Preprint (to be submitted). This manuscript is a working draft prepared for submission to Philosophy. It may change during review. You may cite this version as a preprint; please include the version and posted date below.
Version: v1.0 (January 27, 2026). Author: Nathan G. Whittingham.
Abstract
Scientific objectivity is often pictured, following Alexander, as direct control by reality that bypasses methods. This paper rejects that ideal without collapsing into skepticism. It argues that objectivity is an achievement of disciplined constraint: instruments, models, and concepts do not obstruct reality; they organize interaction so that error becomes difficult to sustain across heterogeneous checks. Robust agreement matters only when methods have different failure modes; it secures reliability within inquiry, not method-independent metaphysical disclosure. The paper formulates a robustness limit: convergence and success warrant epistemic authority for prediction and intervention, but do not by themselves fix ontology or vindicate the No-Miracles inference. Applications trace the same structure in rule-governed legal judgment (Palsgraf), context-conditioned quantum measurement, the present lack of triangulation in consciousness science, and the epistemic gains of inclusive research that decorrelates bias. Objectivity is thus a norm-governed, corrigible practice of constraint rather than a mirror of the world. (Alexander, 1926, pp. 12, 14, 18; Wimsatt, 1981, p. 128; Longino, 1990, pp. 68-71, 74, 80; Laudan, 1981, pp. 19, 22, 45, 47; Hart, 2012, pp. 1-6, 124-128, 272; Chalmers, 1995, pp. 200-202; Cartwright, 1983, pp. 2-6, 163).
1. Objectivity after Alexander: from control to disciplined constraint
Science presents itself as objective. Its claims are meant to hold regardless of who makes them, where they are made, or what interests are at stake. That aspiration helps explain why scientific results can travel across borders, coordinate large-scale action, and settle disputes where trust is scarce. (Porter, 1995, p. ix)
A natural starting point is the first issue of Philosophy itself. Samuel Alexander wrote that science is ‘controlled entirely from the real world it desires to understand’, contrasting it with art, which ‘builds its imputations into the material’. On this picture, science earns objectivity because reality dictates its results. Methods and models are, at best, transparent conduits. Where they intrude, they distort. (Alexander, 1926, p. 5)
That ideal still exerts influence. Even views that reject simple ‘mirroring’ often treat scientific success as evidence that inquiry has, in some important sense, broken through its own mediations. Convergence across instruments and models is taken not only as a sign of reliability, but as a sign that we have approached reality ‘as it is’. (Daston and Galison, 2007, p. 29)
Over the last century, however, the idea of unmediated control has become hard to sustain. The challenge did not come from doubts about science’s practical success. It came from closer attention to how that success is achieved.
One pressure point is theory-ladenness. Observation is not raw intake. Scientists see and record under descriptions supplied by concepts, background assumptions, and measurement conventions. Data are not simply given; they are admitted as data under practices that fix what counts as signal, noise, and relevance. The world constrains inquiry, but it does not ‘speak’ except through the languages and tools that make it answerable. (Hanson, 1958, pp. 19-26)
I will call the instruments, models, concepts, and procedural norms that format access to phenomena mediations. Mediation is not an obstacle to objectivity. It is the condition under which constraint becomes testable and comparable across agents and sites. The central question is therefore not whether science escapes mediation, but how it disciplines mediation so that error becomes difficult to sustain.
A second pressure point is historical change. Kuhn argued that scientific revolutions involve shifts in standards of explanation and evidence, not just the replacement of one hypothesis by another. Competing paradigms can structure inquiry so differently that the idea of steady convergence on a single, framework-independent description of the world becomes difficult to interpret. (Kuhn, 2012, pp. 103, 109-110, and 170-171)
These lessons undercut Alexander’s picture of control. Yet rejecting that picture can create an overreaction. If inquiry is always mediated, one might conclude that objectivity is a myth, or a mask for local values, or merely the consensus of a community. The twentieth-century debate often polarized into two unsatisfying positions: a realism that underestimates mediation, and a skepticism that treats mediation as epistemic defeat. (van Fraassen, 1980, pp. 8, 12, 13-19; Longino, 1990, pp. 22-37, 177)
This paper proposes a different revision of Alexander’s ideal. The account offered here concerns objectivity as an achievement of disciplined inquiry, not a general theory of truth, reference, or metaphysical reality. Science is objective, but not because it is ‘controlled’ by the world in a way that bypasses methods. Science is objective because it disciplines its claims through methods that expose them to constraint, error, and correction. The world does not dictate results directly. Rather, scientific practices organize interaction with the world so that many kinds of mistake become difficult to sustain.
On this view, mediation is not the enemy of objectivity. It is the condition under which objectivity is possible. Instruments, models, and conceptual frameworks do not stand between inquiry and reality as removable filters. They are the means by which reality can push back in a checkable way. Constraint operates through method, not in spite of it. (Wimsatt, 1981, p. 128; Hacking, 1983, p. 146)
Scientific objectivity is an achievement within inquiry: robust agreement across partially independent methods that constrain distinct failure modes. The robustness limit blocks the escalation from procedural stability to metaphysical disclosure unless an additional bridge principle is supplied.
If the argument succeeds, Alexander’s ambition can be retained without his mechanism. Science remains answerable to the world. But its answerability consists in disciplined constraint, not in method-free control. Objectivity is neither an illusion nor a mirror of reality. It is a robust achievement of practice that earns authority while remaining compatible with deep theoretical revision.
2. Robust agreement and the temptation to escalate
Accounts of scientific objectivity increasingly center on agreement across methods. When the same result is obtained through different experimental setups, measurement techniques, modelling strategies, or inferential routes, confidence increases. The result appears less like an artefact of a particular tool and more like a stable finding. (Weisberg, 2006, pp. 730-731; Wimsatt, 1981, p. 128)
This section clarifies what such agreement actually secures, why it is epistemically powerful, and why it so often invites a further inference it cannot support.
2.1 Robustness as constraint, not repetition
The relevant notion of agreement is not simple replication. Multiple replications of the same design can reproduce the same mistake. What matters is heterogeneity: methods that fail in different ways. When results converge despite that heterogeneity, many candidate error stories are ruled out at once. Inquiry earns stability by surviving distinct failure modes. (Wimsatt, 1981, pp. 126, 128)
A result is robust when it persists across variations in how it is produced and checked, even as sources of idealisation, calibration, background assumption, and measurement convention shift. Robustness does not eliminate uncertainty. It restructures it by distributing risk across partially independent routes rather than concentrating it within a single method. (Wimsatt, 1981, p. 128; Weisberg, 2006)
The independence condition does the epistemic work. If methods share a calibration assumption, a modelling idealisation, or a common source of bias, their agreement carries limited weight. They may be wrong together. When dominant error sources are not shared, convergence becomes more probative. Agreement then functions as a constraint on error rather than an accumulation of confirmations.
Independence, in this sense, concerns how methods can fail rather than what the phenomenon is like, and it does not require a prior grasp of the phenomenon under study. It is assessed at the level of method design and failure, not at the level of truth. Distinct methods are independent insofar as they rely on different causal interactions, instruments, inferential procedures, or modeling assumptions, such that their dominant modes of breakdown are detectably different. These differences can be investigated empirically through calibration, stress-testing, and comparison, without presupposing that any method already tracks the phenomenon correctly. Independence is therefore provisional and local, established through ongoing scrutiny of how methods fail rather than through appeal to an external benchmark. This account develops themes in robustness analysis while emphasizing failure modes as the primary unit of independence.
This also distinguishes robustness from consensus. Agreement among many users of the same technique can be epistemically cheap. Robustness requires variation that would expose error if error were present. (Weisberg, 2006, pp. 731-738)
These points clarify why mediation does not undermine objectivity. Scientific claims depend on instruments, models, background assumptions, and standards of inference. Those mediations introduce distortions, but they also create the conditions under which claims can be tested, compared, and corrected. When multiple representations simplify in different ways, convergence can stabilise a claim that does not depend on any single simplification being literally true. (Cartwright, 1983; Longino, 2002; Wimsatt, 1981)
On this view, objectivity is an achievement within inquiry. It is produced by procedures that make certain kinds of error hard to sustain across heterogeneous checks. The practical authority of science flows from this discipline.
Robust agreement is identified only relative to standards that determine what counts as evidence, error, and result-identity within inquiry. Agreement therefore shows that a claim is stable under those standards and has survived the forms of testing they impose. It does not, by itself, establish that the claim is independent of those standards or that it fixes a unique ontology. Any inference from robustness to such independence requires additional premises connecting epistemic success to metaphysical commitment. I call this constraint the robustness limit.
2.2 Escalation: from procedural stability to metaphysical disclosure
The temptation to escalate arises naturally. When results survive diverse tests, agreement begins to feel like bedrock. Stability achieved under variation can be mistaken for independence from the very practices that produced it. What begins as a procedural mark of reliability inside inquiry is treated as evidence that inquiry has disclosed how the world is ‘in itself’.
The inference has a familiar structure. Robust agreement shows that a claim is stable under the tests imposed by existing methods. That stability is then taken to warrant a stronger conclusion: that the claim tracks reality in a way that no longer depends on those methods. Procedural success is redescribed as metaphysical disclosure.
This step is not careless. It reflects a genuine pressure generated by scientific success. Convergence across heterogeneous approaches makes it difficult to explain success without appealing to something more than local reliability. The error lies not in feeling the pressure, but in how it is discharged.
The escalation rests on treating procedural stability under existing standards as warrant for a claim that no longer depends on those standards.
2.3 Scientific realism and the No-Miracles inference
Scientific realism gives the escalation its most explicit form. According to the No-Miracles argument, the predictive and technological success of science would be inexplicable if our theories were not at least approximately true. Robust agreement is therefore asked to do more than secure objectivity within inquiry. It is asked to explain why inquiry succeeds at all. (Boyd, 1983; Putnam, 1975, p. 73)
The objection here is not to realism as such, nor to the acknowledgment that scientific inquiry is fallible, mediated, and revisable. The target is a more specific inferential step: treating robust epistemic success as sufficient, by itself, to warrant determinate ontological commitment. Robust convergence is taken to bridge the gap from stability under inquiry to claims about what exists independently of the conditions under which that stability is achieved. (Laudan, 1981, pp. 19-22; Stanford, 2006, pp. 18-22)
This step is unwarranted. Robust agreement explains reliability by showing how error is constrained across partially independent routes of investigation. It accounts for why inquiry supports prediction, intervention, and coordination. What it does not do, on its own, is fix a unique ontology or supply warrant for metaphysical claims that outrun the structure of inquiry that produced the success. Robust agreement can motivate realist inquiry, but it cannot by itself discharge the realist’s explanatory burden without further premises connecting epistemic success to ontological commitment. In particular, an inference from success to approximate truth requires an independent account of why inference to the best explanation is reliable in this domain, rather than merely an appeal to success itself.
Historical cases sharpen the point. Past theories achieved high levels of empirical success, methodological discipline, and practical utility while being later rejected as fundamentally mistaken about the nature of reality. Phlogiston, caloric, the luminiferous aether, and classical determinism were not marginal failures; they structured inquiry, coordinated research, stabilized results, and supported intervention across wide domains. These cases show that robust epistemic success can coexist with incompatible ontological interpretations, indicating that success alone underdetermines metaphysical commitment. This pattern is the core insight of the Pessimistic Meta-Induction: repeated episodes of sustained empirical success coupled with later ontological rejection undermine any inference from success alone to metaphysical truth. The lesson is not skeptical, but disciplinary: success within inquiry underdetermines metaphysical commitment unless supplemented by an additional bridge principle. (Kuhn, 2012, pp. 2-3, 170-171; Laudan, 1981, pp. 33, 46-47)
Realist responses often attempt to soften this result by weakening truth into an elastic notion of approximate truth or by retreating from ontology to structure. Such moves may preserve a realist orientation, but they do not repair the inferential gap at issue here. They continue to treat robust success as a route from procedural achievement to metaphysical commitment without independently justifying that transition. The explanatory demand remains unmet. (van Fraassen, 1980, pp. 37-40; Worrall, 1989, pp. 117-123)
None of this entails skepticism about the world. The resistance of the world remains indispensable. Failed predictions, recalcitrant measurements, and technological breakdowns all testify to constraint beyond our choosing. The point is that constraint does not, by itself, yield transparency. Scientific success can be fully intelligible as the product of disciplined constraint while leaving open what, if anything, that success licenses about reality beyond the conditions under which it is achieved. (Hacking, 1983, p. 146)
One might press the point further by insisting that disciplined constraint itself constitutes a metaphysical clue: if the world resists inquiry in stable ways across heterogeneous methods, then there must be some underlying structure responsible for that resistance. The present account does not deny this minimal claim. It denies only that causal stability, by itself, fixes a determinate structure rather than leaving multiple structural interpretations equally compatible with the same pattern of resistance. The history of science repeatedly shows that stable causal relations can be captured by incompatible formalisms and ontological frameworks. Causal constraint therefore supports the existence of structure in a minimal sense, but it does not license commitment to any particular structure without an additional bridge principle connecting successful modeling to metaphysical description. (Laudan, 1981; Longino, 1990, pp. 74-80; Worrall, 1989, pp. 117-121)
2.4 Applying the limit
The remaining sections track the same inferential temptation across domains in which epistemic stability is routinely mistaken for metaphysical disclosure.
In law, agreement across doctrinal frameworks stabilises judgment without generating a standpoint outside legal practice. In quantum mechanics, objectivity is secured only under explicitly specified measurement conditions. In consciousness research, the absence of robust convergence explains the present fragility of objectivity without invoking a principled metaphysical barrier. In debates over inclusion, expanding methodological variation strengthens objectivity by decorrelating sources of error.
Across these cases, the issue is not whether inquiry is constrained, but whether stability achieved under constraint is asked to license claims that outrun the conditions under which agreement is assessed.
2.5 Constraint without control
The preceding analysis relies on a notion of constraint that is easily misunderstood. To clarify the revision of Alexander’s ideal proposed here, it is useful to distinguish three senses in which inquiry can be constrained.
First, there is causal constraint. The world limits what can be done and what can occur. Instruments break, measurements fail, predictions miss their targets, and interventions produce unexpected effects. These forms of resistance are not optional. They register the fact that inquiry is directed at a world that does not bend to our wishes. Causal constraint is indispensable, but by itself it does not secure objectivity. Causal impact can explain why beliefs arise without explaining why they are warranted.
Second, there is epistemic constraint. Claims are constrained by standards that determine what counts as evidence, error, success, and failure. Calibration procedures, statistical thresholds, replication norms, and admissibility rules limit which inferences may be drawn from which data. Epistemic constraint is what allows inquiry to distinguish signal from noise and error from achievement. Robust agreement across methods operates at this level by exposing claims to independent ways of failing.
Third, there is normative constraint. Inquiry is governed by commitments to consistency, coherence, accountability, and corrigibility. These norms do not describe how the world behaves; they regulate how claims are assessed and revised. Normative constraint is what gives epistemic standards their authority within inquiry rather than reducing them to mere conventions.
Alexander’s picture of control obscures this layered structure of constraint by suggesting a direct line from world to result. Once control is understood that way, mediation looks like an obstacle and successful convergence invites an illicit upgrade: stability achieved under disciplined procedures is treated as if it warranted conclusions about reality that no longer depend on those procedures. On the present account, constraint names a structured relation instead. The world resists inquiry causally, methods discipline claims epistemically, and norms regulate assessment and revision. These constraints do not bypass mediation; they operate through the practices that render results testable, comparable, and corrigible. This is why ‘control by the real world’ so easily slides into metaphysical inflation, and why the better ideal is disciplined constraint rather than unmediated control.
This constraint follows directly from the structure just described. Because epistemic and normative constraints are inseparable from the standards that govern inquiry, agreement secured under those standards cannot certify independence from them. Robustness shows that a claim has survived demanding forms of constraint; it does not show that constraint has been transcended. Treating disciplined constraint as warrant for ontological conclusions that outrun the conditions under which constraint operates therefore repeats Alexander’s mistake in a subtler form.
Understanding objectivity in terms of constraint rather than control preserves what matters about scientific authority while respecting its limits. Science is answerable to the world, but that answerability is mediated, structured, and norm-governed. Objectivity arises not from escaping those conditions, but from taking them seriously.
Objection: Does this weaken objectivity?
One might object that limiting objectivity to what can be secured within inquiry weakens its authority, replacing truth with mere procedural stability. This worry rests on the assumption that objectivity must involve access to reality independent of the conditions under which claims are assessed. The present account rejects that assumption. Objectivity earns its authority not by escaping mediation, but by disciplining it through methods that constrain error across heterogeneous checks. This constraint does not diminish objectivity; it fixes its scope by blocking inferences that outrun what disciplined inquiry can warrant. What is weakened is not scientific authority, but a mistaken inflation of that authority into metaphysical disclosure.
3. Ontological restraint and epistemic authority
Blocking the escalation from epistemic stability to metaphysical disclosure raises a further question: what posture toward ontology remains once this constraint is respected? The position defended here is ontological restraint. It does not deny a mind-independent world, nor does it rule out metaphysical speculation in principle. It denies only that robustness, convergence, or epistemic success by themselves supply warrant for claims about what exists independently of the conditions under which those achievements are assessed.
Neither conclusion follows. The constraint at issue governs inference, not ontology. It rules out a specific argumentative move – from procedural stability to claims of method-independent truth – without dictating a comprehensive metaphysical stance. What it enforces is not skepticism about the world, but discipline about what epistemic achievements can warrant.
To clarify this point, it is useful to distinguish three postures toward metaphysical inference that are often conflated in debates about objectivity.
The first is escalatory metaphysical disclosure. On this posture, robust convergence across methods is taken to reveal the structure of reality independently of the practices that produced that convergence. Agreement is treated as evidence that inquiry has transcended its own conditions and accessed what the world is like ‘in itself’. The error lies in moving from stability achieved under assessment to claims that purport to be independent of the conditions that make assessment possible.
The second posture is metaphysical silence. On this view, once claims about method-free truth are rejected, nothing substantive can be said about reality at all. Science is reduced to prediction or coordination, with metaphysical questions dismissed as meaningless or misguided. Although this posture is often attributed to empiricist or instrumentalist views, it does not follow from the present account. Rejecting escalation does not require refusing all metaphysical commitment; it requires refusing a particular route to such commitment.
The third posture is ontological restraint, which is the position defended here. On this view, metaphysical claims are not ruled out in principle, but their warrant cannot be supplied by robustness alone. Robust agreement justifies confidence in reliability, intervention, and coordination. It does not, by itself, settle questions about ultimate ontology. Metaphysical claims may still be proposed and debated, but they must be supported by arguments that do not simply piggyback on epistemic stability. Such proposals do not inherit epistemic authority from scientific objectivity itself; they stand or fall on independent philosophical grounds.
This restrained posture preserves a meaningful distinction between epistemic authority and metaphysical ambition. Scientific inquiry earns authority by constraining error, stabilizing results, and coordinating action across heterogeneous checks. That authority is real and normatively significant. What it does not automatically provide is a license to infer what exists independently of the practices through which claims are assessed.
Seen this way, the present account reframes rather than rejects realism. If one wishes to defend a realist metaphysics, one must supply arguments that do not rely on convergence, success, or robustness alone. Appeals to explanation or theoretical virtue may still be advanced, but they must be defended as metaphysical arguments in their own right, not smuggled in under the banner of objectivity.
Separating epistemic authority from metaphysical ambition also clarifies why acknowledging this constraint does not undermine trust in science. Trust is not an all-or-nothing attitude grounded in metaphysical certainty. It is calibrated to the structure of inquiry. Where claims are supported by diverse, partially independent methods that constrain distinct failure modes, confidence in their reliability is warranted. Where such support is absent, disagreement and caution are appropriate. Respecting this constraint does not leave us with less than science has achieved. It leaves us with exactly what science has earned – and no more than that.
What this shows is that any genuine bridge from epistemic success to ontological commitment would have to involve more than robustness itself. It would require an independent account linking successful inquiry to what exists – whether through a theory of representation, a causal account of successful intervention, a semantic account of reference under error, or a domain-specific justification of inference to the best explanation. The present argument does not deny that such bridges could be proposed. It insists only that robustness alone cannot supply them.
Describing metaphysical bridges as optional does not trivialize the aim of truth or the motivations of scientific practice. Belief that inquiry reveals how the world is often functions as a powerful driver of discovery, hypothesis formation, and perseverance. The present claim concerns justification, not motivation. Disciplined constraint explains why scientific results earn epistemic authority regardless of the metaphysical commitments that guide inquiry. Realist beliefs may animate the logic of discovery, but the authority of what is discovered depends on robustness, error-detection, and constraint rather than on the prior truth of those beliefs. Metaphysical bridges, on this view, are layered onto epistemic authority rather than constitutive of it.
4. Agreement under rules: the legal fact-pattern case
Law offers a clear case in which agreement across different approaches strengthens judgment without creating a standpoint beyond those approaches. The value of the case lies not in equating law with science, but in making visible a shared structure: in both domains, stability and authority arise through disciplined constraint rather than through access to standards outside the practice.
The analogy presupposes a procedural or positivist understanding of law, on which legal authority depends on rule-governed practice rather than correspondence to moral reality. On this view, what confers legal authority is not access to truth outside the law, but the disciplined application of shared standards within it. (Hart, 2012, pp. 94-95)
Just as the facts in a case count as facts only because they are admitted under shared rules of evidence, scientific data count as data only because they are admitted under shared methodological standards. And just as hearsay is excluded not because it is false but because it is unreliable under the rules, experimental noise is excluded not because it is unreal but because it fails to support disciplined inference. (Federal Rules of Evidence, Rules 801, 802, 803, Advisory Committee Notes)
The structural point becomes clearest in a hard case. Palsgraf v. Long Island R.R. Co. is a foundational tort decision that turns on a deep doctrinal divide about how legal responsibility should be understood. (Palsgraf v. Long Island R. R. Co., 1928)
The facts are simple and undisputed. A man carrying a package runs to catch a train. Railroad employees assist him aboard. In the process, the package falls, explodes, and dislodges scales at the far end of the platform, injuring Helen Palsgraf. No one disputes the sequence of events. The dispute concerns how those facts should be taken up. (Palsgraf v. Long Island R. R. Co., 1928)
One line of analysis, articulated by Benjamin Cardozo, frames the issue in terms of duty. On this view, negligence depends on whether the defendant owed a duty of care to the plaintiff. Duty is relational and bounded by foreseeability. Because the railroad employees could not reasonably foresee harm to someone standing far away, no duty was owed to Palsgraf. Without duty, there is no negligence, regardless of the causal chain. (Palsgraf v. Long Island R. R. Co., 1928)
A competing line of analysis, articulated by William Andrews, rejects this restrictive conception of duty. On Andrews’s view, negligence turns on proximate cause rather than relational duty. Once negligent conduct is established, liability depends on whether the resulting harm falls within a legally acceptable causal chain. Foreseeability matters, but it does not function as a threshold for duty. Under this analysis, the railroad could be held liable. (Palsgraf v. Long Island R. R. Co., 1928)
These approaches are not minor variations. They reflect different ways of structuring legal judgment. Cardozo emphasizes relational limits on responsibility. Andrews emphasizes causal continuity. Each doctrine highlights different features of the same fact pattern and applies different standards of relevance.
What matters for present purposes is what remains stable across the divergence. Across duty-based and causation-based analyses, certain constraints remain fixed. The railroad employees acted in assisting the passenger. The package fell. The explosion occurred. The injury followed. No doctrinal framework treats these facts as optional. Disagreement concerns how the facts should be legally interpreted, not what happened. (Palsgraf v. Long Island R. R. Co., 1928)
When courts converge on an outcome despite doctrinal disagreement, that convergence is an achievement within legal practice. Here, doctrinal variation supplies genuine independence: duty-based and causation-based analyses expose judgments to different failure modes, so agreement reflects mutual constraint rather than shared bias. Convergence is produced by testing competing interpretations against the same constrained record under shared rules of admissibility and relevance. By applying a social account of objectivity to the law, we see that the tension between Cardozo and Andrews is not a flaw, but a redundant check that increases the robustness of the legal outcome. (Longino, 2002, pp. 129-135; Prosser, 1953, pp. 5-6; Weisberg, 2006)
Agreement across legal approaches earns authority by surviving variation under constraint. Treating such convergence as access to a law-independent reality mistakes procedural stability for warrant to draw conclusions that exceed the standards under which judgment is made.
The analogy to law is structural rather than ontological. In legal practice, constraint is normative through and through: what counts as a fact depends entirely on rules of admissibility, procedure, and institutional authority. In science, by contrast, normative and epistemic standards mediate inquiry, but the ultimate source of constraint is material. The world resists our claims through failed predictions, recalcitrant measurements, and breakdowns in intervention, regardless of how we choose to describe or regulate those claims. Changing legal rules changes legal facts; changing scientific standards does not change what the world will and will not permit. The shared structure is disciplined constraint, but the source of that constraint is fundamentally different. (Hart, 2012)
The force of the analogy is epistemic rather than metaphysical. In both law and science, objectivity consists in the disciplined assessment of claims under shared rules of admissibility that determine what counts as evidence, relevance, and error. The difference lies not in how verdicts are reached, but in what those verdicts concern: legal practice adjudicates institutional facts under normative authority, while scientific practice adjudicates claims about a world that resists inquiry independently of our beliefs. The analogy therefore illuminates the logic by which agreement is stabilized, not the nature of the facts being judged.
5. Quantum mechanics and disciplined constraint
Quantum mechanics provides the clearest case in which objectivity cannot be understood as control over pre-given properties. What is measured is inseparable from the experimental arrangement that makes measurement possible. The question, therefore, is not whether quantum theory undermines objectivity, but what kind of objectivity remains available once this constraint is acknowledged. (Bohr, 1949, pp. 210-211)
On the quantum case, properties are not available to inquiry independently of the conditions under which they are measured. Position and momentum, spin along different axes, or particle number under different experimental arrangements are not revealed by stripping away interference from measurement; they are defined through it. What inquiry secures is not access to properties ‘as such’, but stability of results under a specified experimental constraint. (Bohr, 1949, pp. 214-215)
It is important to distinguish between fixing the conditions under which a property is defined and fixing the outcome of a measurement. In quantum mechanics, the choice of experimental arrangement determines which observable is well-defined, but it does not determine how the system will respond once that observable is fixed. The material constraint lies not in the selection of the measurement context, but in the stable statistical distribution of outcomes that persists across implementations of that context. Changing the apparatus changes the question asked of the system; it does not grant control over the answers the world permits. (Bohr, 1949, pp. 210-211, 232-234)
This point is often miscast as observer-dependence. The relevant dependence, however, is not on a conscious observer, but on an experimental context that fixes what counts as a result, which variables are relevant, and how outcomes are to be compared. Different apparatuses pose different questions to the system. Quantum theory answers those questions with precision once the constraints are fixed. (Heisenberg, 1958, pp. 54-55, 58)
Objectivity survives here precisely because constraint is disciplined. Measurement contexts are specified, repeatable, and publicly accessible. Different laboratories implementing the same experimental arrangement obtain the same statistical results. Objectivity consists in the reproducibility of outcomes across distinct implementations of the same constraint, not in convergence on a description of properties independent of measurement. In this respect, the role of the measurement context in quantum mechanics is directly analogous to the role of rules of admissibility in law: both fix the conditions under which results count as evidence and thereby make agreement possible, without appealing to a standpoint outside the practice. Where experimental contexts are misaligned, no such independent check is available, and objectivity correspondingly fails. (Longino, 2002, p. 127)
Quantum mechanics therefore exemplifies the account of objectivity defended in this paper rather than challenging it. Objectivity consists in producing stable, reliable results under clearly specified conditions of investigation. Quantum theory does this exceptionally well, but only by making explicit the constraints under which claims are meaningful and testable. (van Fraassen, 1980; Longino, 1990)
The mistake is to treat this stability as evidence for claims that extend beyond those conditions – whether in the direction of observer-dependent reality or metaphysical indeterminacy ‘as such’. What quantum mechanics shows instead is that objectivity is conditional: results are stable only relative to specified experimental constraints. Attempting to generalize beyond those constraints does not deepen our grasp of reality; it abandons the very conditions under which objectivity is achieved. (Bohr, 1949, pp. 233-234)
In this domain, the physics itself enforces the constraint model. Attempts to treat quantum properties as fully context-independent generate familiar interpretive puzzles, whereas specifying the measurement arrangement is what renders claims testable, comparable, and corrigible. When constraints are properly fixed, quantum mechanics delivers robust agreement and extraordinary predictive power. The discipline of constraint, not the elimination of mediation, is what makes objectivity possible here. (Bell, 1964; Peres, 1993)
6. Consciousness and the limits of robustness
The question of whether consciousness can be measured objectively is often treated as marking a principled boundary of scientific inquiry. Conscious experience appears private, subjective, and resistant to third-person description. From this, it is commonly concluded that consciousness resists objective treatment in principle. (Chalmers, 1995, pp. 200-202, 216)
This conclusion is premature.
The difficulty consciousness poses for objectivity is not unique in kind; it is structural. The science of consciousness currently lacks the degree of robustness that secures objectivity in more mature domains. The problem is not that consciousness is intrinsically inaccessible, but that available methods do not yet triangulate into a stable fact-pattern. (Cogitate Consortium et al., 2025; Longino, 1990, p. 74; Wimsatt, 1981, p. 128)
Research on consciousness employs several distinct approaches. First-person reports describe what subjects experience. Behavioral measures track performance, attention, and response. Neurophysiological tools such as EEG and fMRI identify patterns of neural activity correlated with reported experience. Each method contributes genuine information. None can be discarded. But taken together, they do not yet converge in a way that fixes a stable target across inquiry under shared standards of assessment. (Cogitate Consortium et al., 2025; Lepauvre & Melloni, 2021; Varela & Shear (eds.), 1999)
This lack of convergence has predictable consequences. Competing theories coexist despite shared data. The same neural signatures are interpreted differently under different background assumptions. First-person reports are alternately treated as privileged access, unreliable noise, or mere heuristics. Disagreement persists not because consciousness is ineffable, but because the discipline lacks settled procedures for coordinating these methods. (Doerig et al., 2021; Ericsson & Simon, 1993; Lepauvre & Melloni, 2021)
Seen through the lens of robustness, the diagnosis is straightforward. The central difficulty is that current methods do not supply cross-modal veto power: first-person, behavioral, and neural measures tend to fail together rather than independently, leaving error sources correlated rather than mutually constraining. In domains where objectivity is secure, methods expose claims to independent failure modes; errors that survive one approach are caught by another. In consciousness research, by contrast, many error sources remain correlated. The same assumptions shape experimental design, data interpretation, and theoretical framing across methods. As a result, apparent agreement is fragile and disagreement is expected. (Cogitate Consortium et al., 2025; Weisberg, 2006, p. 731; Wimsatt, 1981, p. 128)
Even if the so-called Hard Problem remains a substantive metaphysical issue, there is also a distinct methodological problem that bears directly on objectivity: coordinating first-person reports, behavioral measures, and neural data into partially independent constraints. At present, these methods are insufficiently aligned to expose claims to genuinely independent failure modes. The resulting lack of robustness explains why objectivity in this domain remains fragile, without implying that consciousness lies beyond scientific inquiry in principle. (Chalmers, 1995, pp. 204, 216-217; Longino, 1990, pp. 70-75; Lepauvre & Melloni, 2021)
Treating this fragility as evidence that consciousness lies beyond scientific reach mistakes the absence of robust convergence for warrant to draw a principled metaphysical conclusion. What is shown is only that existing methods have not yet stabilized the target of inquiry, not that such stabilization is impossible in principle.
The appropriate conclusion is methodological rather than metaphysical. Objectivity in this domain will not be achieved by eliminating subjectivity, but by disciplining it. First-person reports function as instruments with characteristic limits, noise profiles, and biases within inquiry. Like any instrument, they require calibration and cross-checking against independent measures. Until such coordination is achieved, disagreement reflects insufficient constraint rather than failure of science. (Varela & Shear, 1999)
Treating first-person reports as instruments is a methodological choice made for the purposes of objectivity, not a metaphysical claim about the nature of mind or subjectivity. It does not deny that the first-person perspective is constitutive of inquiry itself, nor does it reduce experience to a mere data source on a par with physical sensors. Rather, it marks a shift in standpoint: when consciousness becomes a target of scientific investigation, reports of experience must be handled in ways that allow their reliability and limits to be assessed alongside other measures. This instrumentalization concerns how claims about experience are stabilized within inquiry, not what experience ultimately is.
Consciousness thus provides a revealing stress test for the present account. It shows that objectivity does not require metaphysical transparency, but it does require methodological maturity. Where robustness is lacking, objectivity will be limited. Where methods can be coordinated so that claims are exposed to genuinely independent tests, objectivity can increase. (Longino, 2002, p. 190; Varela & Shear, 1999, pp. 1, 5)
It is important to be clear about the scope of this claim. The present account does not deny that consciousness may involve an irreducible first-person aspect, nor does it attempt to dissolve the so-called Hard Problem by methodological fiat. Disciplined constraint is not a theory of consciousness, but a criterion of objectivity. Constraint sets the conditions under which theories can earn objectivity, but it is also the mechanism through which more adequate theories are eventually developed as methods become better coordinated and failure modes are progressively exposed. Even if the ‘view from within’ marks a genuine ontological gap, inquiry can still succeed or fail at stabilizing claims about consciousness under shared standards of assessment. The fragility of objectivity in this domain reflects the current inability of available methods to constrain one another independently, not a verdict on whether consciousness is ultimately the right kind of thing to be explained. (Varela, 1996, p. 333)
7. Inclusion, bias, and error detection in scientific practice
The account of objectivity developed in this paper has a direct consequence for how bias and exclusion should be understood in scientific practice. If objectivity consists in agreement across methods that constrain error, then failures of inclusion are not merely ethical or political defects. They are epistemic failures that weaken objectivity itself. (Longino, 1990; Longino, 2002; Weisberg, 2006; Wimsatt, 1981)
This point is illustrated clearly by a well-documented episode in medical research. For much of the twentieth century, clinical trials routinely excluded female subjects. The stated justification was methodological control. Hormonal variation was treated as noise. Male bodies were taken as the stable baseline. Results obtained under these conditions were then generalized to patients as such. (Epstein, 1996, pp. 258-260; Schiebinger, 1999, pp. 113-114; Food and Drug Administration, 1993)
At the time, this practice did not appear unscientific. Trials were controlled, repeatable, and statistically rigorous by prevailing standards. Agreement across studies was sometimes achieved. From within the prevailing methodological frame, inquiry appeared objective. (Epstein, 1996; Schiebinger, 1999; Food and Drug Administration, 1993)
The epistemic flaw in these trials was the correlation of error sources across methods. Treating the resulting consensus as a baseline for ‘the human body’ mistakes stability achieved under homogeneous assumptions for warrant to generalize beyond those conditions. The same background assumptions shaped subject selection, experimental design, and interpretation of results. Variation that might have exposed error was systematically excluded. (Longino, 1990; Longino, 2002; Wimsatt, 1981; Weisberg, 2006; Oreskes, 2019)
The consequences were substantial. Dosage recommendations, symptom profiles, and side-effect risks derived from male-only trials sometimes failed when applied to female patients. Drugs were metabolized differently. Adverse reactions appeared that had not been detected. What had looked like stable knowledge proved brittle under variation. (Schiebinger, 1999, pp. 114–115; Food and Drug Administration, 1993)
Crucially, the problem was not resolved by increasing agreement of the same kind. It was resolved by changing the methods themselves, a shift codified in new regulatory standards. Including female subjects introduced new forms of variation that disrupted the existing consensus. (Food and Drug Administration, 1993; Epstein, 1996; Schiebinger, 1999;)
The epistemic gain came from decorrelating failure modes: background assumptions that had shaped subject selection, experimental design, and interpretation together were forced apart, allowing error to be detected rather than stabilized. Results that had appeared robust were exposed as method-dependent. Claims were forced to survive a wider range of tests. (Longino, 1990; Wimsatt, 1981; Weisberg, 2006)
This episode illustrates the epistemic mechanism at work. Homogeneous communities tend to share background assumptions about which questions matter, which variables can be ignored, and how results should be interpreted. When those assumptions are built into multiple methods at once, agreement becomes cheap: claims survive not because they are well constrained, but because they are insufficiently challenged. Inclusion strengthens objectivity by introducing genuinely unshared assumptions, incentives, and saliencies, thereby decorrelating error sources across methods and expanding the space of possible failure. (Longino, 1990; Longino, 2002; Oreskes, 2019; Wylie, 2003)
Social accounts of objectivity capture this dynamic at the level of community practice. The present account sharpens their insight by locating its force in epistemic structure rather than in norms of interaction alone: inclusion strengthens objectivity precisely insofar as it increases the number of independent ways a claim can fail. (Longino, 2002, p. 135; Weisberg, 2006; Wimsatt, 1981)
Bias, on this view, is a predictable outcome of insufficient methodological variation. Correcting it does not weaken objectivity; it takes its demands seriously. (Longino, 1990, p. 214)
8. Conclusion
This paper has argued that scientific objectivity is best understood as an achievement of disciplined inquiry rather than as access to reality or truth as such. Agreement across methods secures objectivity by constraining error, stabilizing results, and holding claims answerable to shared standards. What it does not secure is a standpoint beyond inquiry from which those standards could be bypassed or rendered transparent.
This constraint fixes the scope of what objectivity can warrant: agreement can earn epistemic authority only under the standards that make agreement assessable, so robustness alone cannot fix determinate ontological commitment without an additional bridge principle. Respecting this limit does not weaken objectivity; it clarifies its scope.
The account is not skeptical. It does not deny the existence of a mind-independent world, nor does it treat inquiry as self-contained. The world resists our claims through failed predictions, recalcitrant measurements, and technological breakdowns. But that resistance is always encountered through methods that fix what counts as success or failure. Constraint is real, yet never unmediated.
Understood in these terms, objectivity preserves what is distinctive about science without inflating its success into metaphysical transparency. It explains how scientific claims can travel across contexts, coordinate action, and guide reliable intervention while remaining open to revision. It also explains why disagreement can persist without failure and why error correction depends on disciplined variation rather than consensus or detachment. (Daston and Galison, 2007, p. 415)
Objectivity, on this view, is neither an illusion nor a mirror of reality. It is an achievement of disciplined practice. Constraint does not function as a pragmatic synonym for truth: it operates negatively, by ruling out what cannot be sustained under disciplined inquiry, rather than positively, by dictating what must be asserted about the world. Science is objective not because it escapes its methods, but because it submits to their constraint.
References
Samuel Alexander, ‘Art and Science’, Philosophy, 1 (1926), 5-19.
John S. Bell, ‘On the Einstein–Podolsky–Rosen Paradox’, Physics, 1 (1964), 195-200.
Niels Bohr, ‘Discussion with Einstein on Epistemological Problems in Atomic Physics’, in Paul Arthur Schilpp (ed.), Albert Einstein: Philosopher–Scientist (La Salle, IL: Open Court, 1949), 199-241.
Richard N. Boyd, ‘On the Current Status of the Issue of Scientific Realism’, Erkenntnis, 19 (1983), 45-90.
Nancy Cartwright, How the Laws of Physics Lie (Oxford: Oxford University Press, 1983).
David J. Chalmers, ‘Facing Up to the Problem of Consciousness’, Journal of Consciousness Studies, 2 (1995), 200-219.
Cogitate Consortium et al., ‘Adversarial Testing of Global Neuronal Workspace and Integrated Information Theories of Consciousness’, Nature, 642 (2025), 133-142.
Lorraine Daston and Peter Galison, Objectivity (New York: Zone Books, 2007).
Adrien Doerig, Aaron Schurger, and Michael H. Herzog, ‘Hard Criteria for Empirical Theories of Consciousness’, Cognitive Neuroscience, 12 (2021), 41–62.
Steven Epstein, Impure Science: AIDS, Activism, and the Politics of Knowledge (Berkeley: University of California Press, 1996).
K. Anders Ericsson and Herbert A. Simon, Protocol Analysis: Verbal Reports as Data, rev. edn (Cambridge, MA: MIT Press, 1993).
Federal Rules of Evidence, Rules 801, 802, and 803, Advisory Committee Notes.
Food and Drug Administration, Guideline for the Study and Evaluation of Gender Differences in the Clinical Evaluation of Drugs (Washington, DC: FDA, 1993).
Ian Hacking, Representing and Intervening: Introductory Topics in the Philosophy of Natural Science (Cambridge: Cambridge University Press, 1983).
Norwood Russell Hanson, Patterns of Discovery: An Inquiry into the Conceptual Foundations of Science (Cambridge: Cambridge University Press, 1958).
H. L. A. Hart, The Concept of Law, 3rd edn, with a new introduction by Leslie Green (Oxford: Oxford University Press, 2012).
Werner Heisenberg, Physics and Philosophy: The Revolution in Modern Science (London: George Allen & Unwin, 1958).
Thomas S. Kuhn, The Structure of Scientific Revolutions, 4th edn (50th Anniversary Edition), with an introductory essay by Ian Hacking (Chicago: University of Chicago Press, 2012).
Larry Laudan, ‘A Confutation of Convergent Realism’, Philosophy of Science, 48 (1981), 19-49.
Alex Lepauvre and Lucia Melloni, ‘The Search for the Neural Correlate of Consciousness: Progress and Challenges’, Philosophy and the Mind Sciences, 2 (2021), Article 4.
Helen E. Longino, Science as Social Knowledge: Values and Objectivity in Scientific Inquiry (Princeton: Princeton University Press, 1990).
Helen E. Longino, The Fate of Knowledge (Princeton: Princeton University Press, 2002).
Naomi Oreskes, Why Trust Science? (Princeton: Princeton University Press, 2019).
Palsgraf v. Long Island R. R. Co., 248 N.Y. 339, 162 N.E. 99 (1928).
Asher Peres, Quantum Theory: Concepts and Methods (Dordrecht: Kluwer Academic Publishers, 1993).
Theodore M. Porter, Trust in Numbers: The Pursuit of Objectivity in Science and Public Life (Princeton: Princeton University Press, 1995).
William L. Prosser, ‘Palsgraf Revisited’, Michigan Law Review, 52 (1953), 1-31.
Hilary Putnam, Mathematics, Matter and Method (Cambridge: Cambridge University Press, 1975).
Londa Schiebinger, Has Feminism Changed Science? (Cambridge, MA: Harvard University Press, 1999).
P. Kyle Stanford, Exceeding Our Grasp: Science, History, and the Problem of Unconceived Alternatives (Oxford: Oxford University Press, 2006).
Bas C. van Fraassen, The Scientific Image (Oxford: Oxford University Press, 1980).
Varela, F. J. (1996). Neurophenomenology: A methodological remedy for the hard problem. Journal of Consciousness Studies, 3(4), 330-349.
Francisco Varela and Jonathan Shear (eds.), The View from Within: First-Person Approaches to the Study of Consciousness (Exeter: Imprint Academic, 1999).
Michael Weisberg, ‘Robustness Analysis’, Philosophy of Science, 73 (2006), 730-742.
William C. Wimsatt, ‘Robustness, Reliability, and Overdetermination’, in Marilyn Brewer and Barry Collins (eds.), Scientific Inquiry and the Social Sciences (San Francisco: Jossey-Bass, 1981), 124-163.
John Worrall, ‘Structural Realism: The Best of Both Worlds?’, Dialectica, 43 (1989), 99-124.
Alison Wylie, ‘Why Standpoint Theory Matters’, in Louise Antony and Charlotte Witt (eds.), A Mind of One’s Own (Boulder, CO: Westview Press, 2003), 339-351.
Nathan Whittingham 