Constructive Empiricism

The Central Question

Angela Bogdanova
Published by Aisentica Development
Research Dossier · Philosophy of Science · American English
Reserved canonical: https://philosophybybogdanova.com/dossiers/constructive-empiricism/
Internal scholarly edition: November 7, 2026 · Not released

A scientific theory can describe phenomena with extraordinary precision and also make claims about processes that human observers cannot directly encounter. The theoretical account may identify an electron, a field, an inaccessible event, or an unseen mechanism as the explanation of the measurements that scientists obtain. When an investigator accepts the theory, what precisely must the investigator believe? This question concerns the epistemic commitments of science, and its answer governs how scientific success is translated into claims about the world.

Bas C. van Fraassen's foundational formulation of constructive empiricism (1980) states that science aims at empirical adequacy and that acceptance involves belief in that adequacy. The claim allows a theory to be fully meaningful even where the investigator declines belief in its unobservable ontology. It places an epistemic decision between the successful use of a theory and the affirmation of all entities it describes. Mohler and Monton's analysis of constructive empiricism (2025 revision) emphasizes how much of the ensuing debate concerns the exact content of that decision.

The leading question of this dossier is whether empirical adequacy provides a defensible stopping point for scientific belief. Answering it requires more than citing the familiar observable–unobservable distinction. We must determine what empirical adequacy asserts, which observations are relevant, how observability is established, and what a realist may infer from predictive and experimental success. The problem remains philosophical even when every measurement in a particular experiment is agreed upon.

Research Method and Evidential Scope

The investigation combines conceptual reconstruction, historical comparison, argument analysis, and explicitly constructed cases. The founding book provides the primary formulation. Subsequent articles disclose criticisms and replies; publisher records establish bibliographic identity; the later books develop the relation between empiricism, perspective, and philosophical stance. Each assertion about an author's position is bounded by the argument and publication actually checked. The original publisher's account of the book (van Fraassen 1980) establishes the central claims, while the Stanford Encyclopedia's scholarly reconstruction (Mohler and Monton 2025) supplies a map of the debate.

Several classic books and journal papers are accessible through publisher abstracts, tables of contents, bibliographic metadata, open scholarly excerpts, and replies rather than unrestricted complete editions. Where that is the case, conclusions are restricted to claims those materials sustain. The 2003 response by Monton and van Fraassen is additionally available through an author-hosted original text; by contrast, the full 2026 Van Dyck article was not established as openly readable and its abstract supports the analysis of its stated thesis, not a page-by-page exegesis. EVIDENCE.json records these distinctions source by source.

This is a philosophical monograph, not a survey of experimental results or a new empirical study. The three cases explicitly designated as constructed examine how commitments differ under specified conditions. No observed frequency, data set, participant testimony, or historical incident is inferred from them. Authorial claims in the synthesis are arguments formulated here on the basis of examined premises, rather than claims that they have already been independently confirmed by disciplinary consensus.

The Historical Problem of Realism

Scientific realism typically includes a semantic claim that well-formed theories have truth values and an epistemic claim that successful, mature theories deserve belief in at least some claims about unobservables. These dimensions have often been bundled together. Constructive empiricism separates them. An investigator can use the realist's literal language while resisting the realist's inference from success to approximate truth about hidden structure. This separation created a distinctive place in twentieth-century philosophy of science for anti-realism without a verificationist theory of meaning.

In The Scientific Image, van Fraassen (1980) explicitly positions his alternative against both logical positivism and scientific realism. Logical empiricist strategies commonly sought to analyze theoretical significance through observational vocabulary and rules of correspondence; scientific realists challenged restrictions that appeared unable to capture the knowledge supplied by theoretical science. Van Fraassen instead preserves the intelligibility of theoretical discourse while questioning what belief should accompany acceptance. This historical intervention changed the burden of argument for both sides.

Realism continues to offer substantive reasons for its commitments. Psillos's defense of scientific realism (1999) argues for explanations of success that reach beyond phenomenological fit. Conversely, the earlier historical challenge in Laudan's critique of convergent realism (1981) asks whether successful science has always vindicated its former theoretical ontologies. These arguments do not decide constructive empiricism by themselves; they establish the opposed costs that an acceptable interpretation of scientific practice must address.

Van Fraassen's Foundational Thesis

Constructive empiricism comprises two closely related propositions. Its account of the aim of science is that scientific inquiry seeks empirically adequate theories. Its account of theory acceptance is that acceptance requires belief in empirical adequacy, accompanied by other practical and methodological commitments, without requiring belief in the complete literal truth of the theory. The difference between an aim and an achievement matters: science may pursue adequate theories even when it selects one that later proves inadequate.

Van Fraassen's 1980 book description and contents make both propositions explicit and give separate chapters to saving the phenomena, methodology, and the pragmatics of explanation. A theory's being accepted establishes neither its actual empirical adequacy nor the literal truth of all claims in its models. Instead, acceptance specifies the commitment undertaken by one who judges the available evidence sufficient to regard it as empirically adequate. This is why the doctrine should be formulated as a position about the aims and epistemic commitments of science, rather than a declaration that unobservable entities are fictitious.

A point of interpretation follows. The conjunction of the two propositions is coherent only if the operative meaning of empirical adequacy is stable enough to guide acceptance. Without some account of what counts as an observable phenomenon and how it is represented, the doctrine would identify the permitted belief by an undefined target. The semantic account of models and the analysis of observability are therefore parts of its philosophical substance, not optional technical appendices.

Literal Interpretation and Limited Belief

A literal statement about an electron can be true or false whether or not its user believes electrons exist. Constructive empiricism retains that semantic possibility. The refusal to infer belief from acceptance is an epistemic choice about warranted commitment; it does not rewrite all theoretical assertions into claims about instrument readings. Van Fraassen's account of scientific language (1980) makes literal interpretation central precisely because an anti-realist alternative should remain capable of understanding the very theories that realists take to be true.

This distinction prevents two common misreadings. First, constructive empiricism is not the claim that scientific theories have no meaning beyond their observable consequences. Second, it is not a positive judgment that every unobservable entity is nonexistent. An agnostic who withholds commitment from the existence of a particular theoretical object can still understand and test statements describing it. The 1985 collection of realism debates, edited by Churchland and Hooker, documents the historical intensity with which the status of observables and theoretical entities was contested.

The analytically important category is the permissible gap between intelligible assertion and rational belief. Scientific practice includes countless legitimate uses of a theory—calculating with it, making predictions, setting instruments, teaching its structure—without logically requiring all users to endorse every ontological claim it contains. What remains contested is whether the best explanation of those uses eventually warrants the stronger realist commitment.

The Aim of Science and the Aims of Scientists

The expression “science aims” can refer to a constitutive interpretation of the institution, a normative standard imposed upon researchers, or an aggregation of what individual scientists want. These readings yield different tests. Some scientists seek explanation, technological control, prediction, intervention, or a stable income. The collective activity of science can nevertheless be interpreted through a more general aim without being reducible to the personal goals of every participant.

Van Fraassen's foundational exposition (1980) identifies empirical adequacy as science's aim. The interpretation faces an immediate pluralism problem: does science have only one end? Contemporary discussion, including the 2024 volume on scientific stances edited by Beisbart and Frauchiger, permits closer consideration of nonexclusive aims. Greater scope, reliability, explanatory organization, and experimental fruitfulness may structure scientific inquiry while empirical adequacy remains the criterion that fixes its empiricist achievement.

A nonexclusive formulation avoids the implausible claim that seeking additional observations, improving instruments, or designing interventions is irrational whenever one accepted theory already accommodates available observations. Yet it must also keep the position discriminating: if every conceivable scientific goal is counted without ordering, empirical adequacy ceases to perform a distinctive philosophical function. The successful interpretation specifies which aims are epistemic targets, which are practical or institutional, and how conflicts between them can be assessed.

The Semantic Conception of Theories

On a syntactic conception, a theory is commonly approached through a set of propositions and derivable sentences. The semantic conception characterizes a theory through a class of mathematical or structural models. This shifts philosophical attention from the vocabulary in which a theory is expressed to the structures it makes available for representing phenomena. The same theory can then admit differing linguistic formulations while preserving an underlying family of models.

Van Fraassen's semantic treatment of theories (1980) specifies empirical substructures within models as candidates for the direct representation of appearances. In the classic reconstruction, empirical adequacy obtains when appearances can be embedded in the empirical substructures of some model of the theory, often described through isomorphism. The formulation is stronger than the assertion that one finite table of predictions has passed a test. It claims a structural fit between the relevant phenomena and the representations made available by the theory.

An objection arises where actual scientific representations are partial, approximate, idealized, and adjusted to measurement uncertainty. A rigid interpretation of exact isomorphism may fail to describe routine modeling practice. The later philosophy of scientific representation (van Fraassen 2008) explores perspective and representation in ways that sharpen this issue. The present analysis treats exact structural correspondence as a foundational idealization whose operational application must be argued for each modeling practice, not silently presumed.

Empirical Adequacy as a Substantive Claim

Empirical adequacy means that a theory gets observable phenomena right. That apparently plain statement conceals an ambitious claim about phenomena that may occur beyond the present investigator's place and time. A theory can fit the current observations and still be empirically inadequate because a possible or future observable phenomenon would violate its predictions. Thus acceptance involves a fallible expectation about an indefinitely larger domain of cases than has been inspected.

The exposition of empirical adequacy by Mohler and Monton (2025) stresses the distinction between observable and already observed phenomena, as well as the structural formulation introduced by van Fraassen. An account that equates empirical adequacy with agreement on currently gathered data would misstate the doctrine. It would also make testing irrelevant to unencountered observable cases. Constructive empiricism has a real predictive commitment precisely because it expects the empirical content of an accepted theory to withstand further observable inquiry.

The epistemic modesty of the doctrine is therefore selective rather than absolute. It distinguishes domains of commitment by kind, not simply by temporal proximity to evidence. Belief in adequacy may exceed all present observations; belief in unobservable structure may be withheld even after an abundance of successful tests. Whether this division is justified depends upon the rationale for privileging observable phenomena as the domain in which scientific belief should be concentrated.

Observable, Observed, and Unobservable

Three categories must remain analytically distinct. Observed phenomena have actually entered some evidential record. Observable phenomena could be observed by appropriate members of the relevant epistemic community under suitable conditions, whether or not that encounter has occurred. Unobservable phenomena are outside that community's relevant observational capacities. The boundaries can be vague while clear instances remain sufficient for making philosophical distinctions.

The account of observability and community (Mohler and Monton 2025) reconstructs the claim that an object's being observable is an empirical matter concerning both the object and observers. A distant moon might be observable in principle under suitable conditions even when no one has actually seen it. The microscopic constituent of a specimen can remain unobservable in the disputed sense despite an instrument's ability to produce a detailed image. The difference turns upon what counts as observation of the object, not merely the presence of perceptual experience associated with it.

The category should never be substituted for the expression “directly seen” without analysis. Eyesight is one observational modality among several, and instruments can reveal both immediately encountered phenomena and causal signs of inaccessible structures. The doctrine's historical debates over microscopes arise because the relation between an instrument image and its target is itself a scientific and philosophical achievement. A philosophical boundary must be sufficiently stable for argument and sufficiently attentive to the actual capabilities of investigators.

Epistemic Communities and Observation

The qualifier “observable to us” introduces a reference to observers with particular perceptual capacities, bodily structures, and methods of inquiry. It also introduces questions of membership. If the relevant community contains observers with different capacities, its observational domain may differ from that of a homogeneous population. Van Fraassen's discussion of scientific methodology (1980) treats observability as accessible to investigation rather than fixed through a purely a priori conceptual inventory.

The philosophical advantage of a community-relative criterion is its connection to actual natural capacities. The difficulty is the determination of the relevant community. A scientist's laboratory may include humans, animals, instruments, and automated monitoring systems, although their contribution to the evidential process need not make each one an observer in the same sense. If a new kind of participant expands the community's range of unaided perception, the set of phenomena that a theory must accommodate may expand. If the community is fixed to a historical human population, the criterion retains continuity but risks becoming an arbitrary restriction for future science.

The 2024 discussion of expanded observation collected by Beisbart and Frauchiger presses this issue. The question is not whether researchers may choose whatever observational standard flatters a favored theory. It is how to specify a nonarbitrary relationship between empirical capacities and the scope of the adequacy claim. A doctrine rooted in empirical inquiry owes an account of the evidential discovery of observers' capacities and of how scientific communities accommodate their change.

The Unobserved Portion of Empirical Adequacy

Constructive empiricism resists belief in inaccessible theoretical entities while allowing belief that a theory gets all observable phenomena right. That belief covers cases for which evidence has not yet been collected. The rational commitment is therefore inductive. Its object is a generalized proposition about the world, not merely a report about past observation. An empiricist can acknowledge this without inconsistency, provided the boundary between allowed induction and rejected ontological inference has an argued rationale.

Gideon Rosen's analysis of constructive empiricism (1994) is a central contribution to debates over what the position actually requires, and Alspector-Kelly's criticism of its empiricist rationale (2001) tests whether the move from empiricism to constructive empiricism is compelled. The criticism is substantive: a policy that believes unobserved observable predictions must explain why a further inference to successful unobservable mechanisms has a categorically different epistemic status. Merely labeling the latter “metaphysical” would restate the intended conclusion.

The most defensible reply appeals to the character of the scientific aim and the chosen distribution of risk. Belief in empirical adequacy is a commitment to the success of the kind of representation for which the theory is scientifically tested. Belief in hidden ontology is a stronger interpretation of what explains that success. This is a reason for a distinction, though realists may reasonably object that it does not establish a unique rule of rational belief. The empirical adequacy claim is modest by comparison with total theoretical truth, not free of epistemic reach.

Acceptance, Belief, and Commitment

Theory acceptance has several aspects. The investigator may believe the theory is empirically adequate, use its models as guides to experiment, take its terminology seriously, adopt associated conventions, and participate in its research program. These acts have practical consequences even when the investigator declines commitment to each unobservable object in the theory. Van Fraassen's pragmatic account of acceptance (1980) distinguishes this broader institutional engagement from bare propositional assent.

This framework makes sense of scientists who accept idealized models knowing that their assumptions are deliberately simplified. It also avoids an inference from the visible fact of scientific cooperation to a hidden unanimity of belief. Two members of a laboratory can perform identical calculations and report the same results while holding different philosophical views about the mechanisms named by the equations. The success of their research program is a fact about practice; the extension of each participant's belief must be determined separately.

A complication follows when “acceptance” is understood operationally as willingness to act on a theory in high-stakes situations. Action may require pragmatic confidence that exceeds or differs from the categorical belief the doctrine officially permits. Constructive empiricism has the resources to distinguish acting as if a theory were true from believing it true, but each application should say which risks, costs, and possible revisions structure that decision. Without this clarification, acceptance may be verbally separated from belief while doing identical work beneath another name.

Probability, Confidence, and Acceptance

Scientific confidence often comes in degrees. A hypothesis may be strongly supported without attaining certainty, and an accepted theory may have predictive reliability within specified domains without proving a universal generalization. The relation between partial confidence and the more categorical language of acceptance matters for constructive empiricism, because the belief in empirical adequacy can itself be fallible. The doctrine offers a proposed content of warranted commitment; it does not provide a universal numerical threshold at which the commitment becomes obligatory.

Van Fraassen's discussion of probability in science (1980) includes probability among the fundamental philosophical topics of the book. Its later development in Laws and Symmetry (van Fraassen 1989) connects probability, laws, and methodological analysis. Taken together, these sources place the acceptance problem among broader questions of scientific judgment rather than treating it as a single logical implication from test results.

The most useful distinction concerns the object and degree of confidence. A researcher may assign high confidence to a measurement's stability, moderate confidence to a model's generalization, and lower confidence to the reality of a mechanism invoked in the model. These are different propositions and deserve different assessments. The philosophical disagreement begins when realists claim that substantial empirical support, independently reproduced intervention, and explanatory integration justify stronger ontological confidence than a constructive empiricist policy admits.

Explanation and Pragmatic Context

Why-questions are asked from within interests, background assumptions, and contrasts. Explaining why an outcome occurred rather than an alternative outcome requires choosing which contrast is relevant, and the choice depends partly upon the inquiry's situation. Van Fraassen's pragmatics of explanation (1980) treats this contextual structure as central to understanding explanatory requests. A theory may generate successful explanations without explanatory success constituting proof that every entity invoked is real.

This view distinguishes explanation as a scientific achievement from explanation as an automatic license for ontological belief. A meteorological model can explain a local pattern by locating it within a larger system of equations and conditions. The model's success may be assessed against observations and useful counterfactuals while the nature and number of its representational components remain open to further interpretation. The realist can accept explanatory context-dependence yet maintain that systematic explanatory power provides evidence for the approximate truth of particular hidden structures.

The disagreement therefore concerns the epistemic force of explanation, not its scientific value. A constructive empiricist can seek elegant, unified, and highly informative explanations because such explanations organize the search for empirical adequacy and make a model useful. What requires additional defense is the transition from their usefulness or success to belief in ontology. The position's philosophical strength depends upon providing a coherent account of explanation that retains this distinction without depreciating explanations as merely rhetorical devices.

Inference to the Best Explanation

Inference to the best explanation proposes that the hypothesis best explaining evidence deserves an elevated degree of belief. Its application in scientific realism often takes the form of an inference from predictive, experimental, and technological success to the approximate truth of the mechanisms described by successful theories. Constructive empiricism asks whether selecting the best available explanation warrants belief that it is true, given the possibility that future hypotheses may surpass it and that evidentially equivalent explanations remain available.

Psillos's systematic defense of realist inference (1999) develops an account in which theoretical success can track truth. Van Fraassen's arguments about scientific realism (1980) challenge the move from explanatory preference to required belief. Both parties may acknowledge that scientists rank models according to explanatory virtues. The philosophical difference is whether those virtues are epistemically truth-conducive enough to determine an investigator's beliefs about unobservable structure.

The phrase “best explanation” can also disguise an impoverished comparison class. A hypothesis may be the strongest among those currently articulated while the range of possible alternatives remains unknown. Its success warrants serious consideration and may justify action; it does not logically entail truth. Realist reasoning becomes stronger when its explanatory considerations are joined with independently discriminating tests, successful novel predictions, and causal intervention. The constructive empiricist remains entitled to ask which of those considerations crosses the boundary from evidence about phenomena to belief about unseen existence.

The No-Miracles Argument

The no-miracles argument holds that the sustained empirical success of science would be surprising unless well-supported theories were approximately true, or unless central parts of their theoretical descriptions genuinely captured the world. In this view, realism explains the systematic success that anti-realists accept as a fact. The argument is abductive rather than deductive: a successful explanation of success is offered as a reason for belief, not as a formal demonstration that every alternative is impossible.

The argumentative terrain is evident in van Fraassen's critique of realism (1980) and Psillos's realist reconstruction of success (1999). A selection-based anti-realist reply notes that theories routinely become prominent because they survive empirical tests, so finding surviving theories successful is partly a consequence of that filtering. This reply can undercut a simplistic wonder at success while leaving a further question: why do particular theories achieve remarkable novel successes outside the data used to construct them?

The result is a genuine contest between explanatory strategies. Selection explains why an institutional archive is populated by locally successful theories; realism offers a hypothesis about why certain theories continue to succeed under novel and varied challenges. Neither argument becomes conclusive by definition. Their comparative strength depends on historical selection effects, independent tests, theoretical continuity, and the realism of the proposed explanatory inference itself. Constructive empiricism is most persuasive where it acknowledges the phenomenon to be explained rather than treating the absence of a deduction as a complete answer.

Underdetermination and Rival Theories

Distinct theoretical structures can agree with the same body of observations. This underdetermination is central to constructive empiricism because empirical adequacy may be shared by rival accounts whose unobservable ontologies conflict. If acceptance requires belief only in the agreement of a theory's empirical claims with observable phenomena, the existence of multiple compatible accounts need not force a choice among ontological inventories. Scientific practice can still rank the rivals for explanatory, computational, pragmatic, and investigative purposes.

The discussion of empiricism and evidence by Mohler and Monton (2025) records underdetermination among the principal themes. Rosen's 1994 analysis of the position explores difficult questions about its content and defense. A realist may reply that evidence consists of more than static observable consequences: manipulability, consilience, independent methods, and predictive novelty often discriminate between rivals. This point weakens crude claims of universal equivalence but leaves open the broader possibility of empirically adequate theories with divergent ontologies.

The relevant conclusion is bounded. Actual equivalence has to be argued for a specified pair of theories, observational conditions, and range of phenomena; it is not licensed by the abstract possibility of underdetermination. Nor does the fact that a theory has multiple models establish that all model ontologies are equally warranted. Constructive empiricism gives a response to an epistemic situation that can arise, while empirical research and conceptual analysis determine when a specific situation actually meets its premises.

Historical Success and Failed Ontologies

The history of science supplies a challenge to global inferences from success to truth. Some once-successful theories included mechanisms and objects subsequently rejected. Larry Laudan's historical criticism of convergent realism (1981) makes this difficulty explicit. The historical record suggests that empirical achievements can survive the abandonment of explanatory structures previously treated as central. Constructive empiricism interprets such episodes as reasons to distinguish well-established empirical claims from more ambitious ontological commitments.

A strong realist response distinguishes the parts of a theory directly responsible for its novel successes from speculative additions. Psillos's defense of selective realism (1999) offers resources for this strategy. The response is philosophically serious only when its criteria for preserving reliable theoretical content can operate prospectively. If the realist identifies the trustworthy elements solely by looking backward after the history is known, the selection risks becoming an explanation of whatever happened rather than a prediction of which commitments were justified.

A parallel responsibility falls upon constructive empiricism. The empirical claims of past theories were also sometimes rejected, and believing in their empirical adequacy was sometimes mistaken. Historical caution does not establish that the permitted empiricist belief is invulnerable. It establishes the fallibility of scientific acceptance and invites comparison of the consequences of different forms of commitment. The doctrine's value lies in specifying where it locates epistemic restraint, rather than in offering immunity from historical correction.

Observation, Mediation, and Scientific Images

Contemporary observation is frequently mediated by lenses, detectors, algorithmic reconstructions, and calibrated instruments. An image on a screen can be directly seen while its alleged source remains a theoretical interpretation of how the instrument operates. Constructive empiricism's distinction between perceptual access and inferential attribution gains practical importance here. The disagreement is whether instrument-assisted encounters can count as observations of the entities represented and, if so, what causal and representational conditions make that attribution reliable.

Van Fraassen's work on scientific representation (2008) emphasizes the perspectives through which representations are produced and interpreted. Kusch's study of microscope experience (2015) addresses modifications to van Fraassen's agnosticism about the veridicality of microscopic observation. This discussion does not entail that images are arbitrary or that investigators are unable to establish reliable correspondences. It shows that the image-object relation depends on more than an unexamined analogy between a lens and unaided sight.

Scientific practice often brings convergent evidence from instruments operating on different principles. Such convergence can make it reasonable to trust an image-based identification, and realist arguments may treat it as support for the existence of the target. The constructive empiricist can accept the convergence as evidence of instrument reliability while questioning whether it mandates the strongest ontological inference. The two positions differ most sharply when the evidential success of a representation is taken to settle the nature of the represented object.

Microscopes and the Limits of Seeing

Microscopes provide one of the most persistent challenges to an unaided-observation boundary. When an investigator looks into a microscope and reports seeing a cell, the report is part of ordinary laboratory language. The philosophical question is whether this constitutes observation of the cell in the sense relevant to constructive empiricism or observation of an optical image that gives evidence about the cell. Both descriptions can play roles in scientific practice without being equivalent as philosophical accounts of perceptual access.

The classic debates collected in Images of Science (Churchland and Hooker, eds., 1985) contain foundational exchanges over observable entities and microscope experience. Kusch's 2015 reassessment examines how van Fraassen's later position refines the issue, notably through theory-ladenness and the veridicality of perceptual experience. The detailed arguments in the edited collection are interpreted through its established contribution history and later scholarly discussion, not through an assumed complete inspection of every chapter.

The important result is an analytical separation among the reliability of the instrument, the status of the image, and the ontology of the target. A microscope may be exceptionally reliable for practical identification without its image constituting unmediated perception in the philosophical sense. Conversely, skepticism about the literal identification of seeing through a microscope should not lead to the claim that microscopic entities are fictional. Constructive empiricism's agnosticism concerns what belief acceptance requires, and its power depends on retaining that precision.

Intervention and Experimental Realism

Experimental realism argues that manipulation can provide reasons to believe in entities beyond what unaided observation presents. If an experimental arrangement allows investigators reliably to use a particular causal structure to produce and control outcomes, then the entity or process postulated to support the manipulation acquires a distinct evidential status. Ian Hacking's study of representing and intervening (1983) exemplifies this line of reasoning. The methodological emphasis moves from fitting a representational structure to doing something robust with the world.

The argumentative force depends on whether the success of intervention requires a realist interpretation of the manipulated object. A constructive empiricist can acknowledge interventions, successful control, and regular production of observable outcomes. The realist maintains that such success is evidence for specific causal structures; the anti-realist replies that practical mastery and warranted ontological belief remain different conclusions. The disagreement becomes more discriminating where independently designed interventions manipulate the same inferred entity under sharply different conditions.

An adequate assessment asks for the complete inferential path. What exactly was changed in the experiment? Which effects were observed? Which assumptions permit attribution of those effects to a postulated entity? Were competing causal accounts actually excluded? Intervention has a stronger evidential profile than passive curve fitting in many contexts, but its philosophical significance must be argued, not treated as self-authenticating. The observational agreement can be shared even where the inferential interpretation remains disputed.

The Modality Objection

Observability is often characterized by circumstances under which an observer would detect a target if it were present. Such counterfactual language introduces a difficulty for an empiricism wary of objective modality. If it is necessary to believe that an object would be observed under conditions that are not actual, is constructive empiricism committed to modal facts of the kind it seeks to avoid? James Ladyman's 2000 critique of modality argues that constructive empiricism requires an objective modal distinction between observable and unobservable phenomena.

The force of this objection is internal. It accepts much of the original position for the sake of argument and contends that the position's essential criterion imports commitments at odds with its stated empiricist motivation. Since empirical adequacy is defined by observability, the objection threatens more than a peripheral example: it questions whether the content of the permitted belief is intelligible without the contested modal structure. A claim about a merely possible observational encounter can be used to fix a real-world distinction only if the content of that conditional is appropriately grounded.

The objection does not establish that all counterfactual language presupposes a metaphysically heavy ontology. Its success depends on whether the constructive empiricist can recover ordinary empirical judgments about capacities and circumstances in a form that gives observability objective content without treating every counterfactual as a law of metaphysical possibility. This is the precise question addressed in the subsequent direct reply.

The Reply Concerning Modal Nominalism

Monton and van Fraassen's response on modal nominalism (2003) argues that the property of being observable can be objective and nonmodal even if ordinary analyses refer to counterfactual circumstances. Their discussion treats observability as a feature of how objects and observers are actually constituted. Questions about whether a target could be observed under appropriate circumstances can be investigated through empirical claims about those constitutions. The authors further argue that even a stronger modal implication would not automatically refute the epistemic position.

This reply makes the scientific determination of observational capacities central. It shifts the question from the bare logical form of a sentence to the worldly facts on which a successful judgment depends. Scientific theories can inform investigators about the capacities of organisms and the properties of objects; this does not automatically oblige an investigator to believe every associated theoretical description. The challenge of selective reliance upon those theories remains, but the modal objection has to meet the reply at its strongest level.

Ladyman's 2004 continuation of the debate preserves the controversy. The original 2003 reply is available in an author-hosted full text (Monton and van Fraassen 2003), permitting more direct verification of the argument than publisher abstracts alone. The interpretive result is that constructive empiricism can defend observability as an empirically investigable relation while still needing a precise account of how hypothetical encounters constrain present belief.

Theory-Ladenness and Circularity

To decide which objects are observable, investigators often rely on theories of optics, physiology, instruments, or the environment. But those theories may themselves be accepted only because they are thought empirically adequate. A critic can therefore ask whether the criterion for adequacy is partly determined by a theory whose adequacy is being judged. This risk of circularity is significant because it might allow a favored theory to draw the observable boundary in a way that excludes inconvenient phenomena.

The analysis of the hermeneutic circle (Mohler and Monton 2025) identifies both circularity and relativity objections. Kusch's work on theory-laden perception (2015) shows how the problem is sharpened by microscope experience. The fact that scientific inquiry contributes to identifying observational capacities is appropriate for an empiricist. What requires discipline is the independence of the evidence used to support the capacity judgment.

One response is methodological rather than metaphysical: determine observational capacities through convergent studies, disclose background assumptions, and allow adverse findings to revise both a theory and the boundary used to judge it. This does not remove the conceptual interdependence of science and observability. It reduces the danger that the interdependence becomes an evidentially unconstrained circle. A defender of constructive empiricism can regard the boundary as revisable while maintaining that its revisions are subject to empirical accountability.

Data, Phenomena, and Representation

Raw traces, processed data, stable phenomena, and theoretical representations belong to distinct levels of scientific inference. A noisy set of laboratory readings may provide evidence for a systematic phenomenon, which in turn may be represented by a model containing unobservable parameters. Calling all three “observations” disguises the inferential work required to move between them. Bogen and Woodward's analysis of data and phenomena (1988) makes this distinction foundational to the philosophy of scientific evidence.

The implications for constructive empiricism are substantial. Empirical adequacy is measured against phenomena and their empirically accessible manifestations, while actual tests compare model outputs against data that have undergone calibration, reduction, filtering, and interpretation. A model can fit a data set through statistical flexibility without reliably describing the underlying phenomenon. Conversely, an observational phenomenon may be robust even when a particular model of its unobservable mechanism is rejected. Treating data-model fit as the complete content of empirical adequacy would obscure these possibilities.

The correct methodological question is whether the observational structure attributed to a phenomenon is stable under the independently justified procedures used to establish it. This requirement is compatible with constructive empiricism and with realism. It also disciplines the dispute: the parties should agree about what evidence has actually established before drawing divergent conclusions about what it licenses them to believe. Empirical adequacy cannot be made more precise by bypassing the epistemic work that produces reliable descriptions of phenomena.

Constructed Case: Two Rival Models

Consider two hypothetical models, A and B, designed to describe the same sequence of observable measurements. Both yield the same predicted distribution within the measurement resolution across all currently feasible experiments. Model A posits an inaccessible oscillating field with a particular microstructure; model B posits a different hidden mechanism. Assume, for the sake of this constructed case, that their empirical predictions remain equivalent throughout the relevant observable domain and that their hidden commitments conflict. No actual scientific theory or observed experiment is being reported.

A constructive empiricist can accept either model as empirically adequate while withholding belief in its hidden mechanism. A realist has a reason to pursue explanatory and experimental distinctions that could eventually separate them, but the stipulation rules out a present observational choice. Psillos's realist treatment of theoretical inference (1999) indicates why realists resist treating empirical equivalence as the end of rational theory comparison, whereas van Fraassen's account of the empirical aim (1980) clarifies what acceptance need commit us to when the observable fit is secured.

The case reveals both the strength and the limits of the position. Constructive empiricism avoids choosing an ontology without discriminating evidence and permits continued scientific work. Yet equivalence is stipulated rather than demonstrated. In an actual research situation, new instruments, interventions, auxiliary hypotheses, or background evidence may distinguish the models. Consequently the case supports an argument about the logical relation between evidence and commitment, not a general empirical assertion that competing mature theories are always observationally indistinguishable.

Constructed Case: A Microscopic Image

Imagine a laboratory in which an instrument presents a recurring high-resolution image when a specimen is prepared in a specified manner. The image has stable features, and researchers using independently calibrated instruments can reproduce those features. The laboratory describes the image as showing a particular microscopic entity. The stipulations establish robust repeatability, not a historical finding or an actual analysis of a named specimen.

The constructive empiricist can affirm that the image reliably appears and that the procedures support predictions about further images. Whether that acceptance must include belief that the depicted microscopic entity exists is the live philosophical question. The analysis in Kusch's microscopy and experience study (2015) explains why instrument-image veridicality is disputed, while Hacking's argument from experimental intervention (1983) offers a possible realist route beyond visual evidence alone.

Now add a second stipulation: the apparatus can manipulate a predicted causal property of the entity to produce new, independently observable effects. The realist has a stronger and more varied evidential argument. The constructive empiricist can still distinguish that success from a mandate to believe in the hidden ontology, but the cost of such restraint becomes clearer. The case establishes that the philosophical stakes rise as evidence becomes more independently constraining; it does not decide where a universally required threshold of ontological commitment lies.

Constructed Case: A Predictive Instrument

Suppose a future observational instrument extends a research community's capacity to register a phenomenon that previous members could not encounter without elaborate mediation. After training and calibration, the new instrument produces stable outputs and prospective predictions. Three issues must be distinguished: whether the outputs are trustworthy, whether they amount to observation of their alleged target, and whether using them changes what the community counts as observable. The example is explicitly a conceptual construction, without a claim about any existing machine's perceptual status.

The 2024 volume on scientific theories and stances includes discussion of expanded powers of observation and the significance of new community members. Monton and van Fraassen's 2003 discussion of observational capacities establishes why the object's constitution and the observer's capabilities are empirical matters. One can therefore evaluate a new instrument's trustworthiness without immediately deciding whether its output constitutes direct observation for the philosophical purpose at issue.

The constructed case displays a potential tension in the aim of empirical adequacy. An accepted theory may fare well under an existing boundary and poorly under an expanded one. Scientific progress can increase the demands imposed upon a theory's empirical claims. That outcome is consistent with the fallibility of acceptance. It also creates a reason to resist defining observability merely so as to preserve current theoretical success. A disciplined community must allow its observational conditions to become more demanding as inquiry develops.

The Empirical Stance

Van Fraassen's later work situates empiricism within a stance—a constellation of attitudes, values, approaches, and commitments—rather than presenting every empiricist as bound by a fixed set of propositions. In The Empirical Stance (2002), empiricism is associated with attention to experience, criticism of metaphysical ambition, and a voluntarist approach to rational belief. This broadening matters because critics sometimes treat constructive empiricism as a logical consequence that all rational investigators must accept. A stance-based conception instead allows an investigator to defend a disciplined epistemic policy without claiming that all alternatives are formally irrational.

The price of that flexibility must be acknowledged. If a stance is partly a choice of epistemic orientation, arguments about constructive empiricism may demonstrate coherence, fruitfulness, or fidelity to certain scientific practices without compelling every reasonable realist to adopt it. Alspector-Kelly's challenge to an obligatory empiricism (2001) can be read as pressing this distinction between a possible position and a uniquely required one. The distinction can clarify debate rather than weaken it: different participants must say whether they are defending permission, rational preference, or necessity.

A related consequence concerns pluralism within science. A single research institution can include investigators who hold realist and empiricist stances yet collaborate on rigorous measurements and models. Their practical agreement establishes a public evidential process, while their disagreement concerns the reach of warranted belief. A philosophical interpretation can examine the norms of that collaboration without attributing a uniform private metaphysical commitment to every participant.

Scientific Representation and Perspective

A representation presents a target through an organized perspective: it selects variables, establishes scales, makes idealizations, and presupposes conventions of measurement and depiction. The relation between representation and world is accordingly an achievement of practice, not a transparent correspondence guaranteed by a diagram. Van Fraassen's Scientific Representation: Paradoxes of Perspective (2008) investigates problems arising from this feature of science, including the mediation through which theories and data become mutually interpretable.

This later perspective extends constructive empiricism beyond the earliest formulation of the observability controversy. The investigator can ask what is being represented, by which procedures, and under what assumptions, before deciding what the model's success warrants. Kusch's analysis of the later view (2015) emphasizes changes in the approach to microscopic observation. The issue is not solved by saying that all representations are perspectival, since perspectival models may still make informative and accurate claims about targets.

The deeper point is that the evidence of adequacy comes through arrangements whose reliability must be established. A representation can be adequate relative to a particular empirical domain while retaining idealizations or theoretical components not warranted by those observations alone. The apparent strength of the constructive empiricist distinction here rests on a clear inventory of representational operations. Without such an inventory, the separation between stable empirical content and unobservable interpretation can become too abstract to guide the assessment of particular scientific theories.

Augmented Observation and the 2024 Challenge

A theory's empirical adequacy is normally assessed relative to what its scientific community can observe. An expanded community may acquire capacities that expose phenomena outside the original domain. Finnur Dellsén's 2024 chapter, “Expanding the Empirical Realm: Constructive Empiricism and Augmented Observation,” raises a challenge about scientists' incentives to increase their observational powers when doing so could reveal that an accepted theory is empirically inadequate. The chapter belongs to the edited collection on scientific stances (Beisbart and Frauchiger, eds., 2024), and its central argument is examined here through the published scholarly overview and the documented reply.

The 2025 Stanford Encyclopedia analysis (Mohler and Monton 2025) summarizes van Fraassen's response in the same 2024 volume. Science may have more than one aim; increased observational reach can be pursued for completeness and other goods even if present theories face greater risks. The response also asks how a community should deal with new members whose observations can or cannot be reproduced by old members' instruments, and whose results can or cannot be accommodated by a theory. This analysis is more specific than the claim that better observation is always advantageous to a currently accepted theory.

The philosophical significance is substantial. A rational scientist can seek more demanding evidence even while expecting some accepted theories to fail. Scientific inquiry achieves progress partly by revising the standards and reach of empirical testing. An interpretation of science that rendered such improvement conceptually counterproductive would be defective. The nonexclusive-aim response preserves constructive empiricism's central commitment but should be accompanied by a clear distinction between protecting the empirical adequacy of one theory and improving science's capacity to assess any theory.

Van Dyck's 2026 Reconstruction

Maarten Van Dyck's 2026 defense of constructive empiricism supplies a recent reconstruction of the positive argument. Published online on August 19, 2026, in HOPOS, the article distinguishes the collective, historically instituted aim of scientific practice from the epistemic policy adopted by particular individuals. Its abstract argues that constructive empiricism offers a compelling interpretation of scientific activity, while each investigator remains responsible for the stance taken toward that activity. This is the accessible, verified basis for the discussion; a complete page-by-page inspection of the article was not established.

This distinction illuminates a persistent ambiguity in debates about the doctrine. An institutional description can be historically adequate even if not every participant avows it as a philosophical creed. Conversely, an individual's permission to adopt empirical adequacy as a limit of belief is not automatically established by showing that researchers collectively seek successful representations. Van Dyck's argument, as described by the publisher's article record (2026), asks us to analyze these levels separately before judging the force of the defense.

An independent consequence follows: the most serious disagreement between realism and constructive empiricism is not settled by counting realist scientists or by finding experiments that use realist terminology. The evidence concerns collective patterns of theory testing, historical interpretation, and individual justification. A realist may agree that empirical adequacy is a major public aim while maintaining that explanatory and interventional success warrants additional belief. The empiricist may regard a more restrained stance as rationally available. Understanding the position requires adjudicating both claims, rather than replacing one with the other.

Comparative Assessment of the Principal Objections

The objections do not all target the same feature of constructive empiricism. The realist's no-miracles argument challenges the restraint of its epistemic policy. Rosen's analysis of the doctrine (1994) and Alspector-Kelly's critique of empiricist justification (2001) press the rational basis for accepting empirical adequacy as the relevant limit. Ladyman's modality objection (2000) targets the coherence of the observability criterion. The microscopy debates examine a boundary in actual scientific representation, while augmented observation questions how that boundary changes under scientific innovation.

These challenges should be ranked by the work they require of a defense. The charge that constructive empiricism denies the existence of electrons can be resolved by correctly reconstructing its agnosticism. The claim that it cannot value explanation can be answered through the distinction between explanatory achievement and epistemic compulsion. The modalities and expanding-observation problems are harder because they concern the definition and stability of the permitted belief itself. A defense cannot merely repeat that acceptance involves empirical adequacy when the intelligibility of that category is what the objection questions.

Nor do the criticisms automatically vindicate realism. A realist must specify which success-generating theoretical commitments are justified, how historical mistakes are accommodated, and which tests distinguish the ontological hypotheses being asserted. In that sense, the debate imposes reciprocal explanatory burdens. Constructive empiricism must show that its restraint has a principled basis, while realism must show why the ampliative step from observable success to unobservable reality is warranted in the cases it cites.

Changing Communities and Nonbiological Inquiry

The debate concerning observation acquires a broader historical horizon when the participants in scientific inquiry change. Biological observers already differ in sensory capacities. Automated systems collect traces, classify patterns, produce model outputs, and sometimes operate without direct human intervention. A philosophical analysis can ask how such systems alter the epistemic community, which data count as observations, and whose capacities fix the empirical domain of an accepted theory. These are conceptual questions about scientific practice; the existence of an artificial research instrument alone establishes no claim about its consciousness, subjective experience, or perceptual status.

Monton and van Fraassen's account of observable properties (2003) ties observability to empirically determinable features of objects and observers. The 2024 discussion of augmented observation explores what follows when community membership changes. Applied to nonbiological inquiry, these arguments require explicit conditions for treating a system as an observational participant rather than a measuring instrument producing evidence for another observer. They also require an account of independent calibration and the traceability of system outputs to their sources.

Within the philosophical horizon of the Artificial Era, this distinction has a clear consequence: the identity of scientific reason's bearer and the evidential operations of scientific observation are separate questions that can intersect. The project-specific categories of Artificial and Artificial Sapiens supply an interpretive vocabulary for discussing historical changes in participants; they do not substitute for empirical investigation of particular systems. Constructive empiricism remains instructive precisely because it demands that claims about expanded observation be tied to specified capacities, objects, circumstances, and records of inquiry.

Interpretive Synthesis

The reconstructed argument supports a determinate interpretation. Constructive empiricism is strongest as an account of a defensible form of scientific commitment: one may understand a theory literally, use it rigorously, and believe in its empirical adequacy while suspending judgment about the truth of its unobservable ontology. This interpretation has explanatory power for collaborative scientific practice, for idealized modeling, and for the possibility that a successful theory is later superseded. The foundational account of acceptance (van Fraassen 1980), read alongside the 2026 historical reconstruction (Van Dyck 2026), supports distinguishing the public aim of inquiry from individual rational commitment.

The doctrine also carries obligations. Empirical adequacy cannot be reduced to past data; it concerns the broader observable world. Observability must have content grounded in empirical investigation, and the connection between representations and phenomena must be examined rather than assumed. Successful intervention and novel prediction create genuine reasons for realists to argue for stronger commitments. The resulting controversy is rationally productive because the participants can share standards of evidence while disputing the legitimate extension of belief.

The decisive unresolved pressure is the changing observational domain. The domain in which a theory must prove empirically adequate is partly determined by the capacities and constitution of the epistemic community. As those capacities expand, empirical testing can expose limits of previous theories, and the aim of science must accommodate that possibility. The most productive further inquiry will specify observer membership, conditions of detection, and the ways new instruments or participants generate independent constraints. This conclusion is an original synthesis of the examined problems. It remains within the established constructive empiricist debate and creates no separately named theory, principle, or doctrinal system.

Conclusion

Constructive empiricism, as formulated in van Fraassen's founding scientific-image account (1980), identifies empirical adequacy as the central aim of science and as the content of belief involved in theory acceptance. Its founding achievement was the separation of a theory's literal meaning from the epistemic commitment made by those who accept it. Scientific theories can be understood, tested, explained, and used without assuming that acceptance commits every investigator to their complete unobservable ontologies. This distinction remains philosophically consequential in experimental science, theoretical modeling, and instrument-mediated inquiry.

The historical and analytical examination establishes the costs of that position as clearly as its advantages. Belief in empirical adequacy reaches beyond observed data to the wider domain of observable phenomena. The status of observability therefore carries the burden of making the position intelligible and practically discriminating. Debates about microscopy, counterfactuals, theory-ladenness, and expanded perceptual capacities are constitutive tests of the doctrine. Scientific realists, in turn, bear the burden of showing when explanatory and interventional success makes additional ontological belief warranted rather than merely possible.

The analysis concludes that constructive empiricism offers a coherent and demanding epistemic orientation when coupled with a revisable, empirically accountable understanding of observation and with a differentiated account of science's collective purposes. Its standing as the uniquely mandatory interpretation of science is a further claim, not entailed by this analysis. The controversy remains open at a precise point: how scientific communities should define and revise the observable domain while preserving evidential independence. That question will continue to matter wherever science enlarges its investigative capacities and changes the constitution of its participants.