richardson and Stephan ). If the mechanistic dimension of the explanatory account takes over the traditional causal explanation in efficient terms, the question is: what is determining which the relevant entities and activities in a mechanism are. These tensions about higher–level properties and the kind of biological interactions captured in mechanistic accounts are nicely expressed by Lewis Carroll’s pun about impossible/impassible. Appealing to inter–level experiments as “tools for determining what the relevant entities and activities in a mechanism are, for determining how they are nested in a multilevel hierarchy, and for showing how a given component is integrated within its mechanistic context” (Darden and Craver , ) does not explain the relative dependence of the parts on the wholes, and does not overcome the ambiguity that the definition of mechanistic explanations entails, like those related
How science works
with tumour cell proliferation mentioned above. The answer to the questions posed by the dynamic behaviour of a biological system in mechanistic terms, i.e. “it works because there are mechanisms”, is in fact always admissible, but hardly useful to understanding what is going on in biological regulatory processes without a clarification of how we get to those mechanisms: i.e. what is passible. Multilevel experiments are actually performed in biological sciences, but their explanatory power cannot rely just on the identification of further mechanisms. Consequently, the commonly accepted idea that “a mechanism must trace how a phenomenon is caused using the objects and activities appropriate to the field and must account for each step in this process, leaving no gaps unaccounted for [my emphasis]” (Delehanty , ) is still challenged by the characterization of “the objects and activities appropriate to the field”. They are, in fact, not autonomous from an epistemological point of view, and a gap has to be filled anyway. In other words, acknowledging that mechanisms are always in place doesn’t settle the issue of the reducibility of biological organizational features, or of an epistemological hegemony of mechanisms in science, as the priority in scientific practice of the PCMS identification shows. Such deadlock is overcome only by a “system level understanding” (Kitano ) of biological phenomena, which I suggest is extremely helpful for the philosophical understanding of scientific practice . Such system level understanding characterizes scientific explanations in biological sciences and is strictly related to the epistemological status of generalizations in biological explanations (see also Chapter ). As Richardson and Stephan put it: “Whether or not this is the only route to explain complex systems, genuine understanding is reached when we are able to redescribe a process, or a complex system, with a grade of resolution that allows us to see the relevant components ‘at work’” (Richardson and Stephan , ). Given the inadequacy of mechanism identification to define the relevant parts and the causal relationship that are supposed to explain the system’s behaviour, we might consider the possibility that mul. This seems to be a promising way to understand the limits and possibilities of Woodward’s account of causation in biology, and to further distinguish the epistemological role of specificity, stability and level of analysis in practice (Woodward ).
. Why reductionism is not impossible
tilevel explanations entail different kinds of biological interactions (or part–whole relationships), whose features might be eventually captured as ‘proximate causes’ through a different strategy, derived from the notion of PCMS, which only implicitly involves notions like ‘proximate causes’, ‘relevant components’, and ‘system level understanding’. Their epistemological role, in fact, requires pondering two points that are left open in Schaffner’s account: a) what determines the causal model system to be ‘preferred’ in nomological terms , and b) a further clarification of the requirements that reductions (also mechanistic–reductions) rely upon — and which justifies their characterization in terms of creeping reductions through the PCMS notion. Briefly, my point is that whenever (almost always) scientific questions are related with the inter–level regulation of a biological system ) the need to bring emergence into the picture is not a claim about the ‘in principle’ impossibility of reduction, but a question on the requirements for reductions, and raises the issue of the epistemological status of the condition for reductions in biological sciences, and ) that these requirements are only implicitly captured by creeping reduction through the notion of PCMS. In partial reductions a different epistemological role should be acknowledged to the parts and to the system by distinguishing, in the process of reduction, different steps to identify the level of generalization, and the dependence of the parts’ identification on the higher–level features through the definition of a PCMS. .. Impassible reductionism Why doesn’t a strong and simple mechanistic reductionist account work for biological phenomena? If relations among parts weren’t relevant in nomological terms, so that parts could be considered homogeneous and thus studied in average terms, reductionism in the traditional sense would work and this has been, in my view, one of the major insights that we ever had in science (cfr. gas laws). Prob. Visually one model is clearly favoured among others in scientific papers, and arguments are given for its superiority in the specific article, in terms of experimental results, but their epistemological import is almost always overlooked as scientists do not need it. There is still a philosophical question here that is related to the nomological dimension of such explanatory models in science.
How science works
lems arise when we assume this requirement — intrinsic equivalence of constituent parts of a system and completeness of explanation in terms of parts — as the only intrinsic conditions of the explanatory power of models. Other relational factors (e.g. like topological or contextual factors) are left aside as a mere additional methodological issue of the explanatory enterprise. Reductionism in these terms is simply ‘impassible’, that is: it does not equip us with the tools to grasp some crucial features of the inter–level regulatory features to which biological complexity is linked. Without these tools we cannot solve some paradoxes that emerge in the abovementioned examples from cancer biology. It is in fact difficult to affirm that reduction without specifying the contextual interactions, is impassible, but that, when inter–relations are provided, an explanation or more sophisticated reduction (like in an innocuous emergence) works. The answer is that this would not be the case if, as I am suggesting, the PCMS element, to which the specification of the contextual interactions belong, implies a conceptual change in the definition of the relata (from genes to cells to, eventually, the tissue organization in the biology of cancer). A series of PCMS can still be identified but their explananda will be different. Further discussion of this point might require a clarification of the historical and evolutionary dimension of the biological systems that is nevertheless beyond our scope in this chapter. The same happens with strong mechanistic claims: they agree, for example, that the context is relevant but there are no tools to explain in what sense it is. It often seems to be an ad hoc assumption or a statement that arises only from methodological concerns or aims. This is in fact the case when supporters of mechanistic explanations advocate for an integrative approach to inter–level phenomena (Darden and Craver ). Biological explanations are multilevel in the sense that they don’t refer exclusively to ‘proximate causes’, although at a concrete level of analysis, proximate causes can always be identified (mechanisms). When scientists make claims about the relevance of emergent properties in scientific practice, they first and foremost draw attention to the epistemological features of the relata that integrate the structure of biological explanations: how they are identified and what kind of relationship holds among them. When this is neglected, impassibility occurs (cfr. figure two from this point of view as well). Not all contexts
. Why reductionism is not impossible
and functional activities of their parts are equally relevant to a specific scientific question. Accuracy of an explanatory account depends not only on the level of details gained through different methodologies, when facing complex, multi–level biological phenomena, but also there is interplay and reciprocal dependence between the scientific question and the phenomenon to be considered in the first place. This is the same problem entailed in the Doorknob argument, which Alice has not been able to grasp. I would even suggest that there is no interest in advocating an ‘in principle’ impossibility of reduction in science. What just seems unavoidable is an analysis of its condition of validity or requirements. Framing the issue in these terms, dichotomies between reductionist and non–reductionist features of scientific explanation can be overcome. On one hand, the reductionist feature of the CM (i.e. the appeal to entities that are parts) is heavily shaped by non reductionist features, that is the causally qualitative dimension of the model entailed by the relationship between the explananda and the explanans (relationship that implies that the definition of the explanans is dependent on the higher–level properties that the explananda entail). On the other hand, some of the conditions that Schaffner lists among the non–reductive elements of CM should be reconsidered. In real terms, adopting this distinction between impossible and impassible reductionism helps us to understand why concerns about the possibility of finding out more relevant or new ‘key elements’ in explanatory terms is no further relevant. Theories and explanations are clearly partial, in the sense that they are open to integration (for example depending on the instrumental tools adopted), but they don’t need to fear new attempts to explain a phenomenon because their explanatory driving force is not based on the kind of proof adopted but on the kind of question to be answered. A real challenging alternative will take the form from a new conceptual generalization of biological properties in a different field of scientific inquiry, not from the same one. Instead, this kind of fear threatens the work of those adopting mechanistic–reductionist perspectives that only apparently seem to be more open to changes in scientific research. Usually the doubt about the existence of ‘new key entities’ that might explain something more about a phenomenon when the right questions have been posed does not arise from the scientific requirement of accuracy of the scientific
How science works
inquiry, but from the doubt that something different than mechanisms is needed for useful reduction in science. The real challenge in biological science, and I would say in science in general, is related to the human capability to frame the question so that relevant entities emerge. Eventually, this is how reference to ‘middle–level’ entities of reductions should be understood. Reduction always has a component that identifies its models as causally qualitative (cfr. discussion about PCMSs) while obviously the test models are at a higher level because of the normative dimension entailed in the kind of generalities characterizing the definition of the PCMS. These components are part of the process of reduction itself. .. Reductions in scientific practice The relationship between the notion of ‘proximate causes’ and relevant or ‘key components’ is now easier to spell out. What determines the causal model system (PCMS) to be ‘preferable’ is in part determined by the resolution of scientists to grasp the mesoscopic level where “organizational principles act on the elementary biological units that will become altered, or constrained, by both their mutual interaction and the interaction with the surrounding environment” (Bizzarri et al. , ). A reduction is thus the process of identifying the explanatory system, often called level of explanation, by means of conceptually linking the dynamic properties of the component parts with the feature of the overall behaviour to be explained. Reduction depends on the scientific question, on the kind of biological properties under inquiry, and on the pragmatic aims of the researcher (i.e. instrumental apparatus and concrete interests). There are degrees and thresholds that can enrich this picture, but it is nevertheless necessary to accept that the generalizations used in the definition of the PCMS –which is the core of the explanatory model– change firstly depending on the issue at stake (genetic accounts explain less than CSS accounts when asking about the rate of cancer spreading and the temporal relevance of epigenetic factors, etc.). What partial reductions do is precisely to distinguish the epistemological roles that parts and contextual factors play in biological explanation, and to leave open the possibility that other scientific questions (usually identified
. Why reductionism is not impossible
by scientists with ‘why’ questions as opposed to ‘how’ questions) can be asked and adequately, scientifically speaking, answered. Moreover, if reductions in science are not meant to explain the world, but to give us pieces of knowledge about it, while defining the way we get those pieces of knowledge through empirical research, concerns about the partiality of reductions disappear. Explanations/models don’t even try to explain all the cases of the scientific problem, in the sense of all the possible questions that a scientific problem might inspire. They are much more ambitious: they try to grasp some features of a phenomenon through models that can eventually be useful to control it. This also implies that even if different levels of explanation are possible for the same phenomenon, they are not usually answering the same question. When the question, instead, is explicitly addressed, the same models tend to converge towards the same level of explanation. Cancer research gives us an example (for a more extended review on this point see also Bertolaso a) and the data in these chapters show that the scientific question is able to drive the evolution of explanatory models towards a different explanatory level (from tumour genes to cancer cells), without an a priori definition of a privileged level of causal explanation. Human reason is able to sail through the deep waters of biological contingency and complexity, bringing into the same explanatory account two epistemological dimensions: the definition of the PCMS and the abstraction of some properties of the parts to explain features of the system’s behaviour. Finally, acknowledging the role that PCMS plays in creeping reductions helps us to understand the relationship between the pragmatic aspect of models and the validity of the elements that contribute to the explanation of the process as well. In other words, an antireductionist claim is intrinsically related to the structure of biological models and explanations (and not to the defence of absolute ‘in principle’ emergent properties): there is no privileged level at which a multi–level phenomena can be causally explained (something supporters of a mechanistic view of science would agree with), and the identification of the relata of any mechanistic explanation is dependent on the levels’ properties to be explained. Within this framework, I suggest that the role and status of connectability assumptions (CAs) should be reconsidered as well. Schaffner
How science works
rightly assumes that they can be either causal sequences or identities. However, their dual features stand in the same relation I described above when saying that structures and functions cannot be separated in scientific inquiry. It is the specificity of the system that calls for both features, and the continuity of levels and of degrees of complexity in the natural world brings them into the picture, so that reductions are not only ‘possible’ (cfr. Alice and the Doorknob dialogue) but also required in science. CAs do not seem to constitute a third condition for reductions, but characterize reductions as a reminder of the non–reductive dimension of any biological explanation. However, further discussion of CAs has to take other features of biological systems into account, features which do not emerge in the examples I have given here or in the bibliography, and therefore remain an open issue for future papers. .. Conclusions Reductions in scientific practice are mainly related to the process of identification of the explanatory model. Once this is set, the requirements for validity of scientific explanations, or reductions, work without requiring any assumption about where the explanatory elements should be found. The explanans are often molecular parts of the organism, but are identified by virtue of their relationship with the higher–level macro property or an end state that specifies the explananda. The non–reductionist dimension that intrinsically characterizes, in my opinion, biological explanations is thus related to the definition of parts and how we understand the structure of the world that is not, instead, mechanistically definable although, as shown, there are intrinsic features of the world, and of the way we know it, that allow mechanisms to be identified. Thus, the central thesis of this chapter is that if reductionist–mechanistic explanations work it’s because of the non–reductionist dimension that characterizes the definition of their relata. When this is not acknowledged, ‘impassibility’ occurs. The discussion about reductionism in biological sciences does not . See Chapter for a clarification in which sense sequences and identities can be seen as different aspects of the same thing. As Silvia Caianiello noted, this might be also applied to the discussion of the relationship between structures and functions of biological systems, as they are collective predicates.
. Why reductionism is not impossible
mainly concern, from this point of view, what mechanisms can explain but why they actually explain something when the question is on the dynamic behaviour of the component parts of the explanatory system.
References B M. (a), Towards an Integrated View of the Neoplastic Phenomena in Cancer Research. History and Philosophy of the Life Sciences, : –. ——— (b), The neoplastic process and the problems with the attribution of function. Rivista di Biologia / Biology Forum, : –. B W., R R. (), Discovering complexity. In Decomposition and Localization as Strategies in Scientific Research. MIT Press, Cambridge, MA. B M., C A., C F., D’A F. (), Beyond the oncogene paradigm: Understanding complexity in cancerogenesis. Acta Biotheoretica, : –. B M., G A., C A., D’A F., S A.M., S C. (), Fractal analysis in a systems biology approach to cancer. Seminars in Cancer Biology, : –. B F.C., B F.J., H J.S., W H.V. (), Systems biology: Philosophical foundations. Elsevier Science, Oxford. C C.F. (), Role Functions, Mechanisms, and Hierarchy. Philosophy of Science : –. ——— (), Beyond Reduction: Mechanisms, Multifield Integration and the Unity of Neuroscience. Studies in History and Philosophy of Biological and Biomedical Sciences, : –. D P., C R.W., C M.F. (), “Cancer Stem Cells: Models and Concepts”, Annual Review of Medicine, : –. D L., C K. (), Reductionism in biology, Encyclopedia of Life Sciences, John Wiley & Sons, pp. –. D S E.M. (), Targeting Wnt signaling in colon cancer stem cells, Clinical Cancer Research, : –. D M. (), Emergent properties and the context objection to reduction. Biology and Philosophy, : –.
How science works
D J. (), It is not possible to reduce biological explanations to explanations in chemistry and / or physics. In Contemporary debate in philosophy of biology. Edited by Ayala J and Arp R, Wiley–Blackwell, Oxford. F, A.P. (), Phenotypic plasticity and the epigenetics of human disease. Nature, : –. ——— O R., H S. (), The epigenetic progenitor origin of human cancer. Nature Reviews Genetics, : –. G S. (), Rethinking mechanistic explanation. Philosophy of Science, : S–S. G V., P E., H W., M E., H B. (), Epigenetic changes may contribute to the formation and spontaneous regression of retinobiastoma. Human Genetics : –. G P.B., C C.L., W R.A. (), Cancer stem cells: Mirage or reality? Nature Medicine, : –. H W.C., W R.A. (), Modeling the molecular circuitry of cancer. Nature Reviews Cancer, : –. H D., W R.A. (), The hallmarks of cancer. Cell, : –. H H.H., L G., B S., Y K.J., S J., Y C.J. (), Clonal and non–clonal chromosome aberrations and genome variation and aberration. Genome, :–. I D.E. (), Can cancer be reversed by engineering the tumor microenvironment? Seminars in Cancer Biology, : –. K S.A. (), Articulation of parts: Explanation in biology and the rational search for them. Boston Studies in Philosophy of Science, : –. K J. (), Emergence: Core ideas and issues. Synthese, : –. K H. (), Systems biology: A brief overview. Science, : –. M P, D L C C (), Thinking about mechanisms. Philosophy of Science, : –. M S. (), The Emperor of All Maladies: A Biography of Cancer. Simon & Schuster, New York. M S.D. (), Emergence: Logical, functional and dynamical. Synthese, : –.
. Why reductionism is not impossible
——— (), Unsimple truths. Science, complexity and policy. University of Chicago Press, Chicago. M L. (), What genes can't do. MIT Press, Cambridge, MA. N C.M., B M.J. (), Of extracellular matrix, scaffolds, and signaling: Tissue architecture regulates development, homeostasis, and cancer. Annual Review of Cell Developmental Biology, : –. R R.C., S A. (), Mechanism and mechanical explanation in systems biology. In Boogerd FC, Bruggerman FJ, Hofmeyr JS., Westerhoff HV Systems biology: Philosophical foundations. Elsevier, New York, – . R H. (), Fields and field cancerization: The preneoplastic origins of cancer. Bioessays, : –. S K.F. (), Discovery and explanation in biology and medicine. University of Chicago Press, Chicago. ——— (), Neuroethics: Reductionism, emergence, and decision–making capacities. In Marcus SJ (Eds), Neuroethics: Mapping the Field. The Dana Foundation Press, –. ——— (), Reduction: The Cheshire Cat problem and a return to roots. Synthese, : –. ——— (a), Reduction and Reductionism in Psychiatry, Draft version: To appear in Fulford KWM et al. (Eds), Oxford Handbook of Philosophy and Psychiatry, Oxford University Press, Oxford (in press). ——— (b), Behaving. What’s genetic and what’s not and why should we care? Monograph (under final review at Oxford University Press). S M. (), Reduction, Emergence and Explanation in Machamer PK and Silberstein M (eds.), Guide to the Philosophy of Science, Blackwell. V A.L., G R., R B.A. (), Brain Tumour Stem Cells. Nature Reviews Cancer, : –. V L., S M.R., K K., S G., M J.P. (), Cancer stem cells – old concepts, new insights. Cell Death and Differentiation, : –. W J. (), Causation in biology: Stability, specificity, and the choice of levels of explanation. Biology and Philosophy, : –.
Chapter V
What does the context matter? .. Introduction Explanations characterize science and the knowledge we get through science of the natural world. Reflections on how different explanatory enterprises and projects emerge and characterize scientific practice have been developed (Brigandt , Potochnik ). However, an analysis of the structure of biological explanation in itself has been often overlooked. In particular, the strict relationship between the structure of biological explanations and their context dependency has been difficult to capture. Like a secondary side issue, the context argument eventually comes into the picture as an ad hoc explanatory tool in mechanistic accounts or is left out as an unresolved issue that still challenges reductionism in biological sciences (Moss ). The context dependency either becomes an argument against reductionism (Brigandt and Love ), although still without a clarification of its epistemological status and role, or is included in an expanded account of mechanisms (Delanthy ). In this case, however, the argument against this inclusive process is the existence of emergent properties that eventually push the whole discussion back to the original criticisms, which the reductionist–mechanistic accounts had to face (Chapter ). Often a way out of these tensions has been to consider the context’s relevance as a methodological recommendation (Potochnik ). However, when the context is seen as a mere methodological feature of the biological explanation, mechanistic accounts shift towards multilevel and more complicated accounts in a never–ending inclusive process of new elements. To the contrary, systemic accounts adopt a more holistic stance, in which the context does play a relevant explanatory role that is not adequately clarified. Cancer research is a paradigmatic case, where
How science works
two different explanatory models seem to challenge each other on these terms (cfr. Section .). My strategy is to analyse what the context matters in discussions about biological explanations and biological behaviours and is based on the above–mentioned scientific literature on cancer. I understand biological behaviour as a dynamic process, which implies an organizational and evolutionary dimension usually characterized by an inter–level regulatory phenomenology. My thesis is that the relevance of the context has to be accounted in order to capture how science works. In fact, acknowledging the role of context in our understanding of the natural world proves fruitful and contributes to reshaping our concept of reduction and explanation in biological sciences. In Sections . and .. I analyse how the context issue emerges and is accounted for within some debates on reduction and biological explanations. In the second part of the chapter (Section .) I will focus on some features of explanatory models of cancer where a debate on reductionism and context dependency arises from the effort to explain the peculiar behaviour of tumour cells. What emerges is a distinction between Type and Token Context–Dependency in biological explanations. The first one is related to the level of generalization or conceptualization of the explanans and the second one to the pragmatic focus on different contrast classes. Moreover, the token context–dependency is secondary, from a procedural point of view, to the type context–dependency, while their intrinsic relationship reframes reductions in scientific practice in terms of type–token relationships. The framework I am proposing also allows us to understand better, on one hand, the process of searching for necessary and sufficient conditions that causally account for dynamic behaviours of biological systems and, on the other hand, the role of mechanistic explanations. In the final Section I will present some arguments to overcome tensions between reductionist and anti–reductionist accounts. .. Framing the picture The debate about reductionism in bio–medical sciences has been historically related to three different dimensions of reductionism: on-
. What does the context matter?
tological, epistemological methodological reductionism (Ayala , Ayala and Arp ). These dimensions have been often treated independently and their relationship remains an open question (cfr. discussion in Brigandt and Love ). Nevertheless they link up in a complex manner with both epistemic and ontological implications when coming to the question if a biological explanation can be reduced to physics and chemistry, reduction that is commonly understood both in scientific and philosophical literature refers to a reduction to molecular terms. Models and mechanisms usually adopt molecular vocabulary. However, to say that models are usually presented in molecular terms does not logically imply that molecular parts hold the explanatory endeavour of such models. It might be the case that molecular identification of the parts is just related with the instrumental dimension of the scientific enterprise and that the explanatory enterprise requires an integration of the molecular subject when explaining biological behaviour in order to explain, for example, to what extent the (behaviour of a) part is relevant for the global dynamics of the system it belongs to. Intuitively this is the case, for example, when we consider that stem cells are clearly cells (and thus molecular parts of a tissue), but that their relevance for understanding the process of tissue differentiation relies upon their qualitative feature of stemness rather than their molecular status as cellular parts (see Chapter ). Different paths have been followed in trying to describe and understand the use that scientists make of terms like mechanisms and models (Craver , Darden and Craver , Schaffner ). These authors acknowledge that science, and especially life science, works on multilevel models that take the form of partial reductions (Schaffner ). The common assumption is that biological explanations work through the identification of molecular parts that, with their interactions, explain the system’s global behaviour. The partiality of the reductive explanation is related with the incompleteness of any explanatory model, with a causally qualitative dimension of models from which axioms are derived in the reductive process and with the role of higher–level, rather than molecular, properties in the design of experimental models (ibidem). A mechanistic–molecular account results, to the least, simplistic with regards to what is required to get a satisfactory explanation of a
How science works
biological behaviour. Moreover, what biological behaviour implies, so that it forces towards non–reductionist accounts in biology and towards inter–level explanations, has not been addressed yet. To clarify how reductionist aims and partial reductions are linked with the structure of biological explanation I will focus on the relation among the relata of an explanation, i.e. on the terms that constitute the explanatory statement. This means to analyse how the concepts of a scientific explanatory reduction of a biological behaviour are identified in the process of scientific inquiry. .. Reducing biological explanations A recent debate about the possibility of reducing biological explanations to explanations in physics and chemistry (Keller , Duprè ) meets these concerns and highlights some interesting points helping us to spell out the terms of the relation that characterizes reductionist attempts. Originally the issue at stake is, as in the traditional debates on reductionism, what should be considered explanatorily fundamental in biological sciences. The debate is articulated on the divergence among reductionist and anti–reductionist accounts of biological phenomena. I will refer to these accounts also in terms of mechanistic and systemic accounts. Under the mechanistic perspective (Keller), physical and chemical parts might be sufficient to explain biological phenomena, while under a systemic perspective this won’t be the case: relations are more relevant than parts and their interactions (Duprè). At the same time, both Keller and Duprè seem to agree on the idea that when addressing higher–level properties or behaviours there is a “dependence of the identity of parts, and the interactions among them, on higher–order effects” (Duprè , , quoting Keller). However this apparent agreement has different meanings for the two . This has been often also presented in terms of necessary and sufficient condition. However I will avoid this terminology in this chapter as the focus is on the characterization of the structure of the biological explanation and not on the process of the experimental design to which, in my opinion, the discussion on necessary and sufficient condition belongs to. The “fundamentality” here is to be understood from an epistemological point of view as will be clearer by the end of the chapter.
. What does the context matter?
authors and mainly depends on how that dependence of the identity of the parts on higher–order effects should be understood. For Duprè this is a fatal objection to a reductionist view, because “properties of constituents cannot themselves be fully understood without a characterization of the larger system of which they are part” (Duprè , ). In Dupré’s account (dispositional) properties are relational, i.e. they cannot be reduced to any information about the parts, and the context is relevant because appealing to the context means “to refer to features of an object’s environment that are necessary to confer on the object a particular capacity (. . . ) Interactions are simply the exercise of such capacities with relation to some other entity that will presumably constitute all or part of that context” (Duprè , ). To the contrary, Keller stresses interactions among parts, leaving apart the definition of the system and its properties. Coherently in her discussion the notion of function is minimalist, a simple feedback mechanism (Keller , ). She criticizes Duprè’s argument by labeling context and interactions as artificial distinctions: “context is simply all those other factors/molecules whose interactions with the object or system in question have not been made explicit and, hence, have not been included in the description” (Keller , ). The context argument becomes the pivotal issue in explaining biological behaviours and their inter–level regulation, and understanding that the context matter is necessary to clarify what is eventually at stake in the debate on reductionism. Dupré’s perspective is more adequate to capture how the explanatory terms are defined, although his anti–reductionist position needs to be integrated by a clarification of the context to account for the structure of scientific explanations. At this point (the context issue) the discussion takes an interesting direction. Supporters of reductionism, like Keller, seem to believe that the whole game consists in proving an explanation of higher–level properties in terms of molecular parts, where the (in principle) logical possibility of such reduction can be easily supported. There are, examples of this kind of reductions in science (typically gas law), and clearly any system in the natural world is constituted of more ‘elementary parts’ and works in accordance with more ‘fundamental laws’. Reductionist therefore concentrate on ‘in what way’ the reductive relation operates. On the other hand anti–reductionist views, like the one defended by Dupré, are more sensitive to the definitions of
How science works
emergent properties, to the discussion of causality notions, and to the identification of the systems (this is why, in this discussion, I call them systemic): they so challenge the in principle possibility of biological explanations to work through mere reductions. Ontological claims typically accompany this kind of stance, although a commitment to a more specific ontological position is usually avoided . Acknowledging this tension, Silberstein offers a more comprehensive framework for the analysis of reductionism (Silberstein ) distinguishing the concern about the relata of an explanation and ‘in what way’ a reductive relationship is construed. Reductionist positions, like the one that Keller embraces, seem ‘impassible’ to this distinction, i.e. unable to grasp the relevance of the specific relationship among the relata of an explanatory account that emerges in the attempt of explaining an inter–level regulatory process (Bertolaso ). Duprè, instead, does distinguish the two dimensions: “I would say that the project of characterizing the entity, which I have said requires reference to the context, and the project of describing what, on a particular occasion, it does, namely interact, are distinct activities” (Duprè , ). I wish to note that to affirm that the capacities of an object of inquiry are not merely consequences of its molecular constitution, but are simultaneously determined by the systems of which the object is part, does not mean that the conceptualization of a part, as an explanatory tool, demands to define the context with which it interacts. This is a further claim, laden with epistemological implications. Dupré seems to acknowledge, to some extent, this point when saying: “the fact that biology — a science — works with concepts that depend on the larger systems of which they are part, as well as on their constituents, is a fatal objection to the claim that ‘it is possible to reduce biological explanations to explanations in chemistry and/or physics’” . “It should be obvious what my worry with such a position will be. Perhaps this would be true for any closed biological system, but then there are no closed biological systems. This is one way of understanding the dependence of the identity of biological entities on context that I have emphasized in this chapter. Bounded biological systems do not supervene on their physical parts because aspects of what they are depend on the context with which they interact, a context always extending beyond any predetermined boundaries. Perhaps I should concede that everything in the universe supervenes on the total physical state of the universe? Perhaps. But, here, we are so deeply into the domain of purely speculative metaphysics that I am more than happy to remain agnostic” (Duprè ).
. What does the context matter?
(Duprè , ). However, this point requires further analysis. It implies to understand in which sense the context could be relevant both in the identification of the parts, which might be just a methodological issue and compatible with mereological or functional account of higher–level properties, and in the conceptual definition of these parts which seems to bear more complex epistemological implications. An example will help in understanding this point. Consider the pancreas, which is supposed to produce insulin. Some cells, once organized in a syncytium, are able to produce that hormone. This means that their connections and interactions within the pancreatic context are crucial for the fulfilment of insulin’s production. However the concept of beta–cells implies a further level of conceptualization. Their definition as beta–cells highlights a specific property of the cells that produce insulin in the pancreas, i.e. their electrical activity and synchronization through a collective behaviour. In the description of their functioning, this electrical property is assumed as default. Both their function and denomination, implying a specific new higher–level property, is determined by their belonging to the organ that produces insulin. The epistemological distinction between the conceptualization of the parts and the description of their functioning within a system (whose representation involves the context argument) will be clarified in the following sections with other examples from cancer biology. For our purposes, at the moment, a first step is possible considering that the tension between reductionist and anti–reductionist positions is due to the different understanding of the interactions that specify the relationship between the parts and the whole, consequently a divergent concept of the context and then of the explanatory role of the context itself. In Keller’s account the relevance of the context is empirical, i.e. there are molecular parts and their interactions, and the context is “simply all those other factors/molecules whose interactions (. . . ) have not been included in the description” (Keller , , already cited). In Duprè’s account it is conceptual, i.e. it is related to the definition of the properties of the parts relevant for the overall behaviour of the system. It remains to be clarified what implication this conceptual dependency of the parts from the whole has from an epistemological point of view and its relationship with ontological claims (in which stronger sense the identity of the parts depends on the whole system they belong to).
How science works
Admitting this sort of articulated context–dependency, I will adopt a double characterization of the context dependency in biological explanation. I call them a type–context–dependency (TyCD) and a token–context–dependency (ToCD). At first sight, the latter is related with an empirical and descriptive issue of the phenomenon object of inquiry (e.g. what interactions are relevant for a realistic explanatory purpose), the former with a conceptual one, (i.e. how definition of parts depend on the global context). From the point of view of the identification of the parts Explaining Type–Context–Dependency Biological Token–Context–Dependency Behaviours
Conceptual Empirical
Although the Type–Token subject clearly needs further discussion, we can now meet the challenge to understand the relation that characterizes reductions in scientific practice, i.e. how the items are linked by a reductive relationship, from the point of view of this context dependency. What remains to clarify now is the nature of the link itself . .. Explaining biological behaviour In the previous section we have seen how the question about the possibility of reduction of biological explanations leads to a discussion on biological behaviours and that the discussion about what the context matters is eventually the core of reductionist debate in biology. What I have also highlighted is that, through the context argument, the focus of my analysis should be on the structure of the biological explanations instead of on the way reductions are performed. What I . This approach allows us to first focus on the explanatory issues and to leave aside the issues related with physicalism and the compositional structure of the world. Note that from this point of view, the conclusion of this study meet the distinction that Mitchell did in a paper on emergence (Mitchell ). Moreover, an extensive literature is already available on molecular biology, its explanatory import and the payoff of the process of the molecularization of biology as well (Dupré , Rosenberg , Okasha , Love b, Robert , Rosenberg , ), but these discussions are not part of our first concern in this study.
. What does the context matter?
am interested to analyse now are the relata of these explanations, that is, what kinds of relata fulfil the requirements for an explanation when the question is on biological behaviours and on the nature of their relationship. This means understanding the conceptual dependency of the parts from the whole in the language of the debate we have previously analysed. Another debate is useful at this point. It dates back to the Nineties when Schaffner published a paper on the Developmentalist challenge (Schaffner , ). His aim was to understand the explanatory role of “simple” models in biological sciences that often implied references to the genetic patrimony of animals. Acknowledging that biology lacks general laws that may allow applying a strict nomological–deductive model (ND–model) in the explanatory enterprise, leads Schaffner to focus on the features of what he called Causal Model (CM) to account for the explanatory enterprise of biological sciences. As he said, in fact, “the structure of biological knowledge, from both epistemic ad logic–of–explanation perspectives, is organized differently from what we find in standards accounts of the physical sciences” (Schaffner b, ch , ). Although the assumption is still that a deductive process holds the explanatory power of the explanation, his conclusion was that the explanans in physics are theories, while in biology they are models. Following authorities in the scientific field, Schaffner first stated that “model systems are a powerful heuristic for biological research” (Schaffner b, ch , ) but he never gave up the possibility to grasp, in epistemological terms, the nomological dimension of biological explanations , going beyond mere heuristic considerations and assuming that models entail a level of idealization that justify their use in different areas and fields of inquiry. The hope was that such models could eventually disclose conserved mechanisms of general applicability. Schaffner is prudent in saying that these kinds of continuities allow us to infer that simple organism models can fully explain the behaviours of much more complex living beings, but he acknowledges that some kind of relevant information is driven by these explanatory models. This is plausible if we consider that some molecular pathways are highly conserved in nature and that also the most complex be. In Schaffner’s account this nomological dimension relies upon the deductive character of explanations as assumed in the ND model as will be clear in the following statements.
How science works
haviours in nature can be heavily affected by the molecular factor. However, this does not necessarily imply the possibility of reduction of higher–level properties’ explanation to explanations in terms of lower level properties. All we can say is that there is continuity and that the challenge is to understand better what this continuity is and how it is captured by the nature of the reductive relationship. As it often happens in history, problems and criticisms bring with them the solution although sometimes implicitly. Schaffner’s support for these models in scientific practice, in fact, immediately faced two challenges that are precisely related with the kind of continuity and reductive relation I am interested in. The first challenge was posed by Gilbert and Jorgensen who claimed that deriving some kind of information from a simple model’s system like C. Elegans (a worm widely used in biological studies for its simple structure of few and well differentiated cells), downsizes and eventually eliminates the amplitude of the developmentalist challenge (Gilbert and Jorgensen , ). The developmentalist challenge, often associated with the Developmental Systems Theory (DST), stated already in the Eighties that the DNA does not contain a program for development and that genes are not the sole and main units of selection. Other factors that are context dependent are relevant in the process of morphogenesis, and evolution equally affects regulatory patterns, other than the genetic ones, that are responsible for the developmental processes in biological organisms. Their concern was articulated in the following way: on one side it is related with an experimental issue — behavioral genetists can only study traits that are present and absent and thus miss variations (examples of this is given in Schaffner b, ch ) , on the other side the question is whether “worm research can say anything useful about interesting research on human cognition” (Schaffner b, ch , ). The latter argument follows some experimental evidence that shows how reduction program for neurogenetics (Gilbert and Jorgensen ) fails because of neurons’ emergent properties. Limits in explain. I will not consider this argument in detail because it is secondary to the main topic of the chapter. However, I believe that clarifying the epistemological role of the context will also contribute to clarifying other experimental issues that are related with the functional definition of explanatory parts. Presence, absence and variations of traits belong, in fact, to this area of discussion as well.
. What does the context matter?
ing any significant human behaviour are highlighted, and Schaffner’s attempts to turn at the worm to look for a basis for behaviour in genes, rejected. As in the previous debate, when the scientific question is on a biological behaviour, the argument of the context dependency is eventually introduced: “a gene may be an essential component of any behaviour, but it does not ‘determine it’ (Gilbert and Jorgensen , ). In the context of a whole organism, single genes cannot determine discrete behaviours. “The predicted behavior of the animal does not emerge from the knowledge of genes and neuronal connections: the player acts independently of his or her teammates” (Gilbert and Jorgensen , ). The challenge is for a deterministic philosophy, which might be dangerous for understanding organisms with consciousness, agency and the relevance of different environments (Gilbert and Jorgensen , ). In spite of this the DST was not interested in the reductionism debate on theories: they react strongly against a reductionist account of their explanatory model. It might not be the formal structure of those reductions that worries them, but the conceptual implication of that reduction for the relata . The distinction I have traced following the previous debate is thus reinforced and also clarified. What the DST supporters claimed was that the relationship between genes and behaviour was not adequately captured by Schaffner’s accounts. The explanans — genes in this case — cannot be considered a satisfactory explanation for an explananda like a biological behaviour. The explananda exceeds its explanans even in the simpler models. In other words, the normative role that genes have in biological development does not follow the deductive structure of the ND model that still holds the explanatory character of Schaffner’s accounts. Genes have no normative priority over development. What DST seems to react . Multiple realizability and variation in biological organisms are often used as an argument at this point, as different pathways are available to perform the same functions and higher–level properties. However I find this argument misleading. I believe, instead, that the approach I am adopting will be able to clarify the real meaning and relevance of this argument as well, but not as an argument against reductionism (see also Sober ). As I have already discussed elsewhere (Bertolaso ) this issue is better understood from the perspective of logical and conceptual derivability (Schaffner ), but this argument is also secondary to the present analysis.l.
How science works
against is the (apparent) attempt of Schaffner to resolve the everlasting tension between nature and nurture through a philosophy of scientific explanation that focuses only on its formal structure. The (apparent) misunderstanding was immediately clarified by Schaffner who reframed and clarified the picture distinguishing a heuristic and epistemological issue: “Worm studies will not tell us anything about consciousness or intention or agency (Gilbert and Jorgensen, ), §, for complexities exist in humans not found in simpler organisms. (. . . ) But some fundamental mechanisms, including simplified analogues of real biological neural nets are emerging in C. elegans studies” (Schaffner , ch , ) and again “A worm is not a human, but worm studies, as well as other animals, may offer important lessons about human psychopathology if yoked to other model systems” (Schaffner , ch , ). In this way, agreement was on the fact that, to some extent, genes control behaviours. There are some aspects of worms and humans behaviours that can be explained at the genetic level. What is at stake is ‘to what extent’ and what determines that ‘to what extent’ or, to put it in a different way, how the ‘fundamentality’ of models and the kind of information we get from simple models should be understood. Now the problem is to clarify the kind of information that these models bear and what universal feature of biological explanations CM conveys . This meets Duprè’s anti–reductionist concern about biological determinism: “I want to deny that the behaviour of the whole is fully determined by the behaviour of, and interactions between, the parts. And hence, the elements of behaviour that are not so determined are what we don’t know when we know everything about the parts and the way they are assembled” (Duprè , ). This is, in my opinion, the crucial issue. It does seem questionable that a mechanistic explanatory reduction, although partial, of biological behaviour is reasonably sufficient to account for the explanatory feature of biological models without any rest. . The specificity of this scientific information also refers to the relationship between the definition of the parts and the actual instantiation in biological explanations. This can easily read in terms of the relationship between genotype and phenotype as well. The point is to understand how genes are defined in order to explain a higher–level behaviour of biological parts, i.e. to spell out the traditional problem of the relationship between the genotype and the phenotype.
. What does the context matter?
It is worthwhile to go deeper in the context argument of these debates. Antireductionist DST positions claim that the context has a relevant role in the explanatory accounts, while reductionist supporters tend to deny it or to confine the argument at a pragmatic level, which is usually understood as the reality based interest of scientists at a given time. As Schaffner put it: “The dangers of DST in its present form, as I see it, is that gives too much to “context” (. . . ) and needs to formulate its categories of interactions more clearly (. . . ). It is not helpful to assert that everything interacts with everything else, but that could be a problem for DST unless it provides us with some form of “priorized ontology” (Schaffner b, ch , ). He is right, but what is missing here is that a “prioritized ontology” is at hand considering how things work from the different perspective of the Type–Token context dependency. Moreover, that “prioritized ontology” is also what developmentalists refer to without being able to clarify its epistemological status. To some extent we can say that, as it is the case for mechanistic and systemic perspectives in the previous debate, genetists and developmentalists stress different aspects of the same picture. Developmentalist’s emphasis is on the context so that the developmentalist challenge has been often addressed as contextualism, i.e. the view that “genes have little meaning . . . per se, only in context with other genes and in the environment that is cellular, extracellular and extraorganismic” (Schaffner , , quoted in Griffiths and Knight ). Geneticists’ emphasis is on genetic parts. However, how genotype and phenotype are related is still an unresolved issue. That not everything matters is commonly accepted. But why some traits matter more that others and why the context matters at all is not yet clear because we are not able to explain in which sense for example genes and environment (the same tension is true in the parts and wholes discussion) can have a priority, again in ‘explanatory’ terms, although in a different sense. The way out for Schaffner and others is to recognize that genes, causally, have parity with other molecules as severally necessary and jointly sufficient conditions (to produce traits); but, epistemically and heuristically, genes do seem to have a primus intra pares status (Schaffner b, Ch , ) . Nevertheless, this issue puts Schaffner . This claim was supported by this author, also referred to as indivisibility, i.e. the idea
How science works
in a peculiar position: as the previous debate shows, how genetic and environmental components are related is not straightforwardly obvious when the claim is that their explanatory power is due to the deductive relationship that involves them. This point is easier to understand, also in the light of the abovementioned biological examples and in the further discussion of cancer biology, if we consider that what is really at stake in the “prioritized ontology” is not a traditional ontological priority (what really exists) but an onto–epistemological priority (how we conceptualize what really exists). There is a non–symmetric dependence in the reductive relationship that defines the explanans and the explananda of a reduction, which the logical framework we inherited from the ND model is not able to capture. For example, we can consider that there are genes, but their definitions depend on the identification and previous characterization of the biological behaviour to be explained. In the part–whole language, we can say that the part and the whole are not ‘parts’ in the same sense. As we will see in the next section, paradoxes emerge when the assumption of a methodological and heuristic priority of genes, as molecular parts, is made explicitly, as it was in some molecular models of cancer. In biology there are two different related senses that hold the relationship among the relata (e.g. genotype and phenotype, or explanatory parts and wholes behaviour). One is represented by the conceptual–empirical relationship, the other by a generalization–instantiation relationship. What both hold is a peculiar relationship among parts and the whole that is related with the normative dimension of the biological explanations. The conceptual link between the interactions and the systemic properties belong to the epistemothat individual genetic and environmental causes cannot be identified by separable effects on the phenotype, and the effect of all causal factors are, in some way, context dependent. Schaffner will never spell out the insight he got through this notion of indivisibility. He instead prefers to adopt a weak notion of emergence where the kind of unpredictability that “means that from total information about genes and environment, we cannot predict an organism’s traits: they are, accordingly, emergent” (Schaffner b, Ch , ). Coherently, he did say that a genetic determinism should be defended, but that developmental noise (an argument for strong emergence in developmental theories) should be viewed with suspicion. In his conceptual framework, in fact, the only way to understand heterogeneity in the behavior of biological “parts” of a “whole” was in terms of noise. This is so because the premises of a deductive model logically entail the terms of the deduction.
. What does the context matter?
logical field. In this sense the normative dimension is related to the specific approach of science to the natural world, i.e. the continuous interplay between the phenomena and our way of abstracting generalizations . Science always moves from a real phenomenon but aims to abstract principles from its dynamics. The pragmatic methodological focus of Schaffner’s strategy on models and their structure gives us further elements to claim that the above sketched relationship is implied in biological models. Schaffner’s “simple system” strategy appears as a variant of the mechanistic and systemic approaches but with an important difference. The context sensitivity, in fact, which is a problem for strongly mechanistic committed accounts, is recovered in the Causal Model (CM) through the concept of Connectability Assumptions (CAs), i.e. conceptual assumptions that, once specified, permit to create a bridge between macrodescriptions to microdescriptions (e.g. a behaviour and genes’ activity). In this way, CAs bring into a unified explanatory framework, although implicitly, the continuity implied by the relationship between the two dimensions of the contextual dependency . Let’s see how this works in more detail. In Schaffner’s account reductions in neuroscience are causal mechanical explanations. The explanans are parts of the organism or process, the explanandum is an event to be explained, usually a macro–property or end state of the system. Referring to concrete examples, Schaffner states that CAs are causal consequences or identities. However, it emerges from my analysis that they are both causal consequences (instantiations, phenotypes) and identities (generalizations, genotypes) because of the peculiar relationship that the type–token dependency established . This point would open an interesting reflection about the use of abstract terms in science. As Buzzoni notes their use is, in fact, legitimate in science as far as we can relate them, conceptually, to observation and our experimental intervention in the world (Buzzoni, ). This reconciles reductionist methodologies because, on one hand, we are aware that there is just an historical difference, at least in principle, between theoretical and empirical terms, and because on the other hand experimental observation depends upon the available tools. . I have shown that only acknowledging this, the explanatory role of the CM models holds (Bertoalso ). Moreover, the same could be said and discussed in terms of biological information and of multiple realizability that characterizes any biological system, so that no common and dominant pathway exists, while the “generative entrenchment” (Wimsatt ) constraints explain why history matters in biology, entrenching and conserving generic and contingent states alike.
How science works
among the conceptual generalization and the instantiation of the scientific explananda. Therefore, the apparent paradoxical dual inter–level character of partial reductions is eventually related to the intrinsic relationship among TyCD and ToCD. The first one is related with the level of generality or explanatory system, the latter with the actual instantiation of that generality usually described or represented in molecular terms. The same process of identification of the explanatory parts I have analysed in the previous section can be now described in terms of generalizations and instantiations.
Explaining Type–Context–Dependency Biological Token–Context–Dependency Behaviours
From the point of view of parts–whole definition
From the point of view of parts’ identification
Generalizations
Conceptual
Instantiation
Empirical
.. Explaining why a Tumour Cell behaves in that way Apparent different issues (genotype and phenotype relationship; macro and micro–descriptions of a systemic behaviour) converge in one, the context dependence argument, when the focus is on the dynamics that a behaviour entails and on the implication of scientific explanations of such dynamics. It is thus legitimate to assume that although this relationship is expressed in molecular terms (genes, proteins, cells, tissues, without any need to appeal to vitalistic forces), this does not imply that its explanatory potential relies upon those molecular, i.e. physical and chemical, parts. Anti–reductionist positions have often found in relevance of the context, for the functioning of a molecular part, an argument to make this point as well. This is also the case in cancer research where different explanatory models try to address the same question. Why does a Tumour Cell (TC) behave like that? Cancer is a multi–level process that involves genes, cells and tissues and the outcome of which is the overall disruption of tissue organization and the progressive tumour phenotype of cancer cells. Such phenotype has been characterized by a set of functions, commonly known as the hallmarks of cancer: self–sufficiency in growth
. What does the context matter?
signals; insensitivity to antigrowth signals; capability to evade apoptosis; a limitless replicationpotential; sustained angiogenesis; capability to invade and metastasize (Hanahan and Weinberg ). From a systemic point of view, these traits can also be listed in terms of loss: loss of feedback control on their proliferation, loss of dependence from growth factors, loss of capability to undergo apoptosis, loss of control of local constraints to differentiate properly (cfr. Bertolaso ). Therefore two different kinds of perspectives are present in cancer research and work through different definitions of the disease. A reductionist mechanistic perspective (MP) focuses on the behaviour of a single tumour cell, so that cancer “is a heterogeneous disease often requiring a complexity of alterations to drive a normal cell to a malignancy and ultimately to a metastatic state” (Edelman et al. , e). From this point of view cancer is understood as a disease involving dynamic changes in the genome. Cancer cells have defects in regulatory circuits that govern normal cell proliferation and homeostasis. “[B]y simplifying the nature of cancer — portraying it as a cell–autonomous process intrinsic to the cancer cell — (. . . ) cancer development depends upon changes in the heterotypic interactions between incipient tumor cells and their normal neighbours” (Hanahan e Weinberg , ). An anti–reductionist Systemic Perspective (SP), instead, looks at cancer as the “the result of the disruption in the tissue’s architecture” (Sonnenschein e Soto ), and avoids references to any molecular part to account for cancer’s origin and its development. What really matters are relations among cells: the organizational structure of the tissue, and the context play a relevant role in understanding the process of tumour progression. The reductionist and antireductionist perspectives or the genetic and developmental account of biological behaviour are eventually merged in these explanatory models of carcinogenesis. An on–going debate in scientific literature is currently discussing which theory of cancer might be more satisfactory. My interest, here, is to analyse what they have in common and at what level of the explanatory process they present convergences that can help us understand the process itself and eventually clarify what the context matters in this case. Let’s start putting in order the terms of the above–mentioned definitions.
How science works
Aiming to give a definition of cancer, a first step is to look at the same explananda, i.e. the aberrant behaviour of tumour cells (TCs). The explanans seems, instead, to be different: a cell, a process of cell differentiation and tissue organization. To clarify how these definitions work in scientific practice we have to analyse the first question they aim to answer: Why a TC does X? TC is a tumour cell, while X in the biology of cancer has been characterized by the abovementioned hallmarks of cancer. However, the process of identification of the relata is different in the MP and SP. In the MP dominates the reductionist assumption that eventually identifies the explanatory parts with the genes, so that the research programs have all been shaped by this assumption. No genes have been, however, identified while a wider and more complex circuit of molecular pathways has been reconstructed in the attempt to find key molecular factors responsible for the neoplastic process. Finally, the context is added as a relevant factor for carcinogenesis, but is still understood in terms of interactions among causally equivalent parts. But, assuming that the features of X belong to TC originally, eventually leads to paradoxes. The mechanistic approach seems to fail because not able to grasp the peculiar relationship that holds the characterization of TC and thus in which sense X properties belong to TC in a necessary, and explanatory sense. Instead, when TC is not understood in absolute terms, i.e. as parts that have X properties in their own right, the explanatory model is to some extent satisfactory. It can explain why tumour cells do not give origin to new tumours with the same frequency, why a tumour mass shows a high level of heterogeneity, etc. This, however, implies recognizing that some features of TC mark their biological identity in virtue of their belonging to a tissue, like in the example I gave above of the beta–cells. The normative dimension of their behavioural pattern has to refer to the context in a very peculiar sense. The context has no causal parity in molecular or functional elements in explanatory models. A TC belongs to a more general class C (cells), and only acknowledging that some features belong, as such, first to that class C, like proliferation, the peculiar behaviour of TC can be understood. Simple causal relationship (TC does this and that) is not sufficient to explain carcinogenesis. The natural story of TC becomes relevant at this point
. What does the context matter?
and it is strictly related with the biological explanatory role commonly attributed to the context. The relevance of the context argument in biological sciences is related not with the logical structure of explanations and formal structure of models, but with the dual meaning that the context acquires in biology It is an ontological category that is physically related with the biological information embodied in what is usually defined as microenvironment or biological context (Nelson and Bissel ), epistemologically relevant in the identification and definition of the explanans. This point is easier to understand if we rely upon the explanatory structure of the SP, which acknowledges the conceptual contextual dependence of tumour cells and of the whole process of carcinogenesis. Such dependence has strong empirical evidence in studies that show how the organizational structure of the tissue plays a crucial role in cancer origin and onset. This approach challenged both the idea that cancer arises from a single somatic cell that has accumulated multiple mutations in its DNA, and that the normal state of cells in metazoans is quiescence — as purported by the MP — and that cancer is a disease linked to cell proliferation caused by mutations in genes that control proliferation and cell cycle (Sonnenschein and Soto, ). This new explanatory account attains the level of a theory not because of the explanatory models of cancer that it offers, but because it provides a general framework for dealing with the neoplastic process. What I mentioned earlier — the natural story — is here spelt out in evolutionary terms. Cells that belong to a multicellular organism are continuously proliferating, while their process of differentiation is progressively constrained by built–in boundaries through different kinds of feedback and inhibitory processes. The organic system takes over the control of the timing of the biological cycles and development. Reference to the “natural story” should be understood as a TyCD that implies a more general categorization of the actual instantiation. This natural story, which is usually translated in biological practice in terms of parts that part ‘belon to’ that system, becomes crucial to identify correctly the contrast class to adopt. What is ‘more fundamental’ should be understood, by means of an identification process of a level of analysis. Instead the question and search of more fundamental explanatory parts is misleading, as the search of necessary and sufficient conditions in terms of molecular parts.
How science works
What is ‘more fundamental’ is defined by the operational identification of local (and in this sense contingent) TyCD (a tissue cell that stops behaving in an integrate way) and ToCD (actual TC) relationship. Once this is acknowledged, different approaches are possible in cancer research. Different contrast classes can be identified, but the explanatory models that focus on, for example, TC’s contrast classes or X’s contrast classes are not only not incompatible because they may eventually have different explananda, but imply each other through the TyCD (i.e. what a thing is) and ToCD (i.e. what parts we select in an explanatory account) . In cancer research, for example, the research program can focus on a (dis)functional property of TC or on the biological origin of those TC, like in the case of CSC (cfr. Chapter ). The explananda is at the tissue level (why aberrant differentiation?) or at the cellular level (why does TC proliferate?) respectively, etc. The SP suggests that cancer is like a process of normal histogenesis and tissue repair, intimately involving the three–dimensional tissue organization (Maffini et al. ). In this way, a tumour would not be a caricature of an organ, but rather a body that has not successfully reached its final configuration. In analogy to recent theories of the histogenesis of development and embryogenesis, the Tissue Organization Field Theory (TOFT) perspective considers the organism as the most appropriate organ to study a complex phenomenon such as cancer (Sonnenschein and Soto ): the genome would not be the driving factor of tumour development.
Type-ContextExplaining Dependency Biological Token-ContextBehaviours Dependency
From the point of view of parts’ definition
From the point of view of parts-whole definition
From the point of view of the dependency of the identity of the parts from the whole
Conceptual
Generalizations
Natural Story
Empirical
Instantiation
Causal Relationships
The explanatory picture, in Schaffner’s account, eventually acquires a very interesting unity at this point: “I want to argue that . Once this process of identification and conceptualization of the explanatory elements is acknowledged, I find the analysis performed by Boniolo interesting from a biological point of view, and sophisticated enough to account for different research programs (Boniolo ).
. What does the context matter?
a satisfactory local explanation model, which I think can illuminate what occurs in partial reductions, has two main substantive components, with each substantive component having a closely related logical/epistemological aspect” (Schaffner b, Ch , ). What I have shown is how there is an “essentiality by locality” of biological explanation (see also Bertolaso et al. ), and that it shall be understood from the TyCD and ToCD perspective. Moreover, the dimensions of the conceptual–empirical and generalization–instantiation relationships mentioned in the previous sections can be distinguished but not separated. Their relationship is not dialectical, it’s relational, so which of them is more relevant depends on the epistemological perspective that is adopted. This kind of relevance is always relative , never absolute and explains well why mechanistic and systemic perspectives in biological sciences are not only complementary but imply each other in different ways . From this point of view, reductionism might be understood as a problem inherent to the Type–Token reductions, while the SP and the MP in cancer research should be understood as mutually related, in the sense that the MP can be considered a particular case of the SP and that the SP is implicitly adopted in the identification and conceptualization of the explanatory molecular parts.
. The relational notion implied in biological explanations emerges from scientific language as well. No scientific research activity in fact sets aside referential intent, even though this doesn’t mean that any scientific argument — that sets as a goal to explain some aspect of nature — will succeed in developing a similar referential intent (Agazzi ). My account specifies in which way this happens given the biological question on the multi–level phenomenology of cancer progression. . What developmentalists usually react against is the separation of these two dimensions. This is why their argument is usually not about non–predictability but about non–explicability of higher–level behaviours. The issue is biological determinism and the real concern, as Griffiths and Knight said, is not with whether we can ‘compute the embryo’ but what we can compute it from. The discussion is on the “whole contextualized system” that recurs from one generation to the next and which accounts for the pattern of inheritance (Griffiths and Knight ), i.e. on the relationship between a level of generalization, that usually entails a notion of biological information, and its instantiations. The context relevance the DST workers were defending was the conceptual one, not the methodological one to which, instead, Schaffner referred to.
How science works
.. Conclusions It has been argued that the requirements for lawfulness rules, in the traditional sense, fails to reflect the reality of scientific practice clearly able to provide viable explanations to account and eventually intervene in the natural, biological, world (see also Mitchell ). My account of “what the context matters” helps to view all these issues from a different and fruitful perspective and to clarify what kind of reductive explanations are performed in science. The relevance of generalization in explanatory accounts is also better understood: it is the context that gives us the interpretative code. Moreover my account contributes to meet the challenge of understanding how we are able to represent biological behaviours without knowing all the molecular details. Actually, the molecular definition of parts always implies different dependencies (TyCD and ToCD) that make sense of unresolved tensions among the different perspectives and the actual instantiations of those conceptual entities. Open to further development remains a more extensive analysis of how we shall understand, from this perspective, biological theories, and a discussion of the specific contribution of evolutionary and developmental perspectives in life sciences.
References A E. (), La questione del realismo scientifico, in Mangione C (Ed), Scienza e filosofia. Saggi in onore di Ludovico Geymonat, Garzanti, Milano, pp. –. A F.J., D T. (Eds), (), Studies in the philosophy of biology. University of California Press, Great Britain. A J., A R. (Eds), (), Contemporary debates in philosophy of biology. Wiley–Blackwell, Oxford. B M. (), Towards an Integrated View of the Neoplastic Phenomena in Cancer Research. History and Philosophy of the Life Sciencies, : –. ——— (), Il Cancro come questione. Modelli Interpretativi e Presupposti Epistemologici. Franco Angeli, Milano.
. What does the context matter?
———, G A., F S. (), The Mesoscopic Level and its Epistemological Relevance in Systems Biology, in Recent Advances in Systems Biology, André XCN Valente AXCN, Gao Y, Sarkar A (Eds), NovaScience (accepted),. ——— (c), Breaking down levels of biological organization, Theoretical Biology Forum, accepted. B G. (),. A Contextualized Approach to Biological Explanation. Philosophy, : –. B I. (), Beyond reduction and pluralism: toward an epistemology of explanatory integration in biology. Erkenntnis, : –. ———, L A. (), Reductionism in Biology, Stanford Encyclopedia of Philosophy. B M. (), Experiment and Theoretical Terms from an Operational Point of view, a cura di Minazzi F. — Roma: Studi internazionali in onore di Evandro Agazzi, Istituto Poligrafico e Zecca dello Stato — Vol. Filosofia, Scienza e Bioetica nel dibattito contemporaneo. C C.F. (), Explaining the brain: mechanisms and the mosaic unity of neuroscience. Oxford University Press, Oxford. D L., C K. (), Reductionism in biology, Encyclopedia of Life Sciences, John Wiley & Sons, pp. –. D M. (), Emergent properties and the context objection to reduction. Biology and Philosophy, : –. D J. (), The disorder of things: metaphysical foundations of the disunity of science. Harvard University Press, Cambridge, MA. ——— (), It is not possible to reduce biological explanations to explanations in chemistry and / or physics. In Contemporary debate in philosophy of biology. Edited by Ayala J and Arp R. Wiley–Blackwell, Oxford. E E.J., G J., C J.T., F P.G., M S. (), Modeling Cancer Progression via Pathway Dependencies. PLoS Computational Biology, : –. K E.F. (), It is possible to reduce biological explanations to explanations in chemistry and/or physiscs? In Contemporary debates in philosophy of biology. Edited by Ayala J and Arp R. Wiley–Blackwell, Oxford. G S.F., J E.M. (), Wormwholes: A commentary on Schaffner K.F. “Genes, Behaviour, and Developmental Emergentism”, Philosophy of Science, : –.
How science works
G P.E., K R.D. (), What is the developmentalist challenge? Philosophy of Science, : –. H D., W R.A. (), The hallmarks of cancer. Cell : –. L A. (), Explaining the ontogeny of form: philosophical issues, in Plutynski A and Sarkar S (Eds), The Blackwell companion to philosophy of biology, Blackwell, Malden, pp. –. M M.V., C J.M., S A.M., S C. (), Stromal regulation of neoplastic development: age–dependent normalization of neoplastic mammary cells by mammary stroma. American Journal of Pathology, :–. M S.D. (), Biological Complexity and integrative pluralism, Cambridge University Press, Cambridge. ——— (), Emergence: Logical, functional and dynamical. Synthese, : –. M L. (), Redundancy, Plasticity and Detachment: The Implications of Comparative Genomics for Evolutionary Thinking , Philosophy of Science, : –. N C.M., B M.J. (), Of extracellular matrix, scaffolds, and signaling: Tissue architecture regulates development, homeostasis, and cancer. Annual Review of Cell and Developmental Biology, : –. O S. (), Evolution and the levels of selection. Oxford University Press, Oxford. P A. (), Explanatory Independence and Epistemic Interdependence: A Case Study of the Optimality Approach. The British Journal For the Philosophy of Science, : –. ———, MG B. (), The Limitations of Hierarchical Organization. Philosophy of Science, : –. R J.S. (), Embryology, epigenesis, and evolution: taking development seriously. Cambridge University Press, New York. R A. (), Reductionism redux: computing the embryo. Biology and Philosophy, : –. ——— (), Darwinian reductionism: or, how to stop worrying and love molecular biology. University of Chicago Press, Chicago. S K.F. (), Discovery and explanation in biology and medicine. University of Chicago Press, Chicago.
. What does the context matter?
——— (), Reduction: The Cheshire Cat problem and a return to roots. Synthese, : –. ——— (), Model Organisms and Behavioral Genetics: a rejoinder. Philosophy of Science, : –. ——— (a), Reduction and Reductionism in Psychiatry, Draft version: To appear in Fulford, KWM et al. (Eds), Oxford Handbook of Philosophy and Psychiatry, Oxford: Oxford University Press (in press),. ——— (b), Behaving. What’s genetic and what’s not and why should we care? Monograph (under final review at Oxford University Press),. S M. (), Reduction, emergence and Explanation, in The Blackwell Guide to the Philosphy of Science, Machamer P, Silberstein M (Eds), Blackwell, Massachussets, USA. S C., S A.M. (), Are Times a’ Changin’ in Carcinogenesis? Endocrinology, : –. ——— (), The Society of Cells: Cancer and Control of Cell Proliferation. Springer–Verlag Inc, New York. W W.C. (), Simple Systems and Phylogenetic Diversity, Philosophy of Science, : –.
Chapter VI
Pluralism out of unification .. Introduction In a paper published in , Mitchell and Dietrich consider the tension between unification and pluralism in biological theory and the fragmentation of, in actual fact, the evolutionary synthesis by molecular evolution. Such analysis suggests the limitations of the ideal of unification in biological sciences but not necessarily for integrating explanations. They thus move on defending an Integrative Pluralism that allows for the integration required for explanations of complex phenomena without unification on a large scale (Mitchell and Dietrich ). In this chapter I wish to explore some epistemological implications of this integrative approach, as I also came to analogous conclusions in another study on complex biological processes (Bertolaso , ch. ). Actually I will consider some arguments for an Integrative Pluralism that Mitchell presents in her publications (Mitchell , ) when trying to couple biological complexity with the possibility of integration, and I will present some considerations that result from the analysis of the previous chapters. While discussing a multi-level approach and its epistemological presuppositions in cancer research, I will leave open some issues, worthy of further research, which are related with biological contingency and its implications in the epistemology of science. The final claim, to some extent still programmatic, is that a pluralistic approach is not just a consequence of our incapability to get a unified view of biological phenomena. More radically, it arises from the intrinsic dynamic unity of the organism that we are able to grasp and adequately study at different levels of its biological organization. The identification of the explanatory level takes place through con
How science works
ceptualization of aspects that are partial, and, precisely because of that, theoretical (cfr. also Buzzoni , and Chapter above). This is also related with the need to black boxing some of the causal dimensions of a biological process, and the way this affects the process of reduction in science. Finally it also asks, in my opinion, for a reconceptualization of biological contingency and of its epistemological role in our understanding of the living systems’ behaviour. .. Integrative pluralism As mentioned in Chapter , a question that has long concerned epistemologists and philosophers of science is whether science is a process, which can be expected at some time to reach a completion. If one does suppose that science will one day come to a complete account of the natural world, one should have a conception of its goal. Traditionally, such unification has been conceived of as proceeding by means of the reduction of each part of science to an ultimate level of explanation. In biology this often means reduction to detailed molecular mechanisms. However, the complexity of biological phenomena still generates a plurality of explanatory theories and models, raising important questions. Different kinds of pluralism have been historically proposed, as reviewed by Mitchell (). Some of them argue for a disunity of science as an epistemological attitude against reductionism, that is inadequate to capture the rich variety of relations among the results of scientific inquiry (Duprè , ). Others are more inclined to a partitioning of biological theories and explanations, given that different questions might arise from the same biological phenomena, so that different answers are to be expected. Pluralism depending on the level of analysis has been defended by Mayr, Tinberg, Sherman (; : ; ) and it is almost a dogma in the biological community. More recently another approach has been presented by Mitchell that focuses on a “more nuanced understanding of the scientific representations of multicomponent, multilevel, evolved complex systems” (Mitchell , . . . ) calling for the integration of the plurality of models typical of biological studies. This plurality reflects the peculiar ontology of complex systems, whose multi-level structure “encourages
. Pluralism out of unification
focused analysis of causal structures” (Mitchell , ). Premises of this pluralism are thus mainly related to the multiplicity of causal paths and to the historical contingency of biological phenomena, so that the type of integration that can occur in the application of models will be itself local. This means that explanations are not necessarily competing. On the contrary, as Mitchell notes, they can be compatible and complementary (ibidem). My interest has been to explore what such compatibility is related to when the question is on inter-level regulatory processes, such as those related to carcinogenesis. In dealing with cancer, different theories and explanatory approaches face the difficulty to couple structural and functional properties at different levels, notwithstanding the fact that its phenomenology is perceived as a disruption of the overall unitary functioning of the organism. Different and plural causes act at the same time; their causal relevance can change over time as well. In some cancers, for example, a mutation can be crucial in the first steps of carcinogenesis although it ends up to be secondary at a second stage . Thus, as in other areas of life sciences, it is the diversity of the historical contingencies and of the adaptive solutions influencing those variable paths that preclude global, theoretical unification. No epistemological tools seem to be available to overcome the original dichotomy between pluralism of explanatory models and unification of our account of cancer under a more essential definition or theory. Either because different perspectives do not seem to come to unification or because they do not seem to account, in a conclusive way, for the intrinsic tension that the part-whole relationships of biological systems presents, the abovementioned dichotomies do not seem to come to an easy solution. Nevertheless, as we have seen in Chapter , systemic approaches constitute an interesting attempt to deal with such tension. They have been developed to describe parts-whole relationships in terms of causal . These examples are discussed in different scientific papers (Sonnesnschein and Soto ; Barker et al ) but have also, to some extent, been discussed in philosophical papers. The issue is how we can understand, that for example, in knockout experiments whereby a gene considered to be essential is inactivated or removed, the neoplastic phenotype is maintained. According to Morange, such unexpected findings are in part due to the fact that gene products are components of pathways and networks (Morange, ) and I find his discussion really satisfactory within some specific cases of cancers.
How science works
relationships among the parts that are, in any case, functional in nature. As I have shown in other studies (Bertolaso b, a, c), in this way the complexity of causes and factors has become a question about the functional definition of the terms that integrate the explanatory account. The epistemological problems related with biological complexity end up to be an issue of the epistemological status of functional explanation in biology. However, and more closely to our current interest, what kind of causal relationships do matter, and how different kinds of causes act at different level of the hierarchical organization is still not clear. This point is evidently directly related with the problem of the choice of the explanatory level and of its definition. Moreover, from a philosophical point of view, another challenge posed by inter-level regulatory processes is that there are, apparently, no necessary relationships among parts and functions to define specific biological behaviours. The discussion about the real existence of natural laws to account for the behaviour of biological systems elaborates this point. On the other hand, the high heterogeneity of parts, functions, levels of organizations and their historical evolution seems to justify the multiplicity of levels of explanation. The failure of biological generalizations to conform to strict lawfulness has been then blamed on the contingency of evolved complex structures. Even if those structures are causally explained, but they are contingent as well, thus universality without exceptions and necessity seems to be out of reach for biology. This led to “the evolutionary contingency thesis”: “To say that biological generalizations are evolutionary contingent is to say that they are not laws of nature –they do not express any natural necessity; they may be true, but nothing in nature necessitates their truths” (Beatty , ). Contingency of biological systems seems to push epistemology of biology against the wall: some kind of reductionist explanation must still be possible or a scientific explanation shall not exist for biology that might be restricted to a descriptive science. However, as Mitchell observes (), the dichotomy between natural contingency and natural necessity traditionally comes from framing natural relations in logical terms. In logic either all statements are necessarily true, in virtue of their form or meaning of their terms, or they are contingent, in that their truth depends on facts. Modelling natural necessity on logical necessity carries with it the presumption that the latter, like the former, is an all-or-none property. A logical
. Pluralism out of unification
statement is either necessary or contingent. Nomologically, a relation between two events in the world, in the traditional sense, is taken to be either necessary or contingent, i.e. accidental. This dichotomy neglects the question whether contingency completely overlaps accidentality in the life sciences; at this stage, however, we are rather concerned with how this issue has been addressed. Mainly, it has been argued that the requirements for absolute lawfulness fail to reflect the reality of scientific practice (see Chapter , Mitchell , ch. ). Mitchell highlights that all laws are contingent in two senses: a) logical contingency, i.e. causal relations depend on some features peculiar to our world; b) the conditions upon which causal structures depend are not equally distributed in space and time. Nevertheless, she notices that, although all laws are logically contingent, there is still a difference between physical laws and biological laws, but it is not the difference between a claim that could not have been otherwise, a ‘law’ in the traditional sense, and a contingent claim, i.e. a ‘non law’. What is required to represent the difference between these laws is a framework which locates different degrees of stability of the conditions upon which the different laws rest, varying with respect to their stability in either time and space. Thus, she remarks that the differences between generalizations in physics and those in biology or the social sciences are inadequately captured by the dichotomy between necessity and contingency. The truth of laws (physical or biological) does not depend on the logical form or definition, but on whether they accurately represent our world. There are differences, but they are differences in degree and origin. The details of the differences demarcate an important domain of study for philosophy of science: “As a consequence, the traditional understanding of laws is incomplete and fails to account for how humans have knowledge of the complexity of our world” (Mitchell , ). .. Biological contingency and context dependency It is thus worthwhile to explore what conditions contingency depends upon. The nature of those conditions, actually, should determine the structure and components of scientific explanations in biology and thus constitute an opportunity to reformulate the question about the
How science works
real existence of laws in biology, then meeting, at least to some extent, the second point. The interesting issue for biological knowledge thus becomes “not charting how results can be made more and more like knowledge of fundamental physics, but how to characterize the type of contingent, complex, causal dependency found in that domain” (Mitchell , -). Different types of contingency have been already described by Sandra D. Mitchell (Mitchell ). She lists them and offers a first discussion of how they are related from an epistemological and ontological point of view: ) ) ) ) ) ) )
logic contingency strong evolutionary contingency weak evolutionary contingency space–time contingency multi–level contingency multi–component contingency redundancy and phase change
Given the analysis of how levels of explanation are identified, of how the relata of the explanatory accounts are shaped by the explananda (Chapter and ), and of how explanatory models of cancer deal with tumour heterogeneity (and thus the stochastic dimension of this process: cfr. Bertolaso , ch. ), my feeling is that not all of the terms listed above encompass the specificity of biological contingency. In brief, and as part of this programmatic reflection on how science works in practice, I believe we can assume that “evolutionary contingency” ( and ) can be understood in terms of multiplicity over time of the type and of contingency. This assumption can be done temporally, from a scientific point of view, relying on the statement that oncogenesis recapitulates phylogenesis (Chapter and ). From an epistemological point of view, it relies on the consideration that the issue beyond contingency is much more related to problems posed by the ‘multiplicity’ of functionalities among parts, i.e. biological heterogeneity, than by the multiplicity of ‘components’, and by the ‘level’ issue. As we have seen, in fact, in Chapter causal relationships can easily be identified with specific parts when the question has been correctly framed. The superiority of the CM proposed by Schaffner
. Pluralism out of unification
over the mechanistic account relies, in my opinion, precisely on this consideration. The question about the parts is secondary respect to the question about the explanatory system. Moreover, redundancy and phase change in biology is more related to stable states than contingency itself, so that it will be necessary to go back to redundancy and phase change when addressing the relationship between stability and contingency . Finally, state-time contingency, in the sense it is introduced — “The conditions upon which causal structures depend are not equally well distributed in space and time” (see Waters ) — from a logical point of view could fit n. and biologically n. and (so that also and ). Their specific interests seems, instead, more related with the notion of law than of contingency, so that we can at the moment circumscribe it too and focus on n. and . For sake of completeness, let’s repeat that issues related with logical contingency have already been cashed out by Mitchell’s arguments on physical and biological laws (cfr. Section refsec:- and Mitchell ). Thus, these kinds of contingencies (-) are not equally relevant to get a deeper understanding of biological change, understood in behavioural or processional terms. As cancer research shows, biological systems are describable in terms of multiple levels of organization that interact in a variety of ways with each other. Additionally, such systems tend to be subject to multiple, interacting causes rather than a single, dominant cause at each of those levels of organization. “In short, they are multi-component, multi-level interacting systems.” (Mitchell , cap ). Contingency and then meet features of complex organizational structures of systems. Cancer is commonly recognized as a complex phenomenon of a hierarchical organized system. State descriptions and process descriptions are equally present in explanatory arguments, while functional accounts define molecular parts and organizational levels that are involved in the neoplastic phenomenon. Organizational context, however, conditions the operation of causal mechanisms. Therefore, the analysis of the conditions on which contingency depends should be somehow related with the context dependency of . This point is definitely beyond the possibility of this volume. For further arguments in favor of this thesis cfr. Bertolaso and c.
How science works
biological properties or behaviours. If contingency has some kind of natural characterization, as I hypothesize (i.e. we are not confronted with just theoretical possibilities or logical relationships), its epistemological status might be better captured looking at those cases in which contingency and context have an explanatory role. The abovementioned hierarchical explanatory models of cancer offer an example of these explanatory issues. The original question about the partswhole relationship can be thus further spelled out. Actually we have to analyse a) the kinds of heterogeneity described in cancer; b) the structure of functional explanations in biology; c) the relation between “stability” and “non-contingency” of association in causal explanations. More extensive analysis of these points have been presented in other papers (Bertolaso a, b, c, d). However, I believe that the approach I have adopted in this volume is valuable to say something about how a), b) and c) are linked in the choice of the explanatory level. Emphasis is on the context argument and its relationship with the concept of mesoscopic level (Chapter ), and on its epistemological relevance in scientific practice. The chapters of this book offer a clarification of the relationship among the terms of explanatory reductions and of how the context should be understood in scientific enterprise . If this approach is able to clarify the abovementioned issues, it is also an indirect proof that a different approach to biological contingency, than a mere formal one, should better fit its epistemological specificity in biological sciences. Therefore understanding contingency from the context perspective might be fruitful. Contingency is just apparently secondary to causal relationships among different and the same levels of hierarchical control in biological systems. Like the context, it plays a relevant explanatory role. Some difficulties and limits in accepting contingent generalizations as explanatory tools in biology might thus disappear . As a referee of a previous version of Chapter and said, such approach could be exported to other files of Philosophy of Biology. Actually, as another referee pointed out with regards to the contents and arguments of the chapter on ‘what the context matters’, the discussion about the context-dependence of fitness and the discussion of the context-dependence of function might be explored fruitfully. To test the type-token context dependency on these other fields (e.g. function and fitness) would probably strengthen the dialectic of the argument, as the type/token context-dependence distinction may go a long way if applied to these debates.
. Pluralism out of unification
if we frame the question not focusing on the relationships among biological facts (parts or events), so that the effort is to explain how stability derives from contingency and is maintained by it, but focusing on the relationship itself that produces contingency as a positive property. This also means that causality belongs to the relationship as well and does not create it. Causal explanations then account for evidences of that relationship whose invariant dimension also grounds the validity of scientific explanation. I believe, nevertheless, that there is a difference in degree of contingency that complex hierarchical organized systems display because they differ in the stability of the conditions upon which they depend, i.e. in the stability of boundaries that structure their functional structures. For example, genes, proteins, cells and tissues show different degrees of functional and structural stability with regards to the same systemic functions, like cell differentiation. In this way also uncertainty is not a problem in biological explanation, although different explanatory tools are required. Some of them are causal, others are descriptive. Biological uncertainty mainly reflects the indetermination or contingency of functional states that characterizes biological systems. Functional explanations, and the derived theoretical terms often used in biological explanations (e.g. functional landscape, fields, etc.), are eventually the best example of how these different kinds of explanatory tools are brought into the same explanatory picture in the process of scientific investigation. The concept of contingency, thus, implies the notion of indetermination and stability at the same time. From an ontological point of view, it is related not with the notion of ‘accidental’ but of ‘potential’. This is reflected also in the use that scientists make of concepts like ‘commitment’ or ‘epigenesis’ to explain morphogenesis, developmental processes or the role that stochasticity plays in them (cfr. as an example of such perspectives Mikkers and Frisen ; Waddington ; Gilbert ). The stability of the conditions upon which a causal relationship depends establishes a continuum, rather than a dichotomously partitioned space of the necessary and the contingent. That is why just saying that a dependency is contingent doesn’t tell you much at all, and certainly doesn’t characterize what is distinctive about the causal structures in complex systems studied by specialized sciences. Moreover, what biologists primarily often investigate is not what is biologically
How science works
necessary, given the constraints upon biological form and behaviour from the actual physical and chemical relationship available. Rather they are concerned with explaining what is contingently present within the wider domain of the biological possible (Dawkins ; Dennett ). However, from a methodological point of view as cancer research shows, it is clear that stability (i.e. what is biologically necessary) has a priority to get a specific causal explanation. The level of inquiry is thus equally important and its choice plays a role in the adequateness of the explanatory argument. This is why contingency of scientific explanations is related with contextuality, and Mitchell can say that she believes that “the problem with evolved, complex systems is that the operation of a single causal factor is contingent on a host of different conditions” (Mitchell , ) and that — considering the epistemological role of context in scientific explanations (i.e. in terms of contributionary premises) — her “approach changes the areas of focus for discussions of how laws are identified, how they are related to evidence, and how they are used to predict and act” (Mitchell , ). .. Epistemological consistency of integrative pluralism in biological sciences From a methodological point of view, unification takes place by means of the integration of different theories and models that address partial causes that that contribute to the generation of biological phenomena (Mitchell , ). However, understanding stability (or robustness) of the biological phenomena is the final way out from the tension that the plurality of causes involved in the biological processes pose. As said in Chapter and supported in Chapter , acknowledging the epistemological priority of stability over specificity of biological systems in the scientific practice and explanatory enterprise will get us further interesting insights about the relationship between the epistemological and ontological sides of the debate on reductionism in biological sciences. Multilevel and multi component contingency can be therefore considered a common feature of complex organizational structures of systems. Cancer is paradigmatic from this point of view, as it is commonly recognized as the disruption of a hierarchical organized system. Can-
. Pluralism out of unification
cer appears as a multi-level phenomenon, because it is a disease of the hierarchical control of multicellular organisms, and of their multi-scale structure. Higher-level explanatory patterns are not usefully reducible to underlying lower-level mechanisms. That is why, “Understanding the complexity of cancer is not only a matter of alternative theories, but requires first and foremost, a methodological approach that reflects a new epistemological view, in light of the fact that cancer is increasingly to be held in systems biology disease” (Rew ). State descriptions and process descriptions are equally present in explanatory arguments about cancer, and functional accounts define molecular parts and organizational levels that are involved in the neoplastic phenomenon. Both state and functional descriptions entail a systemic dimension, which is related with the integrated functional organization of the system, and thus with the coherence of its behaviour. Addressing the question of levels and of their hierarchical organization meets therefore important issues that ground some aspects of the reductionist debate, mainly the challenge posed by connectablity, i.e. the possibility to derive the vocabulary of a science that has to reduce to another science the concepts and terms of the latter (chapter and ). Emergent properties, which I understand as higher–level properties that imply inter-level regulation, force us to consider the difficulty of connectability in reducing them to mechanistic explanations (see Chapter ), so that there are epistemological and ontological questions that have to be answered. It is, in fact, the multilevel phenomenology of biological systems, and the hierarchical control features of their behaviours that justifies focused analysis of causal structure at each level. Integration here does not imply a relativistic position; on the contrary, it is grounded in the intrinsic functioning unity of complex biological phenomena and in our unified perception of it although at different levels and through different organizational features. To put it at its simplest, I suggest that the epistemological need for an integrative approach may be attributed to the following fact. Concepts that would have to be assumed for the derivation of the macro theory cannot be identified with those that are the subject matter of the descriptive theories at the next lower level, although their relationship to the actual level may still be close enough to allow such a derivation to function as an explanation. As already said, the possibility of this non-identity is then to be explained by the fact that
How science works
concepts at different levels are abstractions depending on different points of view. Both models at the macro-scale and descriptive laws at the micro-scale are, in their physical or molecular terms, abstract. But the abstractions involved are not the same in epistemological and ontological terms. In physics it has been shown that theory has had the most success, both in explanatory and predictive terms, as it made use of different levels of abstraction, from examining objects to microscopic particles, to force fields and ultimately to probabilistic distribution functions. This seems to be the case also in biology, where however the success of a similar approach, it is increasingly linked to theoretical constructs based on ontological features of the phenotypes that are applied de facto by scientists when they use concepts such as field, plasticity, robustness, etc. (Aranda–Anzaldo ). The development of interpretative models of cancer, specifically, has shown how the choice of the correct epistemic categories for the study of a phenomenon has important implications in the understanding of the phenomenon itself as well as for the configuration of entire research programs. The fact that, in the transition between different levels or orders of natural phenomena, fundamental contradictions between them do not emerge, is not to say that there is necessarily an identity, and it does not mean that explanation of all reality should be sought in a single physics equation: it just means that there is compatibility between the models that describe it.
References A-A A. (), “Understanding cancer as a formless phenomenon”. Medical Hypotheses, : -. B S.G., S A.M., S C., C A., P J.D., K B.S. (), Plausibility of stromal initiation of epithelial cancers without a mutation in the epithelium: a computer simulation of morphostats. BioMed Central Cancer, : . B, J. (),. “The Evolutionary Contingency Thesis.” In Wolters G., Lennox J.G. (Eds), Concepts, Theories, and Rationality in the Biological Sciences. University of Pittsburgh Press, Pittsburgh, pp. -.
. Pluralism out of unification
B M. (), The neoplastic process and the problems with the attribution of function. Rivista di Biologia / Biology Forum, : -. ——— (), Il Cancro come questione. Modelli Interpretativi e Presupposti Epistemologici, Franco Angeli, Milano. ——— (a), On the Structure of Biological Explanations: Beyond Functional Ascriptions in Cancer Research, Epistemologia XXXVI, pp. -. ——— (b), Sulla ‘irriducibilità’ della prospettiva sistemica in biologia. In Urbani Ulivi L Strutture di mondo. Il pensiero sistemico come specchio di una realtà complessa. Il Mulino, Bologna, in press. ——— (c), Breaking down levels of biological organization, Theoretical Biology Forum, accepted. ——— (d), La Indeterminación Biológica y las Perspectivas Sistémicas de la Biología Contemporáne - Biological Uncertainty and the Systemic Perspectives of Contemporary Biology. Anuario Filosofico, in press. B M. (), Thought Experiments in the Natural Sciences. An operational and Reflexive-Trnscendental Conception, Königshausen&Neumann, Würzburg, pp. -. D R. (), The selfish gene. Oxford University Press, Oxford. D D. (), Darwin’s dangerous idea. Touchstone, NewYork. D J. (), Methahisical disorder and scientific disunity. In Galison P., Stump D. (Eds), The disunity of Science. Stanford University Press, Standford, pp. -. D J. (), The disunity of science, Mind, : -. G S.F. (), Mechanisms for the environmental regulation of gene expression: Ecological aspects of animal development, Journal of Biosciences, : -. M E. (), cause and effects in biology. Science, : -. M H., F J. (), Deconstructing stemness. Embo Journal, : -. M S.D. (), Contingent generalizations: lessons from biology. In Mayntz R (ed), Akteure, Mechanismen, Modelle: Zur Theoriefahigkeit makro-sozialer Analysen, Campus Verlag, pp. -. ——— (), Biological Complexity and integrative pluralism, Cambridge University Press, Cambridge. ———, D M.R. (), Integration without unification: an argu-
How science works
ment for pluralism in the biological sciences, The American Naturalist, (suppl ),: S-. ——— (), Unsimple truths. Science, complexity and policy. University of Chicago Press, Chicago. R D.A. (), Modelling in surgical oncology. Part III: massive data sets and complex systems. European Journal of Surgical Oncology, : -. S P.W. (), The clitoris debates and the levels of analysis. Animal Behaviour, :-. ——— (), The levels of analysis. Animal Behaviour, :-. S C., S A.M. (), Somatic mutation theory of carcinogenesis: why it should be dropped and replaced. Molecular Carcinogenesis, : -. T N. (), On the aims and methods of ethology. Zeitschrift Für Tierpsychologie (Journal of Comparative Ethology), : -. W C.K. (), Causal Regularities in the biological World of Contingent Distributions. Biology and Philosophy, : -. W C.H. (), Tools for thought. Jonathan Cape Thirty Bedford Square, London.
Chapter VII
How does science work? “It is a common idea that some choices of level of explanation or causal description are more appropriate of perspicuous than others, although there is little consensus about what exactly this means”. J. W
.. Choosing the explanatory level Following the quote by Woodward that opened this book, I have developed some arguments to clarify some aspects of how science works practically and to open some issues worthy of further discussion and reflection, like biological contingency and the value of generalization in biological explanations. Adequateness of the choice of an explanatory level is a rational and pragmatic issue. It depends on the interplay of regular patterns we find in nature and our way of conceptualizing them to identify a causal relationship. The definition of such patterns is mediated by technology, but their explanatory power is related with the actual existence of stable functional states in nature that we are able to grasp through the scientific method, i.e. through reductions (Chapter ). In biological sciences we deal with structured entities. However, the dynamic features of such structures are crucial for the understanding of their behaviour, so that hierarchical laws have, at least, an epistemological priority with respect to the laws that govern the behaviours or the properties of the parts (cfr. also Agazzi and on this point ). I have tried to apply this view in my reading of cancer research. The short list of some aspects of the history of cancer research I . Agazzi E, Systems Theory and the Problem of Reductionism, Erkenntnis – –; Agazzi E, “Reductionism as Negation of the Scientific Spirit”, The Problem of Reductionism in Science a cura di Agazzi Evandro — Kluwer Academic Publ, .
How science works
set in Chapter shows that one of the features of the complexity of cancer is tumour cells’ heterogeneity, and reference to a hierarchical organization has been often presented in literature to account for tumour heterogeneity and its complexity. Hence in Chapter I analysed to what extent the neoplastic and metastatic phenomena meet the ND feature of hierarchical organized systems proposed by Simon and what implications this has for a hierarchical account of complex biological systems. The focus is on the organizational properties of cancer and on the control, through differentiation, of biological hierarchically organized systems. As Pattee already noted in : “the relation between the structural and descriptive levels is the central problem that must be solved to have a theory of hierarchical control”. (Pattee , ). In my opinion this point made by Pattee is still an open philosophical issue and a central point in our understanding of biological problems. In Chapter and I partially address this challenge looking at how scientific explanations have been developed to account for regulatory processes on which behavioural dynamics at different scales depend over time. In Chapter , I first addressed the question from the point of view of how reductions work in biological sciences when the question is on the inter–level regulatory processes, and thereafter on the hierarchical control at work in the maintenance of a biological system. I have tried to make the requirements for reduction explicit and to show how they entail an intrinsic non–reductionist relationship among the relata of the explanatory argument. I then give examples of how this conceptual shift occurs in practice, and in which sense a reductionist–mechanicistic approach is unable to capture the explanatory relevance of higher–level features in cancer biology. It emerges that the epistemological issue posed by Pattee can not be considered just a matter of philosophy of science, but is very much related with the practice of science as well. At this point I consider another aspect of the structure of biological explanation, i.e. the strict relationship between the structure of the biological explanations and their context dependency. The strategy I have adopted here is to analyse to what extent the context matters in discussions about biological explanations and biological behaviours, and how it is eventually outlined in the above–mentioned scientific literature on cancer. It follows that the relevance of the context is
. How does science work?
more related to how science works in practice rather than to formal issues of the explanations, and the importance of a distinction between Type and Token Context–Dependency in biological explanations. The former is linked to the level of generalization or conceptualization of the explanans, and the latter to the pragmatic focus on different contrast classes. The intrinsic relationship between levels of stability and contingency at different levels of biological organization opens an interesting issue. The specific complexity — in the sense of the diversity of the contingent — of natural phenomena has to be taken into account. As Mitchell notes, in biology, exceptions depend on context, but we are not dealing with accidental truths. Moreover, in scientific practice we set the experimental control precisely upon this assumption. Biological laws are mainly presented in terms of generalizations. However, “[i]t is not sufficient to say that generalizations in the special sciences are contingent and hence not lawful. Rather one must detail what kind of conditions they depend on and how that dependency works” (Mitchell , ). Biological control, in fact cannot be merely considered the result of feedback mechanisms, as the relevance of the context argument, discussed in Chapter , also suggests. Therefore, in Chapter , I looked at how some of these methodological and epistemological issues have been addressed by the Integrative Pluralism proposed by Mitchell. My interest, in this chapter, has been also to spell out the terms of the relationship between the epistemology of this approach and and some biological features of cancer. I considered how the Integrative Pluralism proposed by Sandra D. Mitchell seems to be a good framework to start from when dealing with the epistemological problems posed by cancer’s complexity. And I framed the question about the possibility to go deeper in the epistemological and ontological foundations of an integrative pluralism in biological sciences. Pluralism of explanatory models seems required in the attempt to understand the intrinsic dynamic unity of the organism we grasp at different levels of its biological organization. In this process, we abstract some features. Conceptualizations of such functional and organizational features are thus partial, and because of that, theoretical. They depend on different points of view, and hence they are useful instruments for scientific knowledge. Conciliation of different perspectives seems to be possible
How science works
precisely because the point of view on the the system (or whole) is not assumed in a rigidlydogmatic way. It is, instead, adopted as a regulatory ideal that leaves open the possibility to reframe the scientific question any time it is necessary, asking for an integration of new variables not taken into account at a different stage of the scientific inquiry . I believe that the PCMS porposed by Schaffner has yet some important contributions to offer and worthwhile to develop further, while taking into account discussions on CA introduced in Chapter . Thus, which level (or levels) is (are) most appropriate will be largely an empirical, pragmatic matter, rather than an a priori one. Moreover, as relationships between factors at the various levels are not independent of each other, different analysis’ can be integrated to understand better the role they play in generating a specific behaviour. “Unification takes place by means of the integration of different theories and models that address partial causes that contribute to the generation of biological phenomena” (Mitchell ). But partial causes are related to specific contexts and their explanatory role to the choice of the level at which they have been identified. That is why Woodward () also says that depending on the level of choice, “causal description/explanation can be either inappropriately broad or general, including irrelevant detail, or overly narrow, failing to include relevant detail” (p. –).
.. Scientific Practice In commenting on the relationship between Philosophy of Biology and Philosophy of Science, Marjorie Grene and David Depew wrote: “Knowledge is a form of orientation, finding one’s way in an environment. For men and women of science, that means orientation in a discipline, a language, a type of laboratory, a style of experimentation, and so on. There is no single, all–inclusive formula for such activities. It is a question of immersing oneself in the detailed history of some particular scientific enterprise, and, it is to be hoped, gaining philo. Buzzoni M., Olismo e causalità in medicina, Anthropos & Iatria — — : –.
. How does science work?
sophical insight from that study” . I am sympathetic with this view of human knowledge and I would like to support it in the future, especially thinking of the kind of knowledge we derive from science and the peculiar relationship between experience and scientific theory. The view of science that emerges from this account of knowledge points to the relevance of the concept of tradition in scientific practice . Woodward’s consideration that “some choices of level of explanation or causal description are more appropriate of perspicuous than others” has been proven fruitful for a first attempt to test the implication of Grene and Depew’s account of knowledge in scientific practice. It has also been useful to clarify, in some specific examples from cancer research and cancer biology, what ‘appropriate’ means and to what extent it is related with the ‘choice’ or the ‘level’ of explanation. Besides the abovementioned challenge to clarify the role of generalization and theoretical concepts in biology (cfr. Section .), I believe that another central challenge here is still the question posed by Mitchell regarding the locality of biological explanation (cfr. Chapter ) and its epistemological implication . The experimental design is dependent on the scientific question. When the question regards inter–level regulatory features of biological behaviour, both contingency and stability have to be taken into account. Emergence and complexity are no longer in the spotlight. Instead, what I suggest is to develop a new epistemology able to cope with tensions and dichotomies generated by the concomitance of contingent and stable features in what we usually call ‘biological change’, i.e. persistence over time of functional properties through (and not merely despite) the . Grene M., Depew D. (), The Philosphy of Biology. An episodic history, Cambridge University Press, p. . . As recently reviewed by Marcos, such concept in Gadamer connects to a compelling idea of practice as a form of rationality that seems to fit very well with what we have seen in science development and progress, especially in life sciences. Marcos M. () “Rescher and Gadamer: Two Complementary Views of the Limits of Sciences”, ponencia leída en las XVIII Jornadas de Filosofía y Metodología Actual de la ciencia, Universidad de La Coruña, de marzo de , Ferrol. Its is going to be published in English as well. . On this point cfr. also Schaffner , (quoted in this volume) and a recent study based on scientific data and instrumental problems I have prepared with some collegues cooperation in “Advances in Systems Biology”, Section: “Essentiality by locality” (Bertolaso M., Giuliani A., Filippi S. ()). The Mesoscopic Level and its Epistemological Relevance in Systems Biology. In: Recent Advances in Systems Biology. Hauppauge NY –:Nova Science Publishers, Inc.
How science works
change of the system’s’ parts. Such epistemology is required also to make justice of why, in experimental practice, stability has a priority over specificity. That is, why, for example, the identification of the mesoscopic level comes before the identification of its explanatory parts in molecular terms and why generalizations play an important role in the definition of the explanatory terms (Chapter ). The ability to grasp the explanatory relationships in causal terms explains why I used in Chapter the adjective of ‘interpretative’ almost as synonymous of ‘explanatory’. I believe that science — in practice — works through different skills that are not only mediated by strictly formal logical requirements . Various approaches can then be possible, contributing differently to our understanding of the world and of our place in it. Methodological reductions are not meant to tell us how the world is; however, they do tell us how science works. The above–described reductions can also make sense of how different biological disciplines are related, and how biological theories can be understood as a collection of overlapping causal and inter–level models. Therefore, although I agree with Wimsatt () that addressing the issue of levels does not require a first and foremost discussion about hierarchies, I also think that it is a useful framework, from a methodological and heuristic point of view, as it is commonly used among scientists to explain such phenomena. On this point, I have taken over some arguments already partially highlighted by Mitchell as well. Science functions through explanations that dissect the natural world into meaningful parts. Behaviours in the natural world are used to explain overall phenomena of a system, but the identification of those parts depends on the system’s properties. Reductions refer to procedures determining a scientific explanation, but the nature of the terms that integrate scientific explanation depends on the causal relationship that holds for the specific relationships among the terms (scientific question) and the system of reference (experimental background). In this interplay of rational and . There is an extensive international literature on this point:let me just mention two books published in Spain and Italy that I have found particularly inspiring as the authors seem to know very well how contemporary science works in practice: Marcos (Marcos A. Filosofia dell’agire scientifico. Le nuove dimensioni. FASTtrack ), Guliani (Giuliani A., Zbilut J.P., L’ordine della complessità, Jaka Book — ).
. How does science work?
experimental endeavour, we grasp pieces of knowledge that are an orientation towards understanding. The non–reductionist dimension that intrinsically characterizes, in my opinion, biological explanations is thus related to the definition of parts and how we understand the structure of the world that is not mechanistically definable. As shown, there are features of the world, and of the way we know it, that can be mechanistically described. Mechanistic explanations are always possible and meaningful, although partial and sometimes more or less interesting or relevant, depending on the scientific question. This different way of looking at the world allows, in science, pluralism without relativism. The reductionist–mechanistic binomial I have adopted at the beginning of Chapter can be at this point disentangled. In my opinion criticisms of mechanistic accounts in biological sciences are not due to the reductionist background. Such criticisms are, instead, related with their weak epistemology. The plurality and diversity of approaches currently adopted to describe biological processes and behaviours in biological sciences can now encourage us to explore in depth how science works and why it works. Science and Philosophy still have a long way to go hand in hand and their dialogue will be even more fruitful and challenging in the near future.
PHILOSOPHY OF SCIENCE . Angelo M Tra ordine e caos. Metodi e linguaggi tra fisica, matematica e filosofia Prefazione di Silvano Tagliagambe ----, formato × cm, pagine, euro
. Andrei–Claudiu H¸ La bioetica e l’uomo del futuro Introduzione di Massimo Losito ----, formato × cm, pagine, euro
. Daniele C Kurt Gödel. Philosophical Explorations. History and Theory Preface by Arcangelo Rossi ----, formato × cm, pagine, euro
. Alfio B L’intenzionalità incarnata. Verso una teoria fra filosofia e neuroscienze ----, formato × cm, pagine, euro
. Nicola S L’organizzazione nel rapporto tra l’organismo e l’ambiente ----, formato × cm, pagine, euro
. Silvano T Il cielo incarnato. Epistemologia del simbolo di Pavel Florenskij ----, formato × cm, pagine, euro
. Marta B How science works. Choosing levels of explanation in biological sciences Preface by Sandra D. Mitchell ----, formato × cm, pagine, euro
Compilato il June , , ore : con il sistema tipografico LATEX 2ε Finito di stampare nel mese di giugno del dalla «ERMES. Servizi Editoriali Integrati S.r.l.» Ariccia (RM) – via Quarto Negroni, per conto della «Aracne editrice S.r.l.» di Roma
