Let’s begin with a statement that is both
provocative and unsettling: we never truly predict anything entirely new. By
‘entirely new,’ I mean empirical facts that cannot be anticipated through
extension or rearrangement of prior experiential patterns. This might sound
counterintuitive - after all, science seems full of predictions. Sure, but as
we’ll see, from a broadly Humean perspective, what we call predictions are
better understood as extrapolations from prior experience. They are reasoned
extensions of patterns we have already observed.
Our knowledge of the world is deeply grounded in experience, in that sensory perception, emotional learning, and many of our expectations about how the world behaves arise from accumulated interactions with reality - whether first-hand or socially transmitted. On this view, reasoning alone is insufficient to generate entirely new facts without experiential content to work upon.
David Hume - probably the most influential empiricist philosopher (and definitely the best empiricist writer) - argued that much of what we know about the world derives from experience, and that our impressions of reality are shaped within the limitations of human perception. Without experiential input, our capacity to form expectations about the world would be drastically constrained. (Hume himself distinguished between “relations of ideas,” such as mathematics and logic, and “matters of fact,” which concern the empirical world - and it is primarily the latter that concern us here.)
To illustrate this, imagine a philosophical thought experiment: a child - let’s call him Henry - is born and raised in conditions of complete sensory deprivation. Suppose, for the sake of argument, that Henry survives physically but receives no sensory data at all. Philosophers have long used similar abstractions to probe the origins of knowledge. In such an extreme scenario, Henry would likely lack language, learned concepts, and expectations about how objects behave. His mental life, at least with respect to empirical knowledge, would be profoundly impoverished.
Of course, modern philosophy and science have strongly indicated that Hume over-stated his assertion – we now know that the human mind possesses innate structures that organise experience, and that infants are not entirely blank slates. Still, Hume’s insight remains one of the great contributions to philosophy.
You still might feel like objecting: but James, don’t we predict new things all the time? In fields like evolutionary biology or physics, scientists often forecast outcomes not yet observed. Even everyday expectations - like “the sun will rise tomorrow” - feel like predictions. But there’s an important distinction to draw here. From a Humean perspective, many such predictions are rearrangements or extensions of existing knowledge rather than anticipations of conceptually independent novelty. What we call “new” often remains anchored in prior patterns and expectations.
Hume also famously argued that while we observe regularities in nature - patterns that lead us to expect one event to follow another - we never directly perceive a necessary connection binding cause and effect. Building on Hume’s insight, in my books I tend to use a working distinction (not a standard Humean term) between “causality,” meaning observed regular patterns, and “causation,” meaning the deeper necessity we often assume but cannot directly observe. Causality, in this sense, is empirical: it arises from repeated observation. Causation, by contrast, represents the unseen necessity we project onto those patterns. We cannot reason our way to that necessity independently of experience.
For example, we expect that if an apple detaches from a tree it will fall to the ground, or that banging your head on a doorframe will hurt. These expectations are grounded in repeated observation and in learned regularities. However, even these expectations are shaped by the models we construct from experience. Classical physics describes gravity one way; quantum theories offer deeper and sometimes counterintuitive accounts. Without the empirical investigation that revealed the quantum world, its laws would have remained inaccessible to us. Our theoretical understanding grows through experience-informed inquiry rather than purely armchair reasoning.
I wrote a neat illustration in one of my books to convey this. Suppose that as of midnight tonight, a fundamental regularity of nature changes: electricity no longer produces a magnetic field. When you wake up, nothing initially seems different - until your radio falls silent, your doorbell fails, and coils of wire refuse to generate magnetism despite functioning batteries. At first you suspect a mechanical fault. Only after repeated failures across society would you infer that a deeper regularity has shifted. The point is not that such a change is scientifically likely, but that without prior experiential clues, we would struggle to anticipate radically unfamiliar behaviour. Our expectations are tethered to what we have already learned.
Even some of the most celebrated scientific “predictions” can be understood, from a Humean angle, as extensions of existing conceptual frameworks. Neptune’s discovery followed unexplained perturbations in Uranus’s orbit; Dirac’s equations mathematically implied antimatter; Maxwell’s equations pointed toward radio waves. Many philosophers of science would still call these genuine predictions - and rightly so within scientific practice. Yet they were not conjured from conceptual nothingness; they emerged from rearranging established ideas within existing theoretical structures. Theory guided discovery, but it did so on the basis of prior empirical and conceptual foundations.
Let us return to Henry. Suppose at age 18 he is suddenly granted the full range of senses. In front of him are two billiard balls, and an experimenter rolls one toward the other. Hume used a similar image to illustrate his point: without prior experience, Henry would have no basis for expecting what happens when ball 1 strikes ball 2. Will ball 2 move? Shatter? Remain still? Only after repeated observations would he begin to associate events and form expectations about physical interactions. What seems obvious to us - that the second ball will roll - is obvious largely because we have learned the regularities of physical reality through experience.
Don’t get me wrong, we mustn’t over-egg Hume here. Hume’s enduring challenge is not to gainsay reasoning’s utility – it is to frame our understanding of empirical reality as inseparable from experiential learning. Of course, Kant developed this by arguing that the mind contributes structuring categories, and modern Bayesian epistemology in some ways extends Humean ideas, somewhat like how Riemannian geometry extends classical Euclidean models. Yet the central problem of how we justify our expectations about the future when they depend so heavily on past experience remains one of Hume’s most enduring philosophical legacies.
