acy may be. Nevertheless it is advisable to separate this species of
inferences from experience--whose certainty is not doubted except by the
philosophers--from uncertain probabilities, as a class intermediate between
the latter and demonstrative truth (demonstrations--proofs--probabilities).
All reasonings concerning matters of fact are based on the relation of
cause and effect. Whence, then, do we obtain the knowledge of cause and
effect? Not by _a priori_ thought. Pure reason is able only to analyze
concepts into their elements, not to connect new predicates with them. All
its judgments are analytic, while synthetic judgments rest on experience.
Judgments concerning causation belong in this latter class, for effects are
entirely distinct from causes; the effect is not contained in the cause,
nor the latter in the former. In the case of a phenomenon previously
unknown we cannot tell from what causes it has proceeded, nor what
its effect will be. We argue that fire will warm us, and bread afford
nourishment, because we have often perceived these causal pairs closely
connected in space and time. But even experience does not vouchsafe all
that we desire. It shows nothing more than the coexistence and succession
of phenomena and events; while the judgment itself, _e. g_., that the
motion of one body stands in causal connection with that of another,
asserts more than mere contiguity in space and time, it affirms not merely
that the one precedes the other, but that it produces it--not merely that
the second follows the first, but that it results from it. The bond which
connects the two events, the force that puts forth the second from the
first, the necessary connection between the two is not perceived, but added
to perception by thought, construed into it.[1] What, then, is the occasion
and what the warrant for transforming perceived succession in time into
causal succession, for substituting _must_ for _is_, for interpreting the
observed connection of fact into a necessary connection which always eludes
observation?

[Footnote 1: The weakness of the concept of cause had been recognized
before Hume by the skeptic, J. Glanvil (1636-80). Causality itself cannot
be perceived; we infer it from the constant succession of two phenomena,
without being able to show warrant for the transformation of _thereafter_
into _thereby_.]

We do not causally connect every chance pair of successive events, but
those only which have been repea