I work at the intersection of philosophy of science, philosophy of biology, and philosophy of medicine, with a focus on the epistemology of competing theories of cancer. My M.A. thesis evaluates whether the metabolic theory of cancer holds epistemic superiority over somatic mutation theory using causal inference and simplicity as criteria — and examines the philosophical implications of adopting a metabolic standpoint on carcinogenesis. My ongoing research extends this into questions of ontological commitment, explanatory conflict, theory choice, and the clinical consequences of undeclared theoretical assumptions in oncology.
I completed my M.A. at Leibniz Universität Hannover under the supervision of Dr. Uljana Feest and Dr. Thomas Reydon, and have been a regular participant in seminars and research exchanges at the IHPST (Université Paris 1 Panthéon-Sorbonne) since 2024. I participated as an invited research member of the SIRIC EpiCure laboratory group at the Hôpital Gustave Roussy under the direction of Dr. Lucie Laplane. Alongside my research, I work in AI and technology, currently serving as an international project team lead overseeing the hiring and training of teams of dozens of international specialists in model evaluation and refinement across English, German, and French. I attend industry events and conferences, maintaining an active interest in developments across the technology sector.
Affiliations & Engagements
Contact
yarentuerkyilmaz@gmail.comThe main aim of this thesis is to compare and contrast two prevalent theories of cancer — somatic mutation theory and metabolic theory — by means of two well-established epistemic criteria within the philosophy of science: causal inference and simplicity. In doing so, the possibility of epistemic superiority of the metabolic theory of cancer is affirmatively explored. Shortcomings of either theory on these epistemic grounds and more broadly — with reference to Pradeau et al.'s criteria for adequate theories of cancer — are discussed, while the question of compatibility of the two competing views is carefully taken into consideration.
The thesis evaluates whether the metabolic theory, as advanced by Seyfried and in line with recent work on the relationship between metabolism and carcinogenesis, holds explanatory advantages over the somatic mutation view — specifically with respect to the clarity of its causal relations and the parsimony of its theoretical commitments. Pradeau et al. (2023) have identified four recurring weaknesses in prevalent theories of cancer: narrowness in scope, speculativeness or limited empirical applicability, disconnection from medical theories of cancer, and reliance on qualitative rather than quantitative description. The thesis takes up these criteria and examines how each framework fares against them.
The clinical and therapeutic implications of adopting a metabolic standpoint on carcinogenesis are also considered: whether the epistemic evaluation of a theory at the level of causal inference and simplicity bears on its practical and experimental consequences. The two epistemic criteria are treated as distinct but related standards. The thesis observes that while they can be usefully separated for analytical purposes, a strict separation proves difficult to maintain consistently — causal clarity and parsimony tend to bear on one another in the evaluation of both theories.
Six papers currently in preparation for submission to journals in philosophy of science, philosophy of biology, and theoretical medicine. Papers 1–2 address the ontological and evolutionary structure of competing cancer frameworks. Papers 3–4 analyse theory choice and causal conflict using interventionist and mechanistic tools. Papers 5–6 turn to clinical practice, examining how framework preference operates in clinical reasoning and what it produces when it goes undeclared.
Multistage progression of cancer from normal epithelium to invasive/metastatic disease. Each stage represents an accumulation of cellular alterations. Classification adapted from standard TNM staging (Brierley et al., 2017) and Vogelstein & Kinzler's (2004) multistep carcinogenesis model.
Vogelstein, B. & Kinzler, K.W. (2004). Cancer genes and the pathways they control. Nature Medicine, 10, 789–799. · Hanahan, D. & Weinberg, R.A. (2000; 2011; 2022). Hallmarks of Cancer. Cell.
The five ontological types of cancer theory, each reflecting a distinct answer to the question: what is the primary unit of disease? Each type generates a different set of experimental targets and therapeutic strategies.
Türkyilmaz, Y. (2024). Competing Theories of Cancer (M.A. Thesis). · Sonnenschein, C. & Soto, A.M. (2011). The tissue organisation field theory. BioEssays. · Hanahan, D. (2022). Hallmarks of Cancer: New Dimensions. Cancer Discovery.
Ontological commitment: Cancer is fundamentally a disease of the cell — specifically, of the cell's DNA. The primary unit of disease is the individual somatic cell bearing one or more driver mutations.
Causal locus: Intracellular — the nucleus, the genome.
Therapeutic implication: Identify and target specific mutant proteins (precision oncology); eliminate mutant clones.
Experimental practice: Genome sequencing, mutation profiling, CRISPR knockout models, targeted drug development.
Ontological commitment: Cancer is a disease of tissue organisation — a disruption in the default state of cellular proliferation caused by altered extracellular signalling fields. The primary unit of disease is tissue architecture, not the individual cell.
Causal locus: Extracellular — the tissue microenvironment, stroma-epithelium signalling.
Therapeutic implication: Normalise tissue organisation; restore epithelial–stromal relationships rather than killing mutant cells.
Experimental practice: 3D organoid culture, ECM disruption models, transplant-reversal experiments.
SMT and TOFT do not merely describe different aspects of the same phenomenon. They make incompatible claims about the primary locus of causal responsibility for cancer initiation. Under SMT, a mutant cell transplanted into normal tissue should give rise to cancer; under TOFT, normal cells placed in a disrupted tissue microenvironment should become tumourigenic. The empirical evidence from transplantation experiments (Mintz & Illmensee, 1975; Sonnenschein & Soto, 2011) is interpreted differently by each framework — each selectively confirming its own causal ontology. The conflict concerns what would count as a cause of cancer, which is a paradigmatic case of alethic conflict between competing causal ontologies.
Sonnenschein, C. & Soto, A.M. (1999). The Society of Cells. Bios Scientific. · Sonnenschein, C. & Soto, A.M. (2011). The tissue organisation field theory of cancer. BioEssays, 33, 332–340. · Vogelstein, B. et al. (2013). Cancer genome landscapes. Science, 339, 1546–1558. · Mintz, B. & Illmensee, K. (1975). Normal genetically mosaic mice produced from malignant teratocarcinoma cells. PNAS, 72, 3585–3589.
Each theoretical framework in oncology identifies a different primary site of causal origin and a different sequence of events leading to cancer. These are not merely different descriptions of the same process — they locate the disease in different biological entities and attribute causal priority to different variables.
Hanahan, D. & Weinberg, R.A. (2011). Cell, 144, 646–674. · Sonnenschein & Soto (2011). BioEssays. · Seyfried, T.N. (2012). Cancer as a Metabolic Disease. Wiley. · Laplane, L. (2016). Cancer Stem Cells: Philosophy and Therapies. Harvard UP. · Nowell, P.C. (1976). Science, 194, 23–28.
Random mutations → clonal expansion of fitter variants. Classic Darwinian logic.
States, not just genes, are heritable. Cells can revert to stem-like states.
Fitness is relational. Competition and cooperation within the tumour ecosystem drive progression.
Darwinian selection acts on genetically variant clones. Fitness is determined by mutation-driven replication advantages. Inheritance is vertical, via cell division.
Selection operates on cellular states and stemness capacities. Cancer stem cells (CSCs) sit atop a differentiation hierarchy; non-CSCs may revert to stemness.
Tumour progression is governed by ecological dynamics within the tumour microenvironment. Fitness is relational, not intrinsic — determined by niche availability and competition.
Three distinct instantiations of evolutionary theory in oncology. Each answers differently: what evolves? what is inherited? what determines fitness? These are not merely different levels of analysis — they generate incompatible predictions about therapeutic resistance and clonal dynamics.
Nowell, P.C. (1976). The clonal evolution of tumor cell populations. Science, 194, 23–28. · Greaves, M. & Maley, C. (2012). Clonal evolution in cancer. Nature, 481, 306–313. · Laplane, L. (2016). Cancer Stem Cells: Philosophy and Therapies. Harvard UP. · Aktipis, A. et al. (2015). Cancer across the tree of life. Phil Trans R Soc B, 370.
The three evolutionary models in oncology each instantiate the basic Darwinian triad — variation, selection, inheritance — differently. These differences are not matters of emphasis but of what counts as a valid evolutionary unit and mechanism.
Nowell (1976) · Greaves, M. & Maley, C. (2012). Clonal evolution in cancer. Nature, 481, 306–313. · Marusyk, A. et al. (2012). Intra-tumour heterogeneity. Nature Reviews Cancer.
In the clone-selectionist model, heritable variation arises through random mutation — not in response to environmental pressure. Selection acts after variation has arisen. Inheritance is strictly vertical: parent cell → daughter cells. The environment (the tumour microenvironment) acts as a selective filter, not a generator of heritable variation.
Key prediction: Resistance mutations pre-exist treatment; they are selected for, not induced by, the therapeutic environment (Luria-Delbrück logic applied to oncology).
In the plasticity-hierarchical and eco-evolutionary models, cellular states can shift in response to environmental signals — and these state-changes can be heritable across cell generations via epigenetic mechanisms (DNA methylation, histone modification). This constitutes a form of environmentally induced heritable variation: a structural parallel to Lamarckian inheritance.
Key prediction: Resistance can be induced by therapeutic pressure through epigenetic reprogramming, not merely selected from pre-existing variants. This has direct consequences for adaptive therapy strategies.
The coexistence of Darwinian and quasi-Lamarckian inheritance mechanisms in cancer cells is not philosophically trivial. It corresponds to a genuine incompatibility within evolutionary oncology about the causal structure of heritable variation. If resistance mutations are pre-selected (Darwinian), sequencing before therapy predicts outcome. If resistance is epigenetically induced (quasi-Lamarckian), pre-therapeutic sequencing has limited predictive value. The frameworks generate conflicting clinical recommendations.
This is not merely a question about mechanisms but about what kind of interventions — on what variables — would be effective. In Woodwardian terms, the two frameworks disagree about which variables are invariantly connected to therapeutic outcomes across the relevant intervention range.
Flavahan, W.A. et al. (2017). Epigenetic plasticity and the hallmarks of cancer. Science, 357. · Laplane, L. (2016). Cancer Stem Cells: Philosophy and Therapies. Harvard UP. · Gould, S.J. (2002). The structure of evolutionary theory. · Woodward, J. (2003). Making Things Happen. Oxford UP.
Primary claim: Driver mutations in oncogenes and tumour suppressor genes are the initiating and sustaining cause of cancer. Metabolic reprogramming is a downstream consequence of genetic change.
Causal direction: Mutation → altered gene expression → metabolic reprogramming
Therapeutic target: Mutant proteins (e.g., BRAF V600E, KRAS G12C inhibitors)
Dominance factors: Technological infrastructure (NGS, CRISPR), actionability in clinical oncology, alignment with pharmaceutical industry investment, institutional embodiment in TCGA, ICGC
Primary claim: Cancer originates from irreversible damage to mitochondrial oxidative phosphorylation. Somatic mutations are a downstream consequence of metabolic dysfunction, not the primary cause. The cell's compensatory shift to fermentation is the initiating event.
Causal direction: Mitochondrial dysfunction → reactive oxygen species → nuclear genome instability → somatic mutations
Therapeutic target: Metabolic environment — ketogenic diet, hyperbaric oxygen, press-pulse strategy (Seyfried et al., 2017)
Both SMT and MTC accept that cancer cells exhibit somatic mutations and metabolic reprogramming simultaneously. The disagreement concerns which is causally prior. That dispute cannot be resolved by pointing to the association itself — two theories may agree on a co-occurrence while disagreeing fundamentally on its direction of dependence. In Woodwardian terms: intervening on the metabolic state would, under MTC, alter mutation rates; under SMT it would not. Intervening on driver mutations would, under SMT, reverse the metabolic phenotype; under MTC it would not. These are empirically distinct predictions about what interventions would do — which is precisely what makes the conflict genuine rather than verbal.
Vander Heiden, M.G. et al. (2009). Understanding the Warburg Effect. Science, 324, 1029–1033. · Seyfried, T.N. (2012). Cancer as a Metabolic Disease. Wiley. · Warburg, O. (1956). On the origin of cancer cells. Science, 123, 309–314.
Otto Warburg (1956) observed that cancer cells preferentially ferment glucose to lactate even in the presence of oxygen — aerobic glycolysis. This is energetically inefficient: glycolysis yields 2 net ATP per glucose, compared with ~30 ATP via oxidative phosphorylation under physiological conditions (theoretical maximum ~38). Despite this, aerobic glycolysis confers biosynthetic advantages for rapidly dividing cells by generating metabolic intermediates for anabolic pathways. The metabolic theory of cancer (Seyfried, 2012; Pedersen, 1978) treats aerobic glycolysis as the proximal cause of cancer, initiated by mitochondrial damage. SMT treats it as a downstream consequence of oncogene activation — MYC upregulation directly induces glycolytic enzymes; HIF-1α is induced via RAS/PI3K/mTOR signalling independently of hypoxia. The same biological phenomenon is therefore interpreted as cause or effect depending on which theoretical framework is operative, illustrating the theory-ladenness of cancer data.
Warburg, O. (1956). On the origin of cancer cells. Science, 123, 309–314. · Pedersen, P.L. (1978). Tumor mitochondria. Progress in Experimental Tumor Research. · Seyfried, T.N. & Shelton, L.M. (2010). Cancer as a metabolic disease. Nutrition & Metabolism.
| Determinant | How it advantages SMT | Effect on alternatives |
|---|---|---|
| Technological infrastructure | Next-generation sequencing, CRISPR, proteomics — all built for genomic analysis. The Cancer Genome Atlas (TCGA) profiled ~11,000 patient samples across 33 cancer types. Tools embody SMT's ontology. | Metabolic and tissue-level frameworks lack comparable data infrastructure; entry costs to competition are structurally higher. |
| Actionability (clinical) | A mutation is 'actionable' if a licensed drug targets it. This concept is built around mutant proteins. Precision oncology's clinical language is SMT-entrenched. | Metabolic interventions (dietary, metabolic drugs) do not fit 'actionability' criteria, creating a systematic disadvantage in clinical trial design and regulatory approval. |
| Pharmaceutical alignment | Small-molecule inhibitors targeting specific mutant proteins are highly patentable and lucrative. Industry investment massively favours SMT-derived targets. | Metabolic approaches (e.g., dietary intervention, repurposed generic drugs) have low commercial incentive, regardless of efficacy evidence. |
| Institutional embodiment | TCGA, ICGC, most major cancer centres organise data around mutational profiling. Peer review norms, grant structures, and training programmes are SMT-aligned. | Alternative frameworks face structural barriers to achieving comparable institutional presence — a non-epistemic source of asymmetric persistence. |
| Experimental translatability | Mouse models with defined driver mutations (e.g., KRASG12D, TP53R172H) are well-established and widely used. Results translate predictably within SMT. | Metabolic and tissue-level models require different experimental systems (e.g., altered diet models, ECM disruption); less established and harder to fund. |
Non-epistemic determinants of framework dominance in oncology. Each factor independently and jointly reinforces SMT's institutional position without necessarily tracking relative empirical adequacy.
Kuhn, T.S. (1962). The Structure of Scientific Revolutions. · Longino, H.E. (1990). Science as Social Knowledge. · Douglas, H. (2009). Science, Policy, and the Value-Free Ideal. · Topol, E.J. (2012). Individualized medicine from prewomb to tomb. Cell.
SMT: Do mutations → metabolic changes? Yes, necessarily.
MTC: Do metabolic changes → mutations? Yes, necessarily.
Both cannot be the primary initiating cause.
SMT: Intervening on mutations will change cancer phenotype.
MTC: Intervening on metabolic state will change mutation rates.
These are different causal variables.
SMT: Primary locus is the nucleus (genome).
MTC: Primary locus is the mitochondrion.
Nuclear transplant experiments test this directly.
Two causal claims C₁ and C₂ are alethically copossible with respect to a system S if and only if there exists a possible world in which both C₁ and C₂ are simultaneously true of S under the same description.
The three conflicts identified above fail the copossibility criterion: under the same system description (a specific cancer cell at a specific time), both the claim that mutations are causally prior and the claim that metabolic dysfunction is causally prior cannot be simultaneously true.
The apparent pluralism — 'both are true at different levels' — cannot discharge the conflict without either (a) changing the system description or (b) changing what counts as a causal claim. Neither move is available to the defender of simple complementarity.
Woodward, J. (2003). Making Things Happen. Oxford UP. · Baumgartner, M. (2010). Shallow analysis and the causation problem. Philosophy of Science. · Kistler, M. (2006). Causation and Laws of Nature. Routledge. · Craver, C.F. (2007). Explaining the Brain. Oxford UP.
Woodward's interventionism evaluates causal claims by asking: would an ideal intervention on X produce a change in Y? But this presupposes that the variables X and Y have already been selected for inclusion in the model. The variable-selection problem asks: what determines which variables enter the causal model in the first place?
Includes: specific driver mutations (KRAS, TP53, BRCA1/2, APC…), signalling pathway activations, copy number variants, chromatin accessibility
Excludes: systemic metabolic state, dietary environment, mitochondrial integrity as primary variables
Includes: mitochondrial membrane potential, ROS levels, ATP/ADP ratio, lactate output, oxygen availability, respiratory capacity
Includes mutations only as downstream markers, not primary causal variables
The two frameworks do not merely weight the same variables differently — they include different variables in their causal models. A Woodwardian evaluation of SMT claims using MTC variable sets, and vice versa, will produce systematically different causal verdicts. This is the diagnostic function of the variable-selection problem: it reveals that the frameworks are not operating on the same causal domain even when describing the same cancer cell.
Woodward, J. (2003). Making Things Happen. Ch. 2–3. · Spirtes, P. et al. (2000). Causation, Prediction, and Search. MIT Press. · Hausman, D. & Woodward, J. (1999). Independence, invariance, and the causal Markov condition. British Journal for the Philosophy of Science.
Craver's (2007) account of mechanistic multilevel explanation holds that higher-level and lower-level descriptions of the same mechanism are mutually constitutive: the upper level is realised by, and only makes causal claims in virtue of, the lower level. The 'nesting assumption' requires that multi-level causal claims refer to nested components of the same mechanism.
If the assumption holds: SMT and MTC would describe nested levels of a single mechanism — mutations would be higher-level descriptions realised by metabolic lower-level processes, or vice versa. The frameworks would be complementary by definition.
This is the standard integrationist response: Hanahan & Weinberg's Hallmarks framework implicitly assumes that all the causal factors it catalogues are nested within a single mechanistic framework of tumour biology.
If the assumption fails: SMT and MTC describe the same cellular events — metabolic reprogramming and somatic mutation — but attribute different causal roles to them within non-nested mechanistic structures. The upper-level SMT claim (mutations cause cancer) is not realised by but rather in competition with the upper-level MTC claim (metabolic dysfunction causes cancer).
This paper argues the nesting assumption fails for the SMT/MTC comparison in the cancer initiation context.
Nuclear/Cytoplasmic Transfer Experiments as Philosophical Test Cases: A set of transplantation experiments — transferring nuclei and cytoplasm between normal and cancerous cells in various combinations — provides a direct empirical test of SMT vs. MTC. Under SMT, transferring a cancerous nucleus into a normal cytoplasm should produce malignancy. Under MTC, transferring normal cytoplasm into a cancer cell should suppress malignancy. The results of such experiments (reviewed in Seyfried, 2012, ch. 11) have been interpreted as partial support for MTC's predictions, but the interpretation remains contested because the experimental results are theory-laden in ways that prevent straightforward adjudication between frameworks.
Craver, C.F. (2007). Explaining the Brain. Oxford UP. · Seyfried, T.N. (2012). Cancer as a Metabolic Disease. Wiley (ch. 11). · Hanahan, D. & Weinberg, R.A. (2011). Hallmarks of Cancer: The Next Generation. Cell, 144, 646–674.
Both SMT and alternative frameworks describe different aspects of cancer. Tools from each are deployed by context. The Hallmarks framework encourages this. But it suppresses genuine incompatibilities when treating a specific patient.
Frameworks are assigned by presumed aetiology: carcinogen-driven lung cancer → SMT; hormone-dependent breast/prostate → microenvironmental models; metabolic syndrome-associated cancers → MTC. The assignment itself is a theoretical claim.
Frameworks are assigned by tumour type: haematological malignancies as primarily genomic; solid tumours as more microenvironmentally complex. Glioblastoma handled differently from NSCLC. Implicit ontological commitment; rarely declared.
One framework is the correct account; alternatives are errors, incomplete descriptions, or empirically disconfirmed. The clinician holds SMT or MTC as true of cancer, treating the other's evidence as explicable within their own framework.
None of these positions is typically declared in clinical reasoning. Framework preference operates as background — structuring which questions are asked and how results are interpreted — while remaining invisible as a methodological variable. A null result in a metabolic intervention trial may reflect therapeutic failure or trial design failure under the wrong framework. These are not the same inference.
The four positions are not merely descriptive categories. Each implies a different set of clinical decisions, a different relationship to negative trial evidence, and a different standard for what counts as therapeutic success or failure.
| Dimension | I — Pluralist | II — Aetiology-Assigned | III — Disease-Assigned | IV — Incommensurabilist |
|---|---|---|---|---|
| How frameworks relate | Complementary descriptions of a complex disease; different tools for different questions | Each framework captures the predominant biology of a specific aetiological subtype | Each framework captures the predominant biology of a specific tumour category | One framework is causally correct; others are incomplete, mistaken, or reducible to it |
| What guides framework choice | Clinical context; current research question; available assays | Presumed aetiology of the patient's tumour (carcinogen, hormone, metabolic) | Tumour type and site (haematological vs. solid; organ of origin) | The clinician's theoretical commitment; evidence interpreted from within that framework |
| How conflict between frameworks is handled | Typically ignored — frameworks seen as operating at different levels or scales | Managed by restricting each framework to its assigned aetiology | Managed by disease-specific specialisation; each subspecialty maintains its own framework | Conflict is an empirical question with a correct answer; competing evidence reinterpreted or explained away |
| How negative trial evidence is interpreted | Failure of the intervention in this context; does not generalise across frameworks | Failure reflects wrong aetiological selection; would succeed in a correctly stratified population | Failure reflects wrong disease-type application; framework remains valid for its assigned tumour category | Failure confirms the competing framework's inadequacy; or attributed to flawed design within the rejected framework |
| Standard for therapeutic success | Defined by whichever framework's endpoint is in use; no cross-framework standard | Framework-specific endpoint appropriate to presumed aetiology | Tumour-type standard of care, which inherits whatever framework built that standard | The endpoint that the correct framework defines; alternatives are category errors |
| Is the position declared? | Rarely — pluralism presented as methodological openness, not theoretical commitment | Rarely — aetiological assignment treated as empirical inference, not framework choice | Never — specialisation by tumour type is institutional, not acknowledged as theory-laden | Occasionally in theoretical contexts; never in trial design or clinical documentation |
Hanahan, D. & Weinberg, R.A. (2011). Hallmarks of Cancer: The Next Generation. Cell, 144, 646–674. · Feinstein, A.R. (1967). Clinical Judgment. Williams & Wilkins. · Longino, H.E. (1990). Science as Social Knowledge. Princeton UP.
Framework preference is not confined to theoretical debate. It is present in clinical guidelines, tumour board composition, pharmaceutical R&D priorities, and patient consultations. In each setting, the theoretical commitment is operative but undisclosed.
| Clinical or Research Context | How Framework Preference Manifests | Declared? | Consequence of Non-Declaration |
|---|---|---|---|
| NCCN / ESMO clinical guidelines | Guidelines organised by tumour type, then by mutation status. Treatment pathways built around targetable mutations (KRAS, EGFR, BRAF, HER2). Metabolic, stromal, and ecological interventions listed as investigational — not primary pathway options. | No. Mutation-centred organisation presented as standard of care, not as a framework-specific recommendation. | Clinicians following guidelines enact SMT commitments without recognising the choice. Departure requires conscious defection from institutional authority — a high friction cost in routine practice. |
| Multidisciplinary tumour boards (MDTs) | Standard MDT: surgical oncologist, medical oncologist, radiation oncologist, pathologist, radiologist. Molecular oncologist increasingly standard. Metabolic specialist, microenvironment researcher, or evolutionary oncologist: absent or exceptional. | No. MDT composition presented as ensuring comprehensive assessment, not as reflecting a theoretical framework about which aspects of the disease are most causally relevant. | The questions that can be raised in an MDT are constrained by who is present. A metabolic hypothesis about a patient's resistance cannot be discussed if no one in the room holds the relevant framework. |
| Clinical trial reporting | CONSORT and SPIRIT standards require disclosure of statistical methods, randomisation, blinding, and outcome measures. No requirement to state the theoretical framework underpinning target selection, patient stratification, or endpoint definition. | No. Framework assumptions embedded in the protocol but never listed as assumptions. | When a metabolic or ecological trial fails its primary endpoint, the null result enters the evidence base without the caveat that it was evaluated using SMT-derived criteria. The literature accumulates misleading negative evidence. |
| Pharmaceutical R&D priority-setting | Drug pipeline organised around actionable mutations. Investment concentrated in KRAS, EGFR, BRAF, ALK, CDK4/6, PI3K inhibitors. Metabolic targets remain in early clinical development. Dietary and press-pulse protocols: no commercial pathway. | No. Priority-setting justified by "unmet need" and "biological rationale" — both criteria that are framework-internal. | The absence of approved metabolic drugs is used as evidence of non-actionability. Non-actionability is used to justify non-investment. The circularity is invisible because the framework is undisclosed. |
| Patient consultation | Patients told their tumour has a specific mutation and offered targeted therapy or chemotherapy. The metabolic or ecological interpretation of their disease is not offered as an alternative account. | No. The SMT framing is presented as the diagnosis, not as one of several theoretically possible descriptions of the patient's disease. | Where alternative frameworks generate different treatment options, non-disclosure is epistemically and ethically significant. Patients cannot choose between accounts of their disease they have not been told exist. |
Prasad, V. & Mailankody, S. (2017). Research and development spending to bring a single cancer drug to market. JAMA Internal Medicine, 177, 1569–1575. · Fojo, T. et al. (2014). Unintended consequences of expensive cancer therapeutics. JAMA Otolaryngology, 140, 1225–1236. · Sackett, D.L. et al. (1996). Evidence based medicine: what it is and what it isn't. BMJ, 312, 71–72.
Longino (1990) identified background assumptions as the tacit theoretical commitments that determine relevance relationships between evidence and hypothesis — what counts as data, what counts as confirmation, what counts as a plausible alternative. Background assumptions are not usually stated because they are shared by all parties to the inquiry; their invisibility is a mark of consensus, not of absence.
In oncology, SMT functions as the dominant background assumption. It determines which molecular changes count as primary evidence, which endpoints count as response measures, and which interventions count as therapeutic candidates. Competing frameworks are not excluded by explicit decision; they simply do not fit the background against which evidence is evaluated.
Foucault's episteme — the historical set of conditions that determine what can count as a legitimate statement in a given domain — does not require conspiracy or false consciousness. A knowledge-power structure reproduces itself precisely because it determines what questions are intelligible, not by forbidding certain answers. The clinician who orders mutational profiling is not suppressing metabolic hypotheses; metabolic hypotheses are simply not present in her clinical episteme.
This is what makes framework preference an ideological structure rather than merely a methodological preference: it operates at the level of what can be seen, not at the level of what is permitted. The corrective is not deregulation but excavation — making the background visible so that it can be examined as what it is.
The argument does not require abandoning SMT or treating all frameworks as equally valid. It requires only that framework position be declared — in trial protocols, in clinical guidelines, in MDT discussions, and in patient consultations where alternative accounts generate different treatment options. A trial testing a metabolic intervention against an SMT-derived endpoint should say so. A guideline organised around mutation status should acknowledge that this organisation embeds a framework that competing accounts of cancer would dispute.
Whether a clinician holds Position I, II, III, or IV should be as reportable as the statistical assumptions underpinning their trial design — not because it will always change what they do, but because undeclared theoretical commitments distort the evidential record in ways that accumulate over time and cannot be corrected retrospectively.
Longino, H.E. (1990). Science as Social Knowledge. Princeton UP. · Foucault, M. (1970). The Order of Things. Pantheon. · Douglas, H. (2009). Science, Policy, and the Value-Free Ideal. Pittsburgh UP. · Kuhn, T.S. (1962). The Structure of Scientific Revolutions. Chicago UP.
| Clinical Domain | SMT / Precision Oncology | Metabolic Theory (MTC) | Tissue / TOFT | CSC Theory | Eco-Evolutionary |
|---|---|---|---|---|---|
| Primary drug target | Mutant oncoprotein or kinase (e.g. BRAF V600E → vemurafenib; KRAS G12C → sotorasib; EGFR → erlotinib) | Glycolytic enzymes (LDHA, HK2), mitochondrial complex I (metformin, IACS-010759), glutamine metabolism | Stromal signalling molecules; ECM components; fibroblast activation; TGF-β pathway normalisation | CSC surface markers (CD44, CD133); stemness transcription factors (SOX2, OCT4, NANOG); Wnt/Notch pathways | No single molecular target. Ecological leverage points: competitive dynamics, resource limitation, spatial structure |
| Treatment goal | Eliminate cells carrying driver mutations; achieve molecular remission defined by absence of detectable mutant clones | Normalise mitochondrial function; restrict glucose availability; starve the Warburg phenotype | Restore normal tissue organisation; reverse epithelial-mesenchymal interface disruption | Eliminate or differentiate CSC subpopulation; prevent tumour regeneration from residual stem-like cells | Manage tumour population dynamics; maintain sensitive cells that suppress resistant clones; delay progression rather than achieve remission |
| Response biomarker | ctDNA mutation frequency; allele frequency of driver mutations; molecular response rate | PET-FDG uptake (glycolytic rate); lactate levels; mitochondrial membrane potential | Tissue architecture scores; stromal marker normalisation (FAP, α-SMA); ECM stiffness | CSC frequency in tumour biopsy; ALDH activity; sphere-forming assay | Clonal diversity index; tumour volume dynamics; frequency of sensitive vs. resistant populations |
| Resistance interpretation | Emergence of secondary mutation (e.g. T790M, C797S); bypass pathway activation via alternative oncogene | Upregulation of alternative metabolic pathways; shift between glycolysis and OxPhos states | Stromal remodelling enabling bypass of normalising signals; fibrotic progression of TME | De-differentiation of non-CSC cells to CSC phenotype under therapeutic pressure; epigenetic plasticity | Competitive release of resistant clones; elimination of sensitive cells removes ecological suppression of resistant variants |
| Dietary / lifestyle | Not a primary therapeutic target under standard SMT-guided care; diet not expected to alter mutational landscape | Ketogenic diet as adjunct therapy (restricts glucose substrate); intermittent fasting; hyperbaric oxygen (press-pulse protocol, Seyfried et al.) | Dietary and endocrine factors relevant insofar as they alter systemic hormonal signals reaching the tissue field | Not a primary modality; relevant only if dietary factors alter CSC niche maintenance | Relevant as modifier of ecological parameters (nutrient availability, immune composition of TME) |
What a clinician does in each clinical domain depends on which framework they are working within — whether consciously or not. The table below shows how each framework answers the same clinical questions differently, making explicit commitments that routine practice leaves unstated.
Vander Heiden, M.G. et al. (2009). Understanding the Warburg Effect. Science, 324. · Gatenby, R.A. et al. (2009). Adaptive therapy. Cancer Research, 69, 4894–4903. · Seyfried, T.N. et al. (2017). Press-pulse: a novel therapeutic strategy. Nutrition & Metabolism, 14:19. · Laplane, L. (2016). Cancer Stem Cells: Philosophy and Therapies. Harvard UP. · Sonnenschein & Soto (2011). BioEssays.
Resistance management is where framework preference has the most direct clinical stakes. The question is not merely which drug to use next, but what kind of biological event resistance represents — and this question receives different answers depending on which evolutionary model underpins the clinician's reasoning. Under clone-selection logic (SMT), resistance is a selection event and was predictable in principle by pre-treatment sequencing. Under eco-evolutionary logic, it is an ecological consequence of the treatment strategy itself, and maximum-dose therapy may be its primary cause. Under the plasticity model, it is partly an epigenetically induced phenotypic shift, not a pre-existing variant. These are not complementary accounts of resistance; they assign responsibility to different variables and prescribe different responses.
Mechanism: Resistant mutations pre-exist treatment at low clonal frequency. Drug-induced selective pressure eliminates sensitive clones, allowing pre-existing resistant variants to expand — classic Darwinian selection (Luria-Delbrück logic applied to oncology).
Clinical prediction: Resistance is detectable before treatment by deep sequencing. Acquired resistance mutations (e.g. EGFR T790M after erlotinib; BRAF V600E after vemurafenib) are selection events, not induction events.
Therapeutic response: Combine agents targeting different vulnerabilities simultaneously (vertical and horizontal pathway blockade). Next-generation inhibitors for secondary mutations (osimertinib for T790M). Liquid biopsy monitoring for emergent resistance clones.
Mechanism: Maximum-dose therapy eliminates drug-sensitive cells, creating an ecological vacuum. Resistant cells — previously suppressed by competition with sensitive cells — expand into vacated niche space. This is competitive release, not mutation selection alone (Gatenby et al., 2009).
Clinical prediction: High-dose treatment paradoxically accelerates resistance by eliminating the sensitive population that was ecologically suppressing resistant clones. Standard response criteria (tumour volume reduction) are inadequate; clonal diversity must be tracked.
Therapeutic response: Adaptive therapy — modulate dose dynamically to maintain a population of sensitive cells, preserving ecological suppression of resistant variants. Exploit evolutionary trade-offs: resistance to one drug may confer sensitivity to another (collateral sensitivity).
CSCs are intrinsically resistant to cytotoxic therapy: typically quiescent, expressing multidrug-resistance pumps, with elevated DNA repair capacity. Under therapeutic stress, non-CSC cells can undergo epigenetic de-differentiation back to a stem-like state — resistance is partly induced, not just selected. This is a quasi-Lamarckian dynamic absent from SMT's pre-selectionist account of resistance. Treating the bulk tumour leaves the CSC subpopulation intact; the tumour regenerates from residual stem-like cells. CSC-directed agents or differentiation-inducing therapies are required in combination — a different treatment logic entirely.
Under metabolic theory, resistance reflects the metabolic flexibility of cancer cells: the capacity to shift between glycolytic and oxidative phenotypes when one pathway is pharmacologically suppressed. Targeting glucose restriction alone may be countered by upregulation of glutamine as an alternative carbon source. The press-pulse protocol (Seyfried et al., 2017) addresses this directly: a chronic metabolic constraint (ketogenic diet, glucose restriction) combined with acute drug pulses targeting remaining metabolic pathways exploits the reduced buffering capacity of metabolically stressed cells. A clinician reasoning within SMT has no basis for prescribing this combination — the intervention is invisible as a category within that framework's account of what the disease is.
Gatenby, R.A. et al. (2009). Adaptive therapy. Cancer Research, 69, 4894–4903. · Aktipis, C.A. et al. (2012). Life history trade-offs in cancer evolution. Nature Reviews Cancer, 13, 342–350. · Gupta, P.B. et al. (2011). Stochastic state transitions give rise to phenotypic equilibrium in populations of cancer cells. Cell, 146, 633–644.
Trial design is not framework-neutral. When a metabolic intervention fails to meet a RECIST endpoint, this is standardly read as evidence against the metabolic hypothesis. But RECIST measures tumour volume on imaging — a criterion derived from SMT's model of disease, in which tumour mass corresponds to cell-population burden of mutant clones. Under MTC, the primary endpoint would be metabolic normalisation, not volume reduction. A negative RECIST result in a metabolic trial does not disconfirm MTC; it may instead record that a metabolic intervention was evaluated with the wrong instrument. Framework preference, when undeclared, converts a methodological choice into an apparently empirical verdict.
| Trial Design Element | SMT-Derived Standard (dominant) | What Non-SMT Frameworks Require | What the Mismatch Produces |
|---|---|---|---|
| Patient stratification | Biomarker-selected by mutation status (e.g. KRAS wild-type, HER2-amplified, MSI-H). Patients enrolled based on genomic profile of tumour. | MTC: stratify by baseline metabolic phenotype (FDG uptake, lactate, mtDNA copy number). TOFT: stratify by stromal composition. Eco: stratify by clonal diversity index. | Metabolic and tissue-level therapies trialled in unselected populations that include patients unlikely to respond by that theory's own criteria. Diluted effect sizes lead to false negative trials. |
| Primary endpoint | Overall survival (OS) or progression-free survival (PFS) measured by RECIST criteria (tumour volume on imaging). Molecular complete response (absence of ctDNA). | Eco: clonal composition dynamics; resistant fraction over time. CSC: CSC frequency post-treatment. MTC: metabolic phenotype normalisation. | Adaptive therapy trials that increase clonal diversity (not reduce tumour volume) may fail RECIST endpoints while achieving their actual therapeutic goal. Trial infrastructure cannot detect success. |
| Comparator / standard of care | Best available molecularly targeted therapy or chemotherapy. Control arm reflects SMT-based treatment as background. | Metabolic interventions (ketogenic diet + press-pulse) tested as adjunct to, not replacement of, standard care. Creates confounding — hard to isolate metabolic contribution. | Metabolic and dietary approaches systematically evaluated in the least favourable conditions: as add-ons to high-dose cytotoxic regimens that conflict with the metabolic hypothesis. |
| Resistance definition | Disease progression on imaging. Emergence of resistance mutation (secondary sequencing). Defined as treatment failure. | Eco: progression of sensitive-cell depletion is itself a sign of therapeutic success in adaptive therapy — not failure. Resistance emergence is an expected and managed feature, not endpoint. | Adaptive therapy trials may be prematurely halted or declared failures by regulatory endpoints designed around maximum-dose SMT logic. The regulatory framework embeds SMT assumptions. |
| Regulatory pathway | FDA/EMA accelerated approval based on objective response rate (ORR) for molecularly targeted drugs. Companion diagnostic (CDx) required for biomarker-selected treatments. | Metabolic and dietary interventions do not fit standard CDx framework. Adaptive therapy requires novel regulatory category — dynamic dosing is not evaluable under fixed-dose trial assumptions. | Non-SMT frameworks face structural regulatory barriers not arising from scientific inadequacy but from framework mismatch. Institutional infrastructure embeds SMT's ontological commitments. |
In precision oncology, a target is "actionable" if a licensed drug exists that binds it. The term presents itself as a clinical-epistemic criterion — targeting what is known to be druggable — but it is itself a product of the infrastructure SMT has generated. Metabolic, stromal, and ecological targets are not actionable because the drug-development pipeline was built around mutant proteins. The absence of approved drugs is not evidence that these targets are less causally fundamental; it is evidence that the pathway from hypothesis to approval was never structured to accommodate them.
What is at stake is a Foucauldian dynamic: the theoretical commitment to SMT has been institutionalised to the point where it determines not just what is studied, but what counts as a legitimate question. The clinician who orders mutational profiling is not making a conscious philosophical choice — she is acting within a field in which that choice has been made for her, in advance, by the conditions of possibility of contemporary oncology. Making framework preference explicit — as a declared assumption in trial design and clinical reasoning — is the minimal corrective. Whether a clinician holds position I, II, III, or IV with respect to competing frameworks should be as reportable as the statistical assumptions of the trial they design.
Topol, E.J. (2012). Individualized medicine from prewomb to tomb. Cell, 148, 1207–1214. · Gatenby, R.A. & Brown, J.S. (2020). Integrating evolutionary dynamics into cancer therapy. Nature Reviews Clinical Oncology, 17, 675–686. · Prasad, V. & Mailankody, S. (2017). Research and development spending to bring a single cancer drug to market. JAMA Internal Medicine, 177, 1569–1575.
Target selection, patient stratification, and response biomarker each encode a theoretical commitment. SMT → mutant proteins, mutation status, ctDNA. MTC → glycolytic enzymes, metabolic phenotype, FDG-PET. Different framework, different molecule, different patient.
Clone-selection → sequence for pre-existing resistant variants, combine inhibitors. Eco-evolutionary → preserve sensitive cells, adaptive dosing. CSC → target stemness, prevent de-differentiation. Three evolutionary models, three therapeutic responses.
RECIST, ctDNA, ORR — SMT-derived endpoints. Metabolic normalisation, clonal diversity, CSC frequency — alternative endpoints, rarely used. Null results under mismatched endpoints enter the literature as disconfirmation of the hypothesis, not the design.
The treatment decisions available to a clinician in each domain are shaped — in part — by which framework they are working within, whether consciously or not. The following maps how each framework would answer the same clinical questions, making explicit the theoretical commitments that routine practice typically leaves unstated.
| Clinical Domain | SMT / Precision Oncology | Metabolic Theory (MTC) | Tissue / TOFT | CSC Theory | Eco-Evolutionary |
|---|---|---|---|---|---|
| Primary drug target | Mutant oncoprotein or kinase (BRAF V600E → vemurafenib; KRAS G12C → sotorasib; EGFR → erlotinib) | Glycolytic enzymes (LDHA, HK2), mitochondrial complex I (metformin, IACS-010759), glutamine metabolism | Stromal signalling molecules; ECM components; fibroblast activation; TGF-β pathway normalisation | CSC surface markers (CD44, CD133); stemness transcription factors (SOX2, OCT4, NANOG); Wnt/Notch pathways | No single molecular target. Ecological leverage points: competitive dynamics, resource limitation, spatial structure |
| Treatment goal | Eliminate cells carrying driver mutations; achieve molecular remission defined by absence of detectable mutant clones | Normalise mitochondrial function; restrict glucose availability; suppress the Warburg phenotype | Restore normal tissue organisation; reverse epithelial-mesenchymal interface disruption | Eliminate or differentiate CSC subpopulation; prevent tumour regeneration from residual stem-like cells | Manage tumour population dynamics; maintain sensitive cells that suppress resistant clones; delay progression rather than achieve remission |
| Response biomarker | ctDNA mutation frequency; allele frequency of driver mutations; molecular response rate | PET-FDG uptake (glycolytic rate); lactate levels; mitochondrial membrane potential | Tissue architecture scores; stromal marker normalisation (FAP, α-SMA); ECM stiffness | CSC frequency in tumour biopsy; ALDH activity; sphere-forming assay | Clonal diversity index; tumour volume dynamics; frequency of sensitive vs. resistant populations |
| Resistance interpretation | Emergence of secondary mutation (T790M, C797S); bypass pathway activation via alternative oncogene | Upregulation of alternative metabolic pathways; shift between glycolysis and OxPhos states | Stromal remodelling enabling bypass of normalising signals; fibrotic progression of TME | Epigenetic de-differentiation of non-CSC cells to CSC phenotype under therapeutic pressure | Competitive release of resistant clones following elimination of ecologically suppressive sensitive cells |
| Dietary / lifestyle relevance | Not a primary therapeutic variable; diet not expected to alter the mutational landscape | Ketogenic diet as adjunct therapy; intermittent fasting; hyperbaric oxygen (press-pulse protocol, Seyfried et al., 2017) | Dietary and endocrine factors relevant as modulators of systemic hormonal signals reaching the tissue field | Relevant only if dietary factors alter CSC niche maintenance; not a primary modality | Relevant as modifier of ecological parameters: nutrient availability, immune composition of the TME |
Vander Heiden, M.G. et al. (2009). Understanding the Warburg Effect. Science, 324, 1029–1033. · Gatenby, R.A. et al. (2009). Adaptive therapy. Cancer Research, 69, 4894–4903. · Seyfried, T.N. et al. (2017). Press-pulse: a novel therapeutic strategy. Nutrition & Metabolism, 14:19. · Laplane, L. (2016). Cancer Stem Cells: Philosophy and Therapies. Harvard UP.
Resistance management is where undeclared framework preference may have the most direct clinical stakes. Clone-selection, eco-evolutionary, and plasticity-hierarchical accounts do not agree on whether resistance was selected, induced, or ecologically released, and they point toward different therapeutic responses accordingly.
Mechanism: Resistant mutations pre-exist treatment at low clonal frequency. Drug-induced selective pressure eliminates sensitive clones, allowing pre-existing resistant variants to expand — Luria-Delbrück logic applied to oncology.
Clinical prediction: Resistance is detectable before treatment by deep sequencing. Acquired resistance mutations (EGFR T790M after erlotinib; BRAF re-activation after vemurafenib) are selection events, not induction events.
Therapeutic response: Combine agents targeting different vulnerabilities simultaneously. Next-generation inhibitors for secondary mutations (osimertinib for T790M). Liquid biopsy monitoring for emergent resistance clones.
Mechanism: Maximum-dose therapy eliminates drug-sensitive cells, creating an ecological vacuum. Resistant cells — previously suppressed by competition with sensitive cells — expand into vacated niche space. This is competitive release, not mutation selection alone (Gatenby et al., 2009).
Clinical prediction: High-dose treatment may accelerate resistance by eliminating the sensitive population that was ecologically suppressing resistant clones. Clonal composition, not tumour volume, must be tracked.
Therapeutic response: Adaptive therapy — modulate dose dynamically to maintain a population of sensitive cells. Exploit evolutionary trade-offs: resistance to one drug may confer sensitivity to another (collateral sensitivity).
CSCs are intrinsically resistant to cytotoxic therapy: typically quiescent, expressing multidrug-resistance pumps, with elevated DNA repair capacity. Under therapeutic stress, non-CSC cells can undergo epigenetic de-differentiation back to a stem-like state — resistance is partly induced, not just selected. This is a quasi-Lamarckian dynamic absent from SMT's pre-selectionist account of resistance. Treating the bulk tumour leaves the CSC subpopulation intact; the tumour regenerates from residual stem-like cells. CSC-directed agents or differentiation-inducing therapies are required in combination.
Under metabolic theory, resistance reflects the metabolic flexibility of cancer cells: the capacity to shift between glycolytic and oxidative phenotypes when one pathway is suppressed. Targeting glucose restriction alone may be countered by upregulation of glutamine as an alternative carbon source. The press-pulse protocol (Seyfried et al., 2017) addresses this by combining chronic metabolic constraint with acute drug pulses targeting remaining metabolic pathways. A clinician reasoning within SMT has no basis for prescribing this combination — the intervention is invisible as a category within that framework's account of what the disease is.
Gatenby, R.A. et al. (2009). Adaptive therapy. Cancer Research, 69, 4894–4903. · Aktipis, C.A. et al. (2012). Life history trade-offs in cancer evolution. Nature Reviews Cancer, 13, 342–350. · Gupta, P.B. et al. (2011). Stochastic state transitions in populations of cancer cells. Cell, 146, 633–644.
Trial design is not framework-neutral. When a metabolic intervention fails to meet a RECIST endpoint, this is standardly read as evidence against the metabolic hypothesis. But RECIST measures tumour volume on imaging — a criterion derived from SMT's model of disease, in which tumour mass corresponds to the cell-population burden of mutant clones. Under MTC, the primary endpoint would be metabolic normalisation, not volume reduction. A negative RECIST result in a metabolic trial does not disconfirm MTC; it records that a metabolic intervention was evaluated with the wrong instrument. Undeclared framework preference converts a methodological choice into an apparently empirical verdict.
| Trial Design Element | SMT-Derived Standard (dominant) | What Non-SMT Frameworks Require | Expected Consequence of the Mismatch |
|---|---|---|---|
| Patient stratification | Biomarker-selected by mutation status (KRAS wild-type, HER2-amplified, MSI-H). Patients enrolled based on genomic profile. | MTC: stratify by baseline metabolic phenotype (FDG uptake, lactate, mtDNA copy number). TOFT: stratify by stromal composition. Eco: stratify by clonal diversity index. | Metabolic and tissue-level therapies would be trialled in unselected populations including patients unlikely to respond on that framework's own criteria. Diluted effect sizes would generate false negative trials that enter the literature as disconfirmation. |
| Primary endpoint | OS or PFS by RECIST. Molecular complete response (absence of ctDNA). | Eco: clonal composition dynamics; resistant fraction over time. CSC: CSC frequency post-treatment. MTC: metabolic phenotype normalisation. | Adaptive therapy trials that increase clonal diversity rather than reduce tumour volume may fail RECIST while achieving their actual therapeutic objective. The trial infrastructure would not detect success on its own terms. |
| Comparator / standard of care | Best available molecularly targeted therapy or chemotherapy. Control arm reflects SMT-based treatment as background. | Metabolic interventions tested as adjuncts to standard care, preventing isolation of the metabolic effect and creating systematic confounding. | Metabolic approaches would be evaluated in the conditions least favourable to them: as add-ons to high-dose cytotoxic regimens that themselves conflict with the metabolic hypothesis. |
| Resistance definition | Disease progression on imaging. Emergence of resistance mutation by secondary sequencing. Defined as treatment failure. | In adaptive therapy, sensitive-cell depletion signals failure of the ecological strategy. Resistance emergence is an expected managed feature, not an endpoint. | Adaptive therapy trials may be halted or recorded as failures by endpoints designed around maximum-dose SMT logic. The regulatory framework would embed the theoretical assumption. |
| Regulatory pathway | FDA/EMA accelerated approval based on ORR. Companion diagnostic (CDx) required for biomarker-selected treatments. | Metabolic and dietary interventions do not fit the CDx framework. Adaptive therapy requires a regulatory category that does not exist — dynamic dosing is not evaluable under fixed-dose assumptions. | Non-SMT frameworks would face structural regulatory barriers arising not from scientific inadequacy but from framework mismatch with the approval infrastructure. |
In precision oncology, a target is "actionable" if a licensed drug exists that binds it. The term presents itself as a clinical-epistemic criterion, but it is a product of the infrastructure SMT has generated. Metabolic, stromal, and ecological targets are not actionable because the drug-development pipeline was built around mutant proteins. The absence of approved drugs is not evidence that these targets are less causally fundamental; it is evidence that the pathway from hypothesis to approval was never structured to accommodate them.
The clinician who orders mutational profiling is not making a conscious philosophical choice — she is acting within a field in which that choice has already been made by the conditions of possibility of contemporary oncology. Making framework preference explicit — as a declared assumption in trial design and clinical reasoning — is the minimal corrective. Whether a clinician holds Position I, II, III, or IV should be as reportable as the statistical assumptions of the trial they design.
Topol, E.J. (2012). Individualized medicine from prewomb to tomb. Cell, 148, 1207–1214. · Gatenby, R.A. & Brown, J.S. (2020). Integrating evolutionary dynamics into cancer therapy. Nature Reviews Clinical Oncology, 17, 675–686. · Prasad, V. & Mailankody, S. (2017). Research and development spending to bring a single cancer drug to market. JAMA Internal Medicine, 177, 1569–1575.
My B.A. thesis, Language as Reality: A Critical Reflection on Meaning and Reference — Realism: Hilary Lawson vs. Paul Boghossian (Leibniz Universität Hannover, 2022, supervised by Dr. Uljana Feest and Dr. Mathias Frisch), examines what role language plays in constituting or accessing reality, using Lawson's post-realist theory of closures and Boghossian's realist rejoinder as the central case.
Lawson's theory of closures holds that the world is fundamentally open — undivided and formless — and that human engagement with it consists in "closing" the world through the stories, models, and concepts we create. No closure accurately represents the world; all are temporary, provisional, and destined to fail. Language is one among several tools (alongside perception and action) that enables closure, but it does not refer to a mind-independent reality.
Science is one story among many. It does not occupy a privileged epistemic position over art, religion, or other ways of closing the world. Lawson views the scientific realist's pursuit of truth as a "theological" project — the assumption that there is one correct description of the world is itself a closure, not a fact about the world.
Lawson distinguishes his position from language idealism by arguing that there is a world — the "indefinite other" — accessible only through closures, none of which accurately reflects it. Unlike relativists such as Rorty or Nietzsche, who make implicit truth-claims while disavowing truth, Lawson incorporates the self-referential character of his theory into it: the theory of closures is itself a closure, and will also fail.
Boghossian defends a robust realism: there is a mind-independent external world, facts about which are not constituted by our practices, beliefs, or social conventions. Language is a real phenomenon within that world — one that allows us to form, communicate, and evaluate beliefs about mind-independent facts. A world prior to and independent of human language is not a philosophical fantasy; it is the starting assumption that the history of science, and of ordinary description, presupposes.
Against relativism, Boghossian argues that global relativism is self-refuting: it cannot be asserted as absolutely true without undermining itself, nor can it be asserted as only relatively true without collapsing into incoherence or infinite regress. Against constructivism, he argues that the cookie-cutter model commits to things having mind-independent properties that constrain our ability to carve them — properties the constructivist cannot account for.
Against epistemic relativism, Boghossian argues that any epistemic system contains general propositions about what justifies what — and these cannot be both system-relative and action-guiding. Absolute epistemic facts cannot be abandoned while retaining the normative force of any epistemic system.
The thesis defends Boghossian's realist position while taking seriously the difficulties Lawson's theory raises. The left column develops objections to the theory of closures on its own terms; the right column argues for the positive realist view.
If the world is genuinely "open" and formless, it is unclear what enables some closures to work better than others. Lawson does not explain the mechanism by which the "indefinite other" constrains our closures. The very fact that closures vary in their efficacy implies that the world has mind-independent properties that resist certain closures — a concession toward realism.
Lawson critiques relativists for the self-refuting character of their claims. Yet the theory of closures is itself a closure — a story about stories. Lawson acknowledges this ("bootstrap theory") but the acknowledgment does not dissolve the inconsistency; it merely relocates it. The theory cannot simultaneously describe all human engagement with the world and claim epistemic modesty about its own status.
Lawson provides no normative framework by which competing or conflicting closures can be evaluated. When closures clash — between communities, individuals, or epistemic systems — the theory has no resources beyond perspectivalism. This is especially acute in ethically contentious cases, where the theory's silence on value-comparison becomes a practical and philosophical deficiency.
Lawson applies a version of pessimistic meta-induction to science — because past theories failed, current theories should not be trusted. But this ignores the cumulative, approximating character of scientific progress. Scientific theories are works in progress, not categorically equivalent failures; their replacement by better theories is a mark of epistemic success, not of indiscriminate closure-production.
Lawson characterises the world as constitutively "open" and "formable" — but these are positive metaphysical claims about the nature of the world, not the modest non-assertions he intends. In attributing properties to the "indefinite other," he engages in exactly the kind of ontological commitment he aims to avoid. The theory is, the thesis argues, more metaphysically assertive than it appears.
Much of reality — stars, chemical processes, evolutionary history — preceded human existence and human language. No argument from constructivism or relativism has given sufficient reason to abandon the default that our beliefs are genuinely about that world. As Boghossian argues, the difficulty of articulating realism is not evidence against it.
Scientific progress in medicine, physics, and engineering is not explicable on a purely constructivist account. The ability to intervene reliably in the world — to treat disease, predict natural phenomena, build technology — implies that our theories are tracking something real. No closure-based account of science explains why the closures keep working.
The plurality of languages, cultures, and forms of life does not preclude objectivity. We are not limited to one language game or cultural perspective; the diversity of epistemic access is compatible with the existence of a single external world that all of these perspectives are attempting, with varying success, to describe. Cultural and linguistic diversity is evidence of different angles on the same world, not of incompatible worlds.
Following Boghossian: any epistemic system contains general propositions about justification — propositions that claim to be action-guiding regardless of who holds them. The relativist cannot coherently maintain both that epistemic facts are system-relative and that any particular epistemic system's norms are action-guiding. Abandoning absolute epistemic facts dissolves the normative force of the relativist's own theory.
These are not purely academic questions. In periods of epidemic, climate crisis, and political extremism, acting on mind-independent facts — about viral transmission, atmospheric chemistry, institutional corruption — presupposes that such facts exist. Radical subjectivism is not merely philosophically weak; it is practically inadequate to the demands placed on it.
| Dimension | Lawson (Post-Realism) | Boghossian (Realism) | Türkyilmaz's Assessment |
|---|---|---|---|
| Status of the external world | Exists as the "indefinite other" — open, formless, and constitutively inaccessible. Not reducible to mind, but not describable as it is in itself. | Mind-independent, real, and in principle knowable. Pre-exists human beings and human language. Our beliefs can be genuinely about it. | Lawson's "indefinite other" is itself a positive ontological claim: to characterise the world as constitutively open and formless is to attribute properties to it. The inaccessibility thesis is no more epistemically modest than the realism it purports to replace. |
| Role of language | One tool of closure among others (alongside perception and action). Does not refer to or represent the world; creates closures that enable engagement without describing reality. | A real phenomenon within the world. Enables genuine reference to and description of mind-independent entities. Language's success at communication is evidence of shared reference. | If language merely closes an undifferentiated openness, the differential success of descriptions across contexts and cultures is unexplained. That some accounts work better than others — persistently, across centuries — is evidence of a world that constrains description from without. |
| Status of science | One story among many. No privileged epistemic access to the world. Scientific theories have always failed and will fail again; this is evidence against scientific realism. | Provides genuine, cumulative knowledge of the world. The track record of scientific intervention (medicine, technology) supports the inference to an independently existing world. | Pessimistic meta-induction is misapplied: the progressive revision of scientific theories is the mechanism of convergence toward truth, not grounds for abandoning truth as a regulative ideal. Lawson's inference runs in the wrong direction. |
| Relativism and constructivism | Lawson distances himself from naive relativism (self-refuting) while retaining a constructivist stance. Closures are not relative to a framework but simply provisional and failing. | Both global relativism and fact-constructivism are internally incoherent. Global relativism either refutes itself or collapses into infinite regress. Constructivism requires mind-independent features it denies. | Lawson's "bootstrap" acknowledgment relocates rather than dissolves the self-reference problem. A theory that cannot substantiate its own claims without self-refutation is not made coherent by admitting this — it is made more candidly incoherent. |
| Problem of disagreement | No universal framework exists by which to adjudicate between competing closures. Disagreement is a feature of the plurality of human engagement; resolution comes from shared closures, not objective facts. | Absolute epistemic facts — about what justifies what — provide a framework for adjudicating disagreement. Without these, no epistemic system can claim to be action-guiding for anyone beyond its own adherents. | Without absolute epistemic facts, normative adjudication between competing closures becomes impossible — and the absence of shared normative ground has its greatest cost precisely in ethical and political conflict. |
| Self-reference | Acknowledged as a "bootstrap" feature of the theory. Lawson incorporates it: the theory of closures is itself a closure and will also fail. This differentiates him from unreflective relativists. | Realism does not face a self-reference problem of the same kind. Claims about the external world are not undermined by their own content in the way relativistic claims are. | Candour about a structural problem is not its solution. The theory still makes general claims about human engagement with the world — claims it cannot, by its own lights, substantiate. Acknowledging self-reference and dissolving it are not the same move. |
| Wittgenstein's legacy | Draws on Wittgenstein's later philosophy — language games, forms of life, the diversity of linguistic practices — to support the view that language does not describe a mind-independent world. Also invokes the Tractatus dictum that the limits of language are the limits of our world, treating it as confirmation that we cannot reach beyond our own conceptual closures. | Language is one of many real phenomena within the world, not a screen separating us from it. The success of communication across cultures and centuries — the fact that speakers of different languages pick out the same entities — is itself evidence that language tracks something mind-independent. Language-independent knowledge remains possible. | The anti-realist reading of Wittgenstein is contested at source. The Tractatus commits to correspondence between language and atomic facts; the later Wittgenstein's silence on metaphysics is a refusal to theorise, not a denial of the world's existence. Whether Wittgenstein himself endorsed a reductionist or anti-realist account remains a genuinely open question in the secondary literature. |
| Overall verdict | Philosophically serious attempt to navigate post-relativist antirealism; ultimately faces unresolved problems of causality, self-reference, disagreement, and implicit realist commitments. | Provides a more coherent framework: defends intuitive realism, demonstrates the incoherence of its main competitors, and retains normative resources for epistemic and ethical adjudication. | "I believe it to be valuable for political and ethical debates [...]. Boghossian, in thoroughly analyzing the different versions of relativism and constructivism [...], provided reasonable grounds to believe that there might be hope for realism after all." |
Lawson, H. (2001). Closure: A Story of Everything. Routledge. · Boghossian, P. (2006). Fear of Knowledge: Against Relativism and Constructivism. Oxford UP. · Westerhoff, J. (2020). The Non-Existence of the Real World. Oxford UP. · Wittgenstein, L. (2009). Philosophical Investigations. Wiley. · Kripke, S., Lawson, H., Boghossian, P. & Jones, S. (2019). Lost in Language. Institute of Art and Ideas.