Bas C. van Fraassen (born 5 April 1941) is a Dutch-American philosopher of science, best known for developing constructive empiricism, a position that holds that the goal of science is empirical adequacy rather than truth about unobservable entities. He has held faculty positions at Yale University, the University of Toronto, Princeton University, and San Francisco State University. His work spans philosophy of science, philosophy of logic, philosophy of language, and formal epistemology.
Van Fraassen was born in Goes, Netherlands, and emigrated to Canada as a young man. He completed his doctorate at the University of Pittsburgh in 1966 under the supervision of Adolf Grünbaum. His early work focused on the logic of conditionals and the semantics of free logic. He later turned to the philosophy of science, where his 1980 book The Scientific Image established him as a major figure in the field. That book introduced and systematically defended constructive empiricism, presenting it as a coherent alternative to scientific realism. He subsequently developed positions in formal epistemology, including a voluntarist approach to belief and probability, and explored the role of symmetry and laws in physics.
Constructive empiricism holds that a scientific theory is acceptable if it is empirically adequate - that is, if it correctly describes the observable phenomena - and that acceptance of a theory does not require belief that it is true with respect to unobservable entities. Van Fraassen draws a sharp distinction between what is observable and what is merely detectable through instruments, though the location of that boundary is itself a subject of philosophical dispute. On his view, to accept a theory is to believe only its claims about observables, while remaining agnostic about its claims regarding unobservables such as electrons or quarks. This position is anti-realist in a specific, limited sense: it does not deny that unobservable entities exist, but holds that science as such does not give us reason to believe in them. Van Fraassen distinguishes his view from older forms of positivism and instrumentalism, arguing that theories can be literally construed as true or false while still holding that acceptance warrants only a weaker doxastic commitment. A full treatment of the realism debate is linked below.
In later work, particularly The Empirical Stance (2002), van Fraassen reframed empiricism itself not as a doctrine but as a stance - a cluster of attitudes, values, and commitments rather than a set of propositional beliefs. This move was intended to address the self-undermining objection that empiricism, if taken as a doctrine, would require empirical justification for empiricism itself. His voluntarist epistemology holds that rational agents have significant latitude in forming and revising beliefs, constrained by coherence requirements such as probabilism and conditionalization, but not by a single uniquely rational prior. This position has been criticized as opening the door to epistemic permissiveness and has been debated extensively in Bayesian epistemology.
Van Fraassen's work has shaped debate in several areas. The Scientific Image reopened the scientific realism debate, which had been somewhat dormant in English-language philosophy, and prompted responses including the development of structural realism and entity realism. His semantic or model-theoretic view of theories - which holds that theories are best understood as families of models rather than sets of sentences - has been influential in philosophy of biology and philosophy of physics. His work on laws of nature, particularly his critique of nomic necessity, contributed to a deflationary tradition that includes the regularity theory and its successors.
There is no settled consensus in academic philosophy of science on scientific realism versus anti-realism; both positions command significant support. Van Fraassen's constructive empiricism remains a live research program. See Scientific Realism - Philosophy of Science Consensus for a survey of the debate's current state.