Category: epistemology
Type: paradox
Associated thinker(s): Henry Kyburg
Original publication or approximate period: Kyburg, Probability and the Logic of Rational Belief (1961)
Difficulty: Advanced
Themes: rational belief, probability, knowledge
Related experiments: Gettier Cases; Sleeping Beauty
Suppose a fair lottery has a very large number of tickets and exactly one winner. For any particular ticket, the probability that it loses is very high. If full belief is tied to a fixed probability threshold below certainty, as in a simple Lockean rule, the lottery can be made large enough that each individual claim—“ticket i will lose”—passes that threshold.
If rational belief is also closed under conjunction, you should then believe that all the tickets will lose. But you also believe that exactly one ticket will win. Those beliefs are jointly inconsistent. The tension is among high-probability belief, conjunction closure, and avoiding inconsistent belief sets; the setup alone does not tell us which principle to give up.
Can each claim in a set be rational to believe when the claims cannot all be true, and what does that show about the relation between probability and full belief?
The paradox puts pressure on three attractive ideas: sufficiently high probability can warrant full belief; rational belief is closed under conjunction; and rational beliefs should not be jointly inconsistent. In the standard finite-lottery setup, these principles cannot all be retained together.
It also connects to questions about knowledge. Whether statistical evidence can establish knowledge is a related debate, not a conclusion settled by the lottery argument itself. The puzzle does not, by itself, show that probabilistic reasoning is unreliable or decide which principle should be revised.
There is no single agreed resolution. One response, associated with Kyburg, rejects conjunction closure: a person may hold individually rational beliefs that are jointly inconsistent without thereby believing a contradiction. Another response rejects or qualifies a simple rule that any sufficiently probable proposition warrants full belief; some proposals add conditions that block lottery cases. Other formal approaches model the relation between degrees of confidence and full belief differently. These are families of responses, not an exhaustive or settled list.
The preface paradox is related but distinct. An author may accept each assertion in a book while also believing that the book contains an error; this challenges conjunction closure without relying on a lottery or a probabilistic acceptance rule. Roy Sorensen’s SEP account attributes the published preface paradox to D. C. Makinson (1965), drawing on Raymond Wilder’s preface.
The puzzle is not a mistake in the arithmetic of lotteries, nor does it depend on a particular ticket count. The issue is how high-probability claims relate to full belief and whether such beliefs combine under conjunction. It does not establish that statistical reasoning is generally unreliable, and it does not settle the debate about knowledge from statistical evidence.
For any fixed probability threshold below certainty, increasing the number of tickets can make each individual loss claim exceed that threshold; the tension remains if the other assumptions are retained. The preface paradox, associated with D. C. Makinson, presents a related challenge without a lottery or probabilistic acceptance rule.
The lottery setup and its core tension are presented here as an educational explanation, informed by the SEP discussions and Wheeler’s scholarly review. The primary book is bibliographically dated 1961, but the book’s argument pages were not accessible for this review: the Internet Archive record marks the item as access-restricted. Wheeler identifies the book as the first published statement of the paradox and notes an earlier conference formulation (1959/60) published in 1963. The exact original wording is therefore not quoted here.
The alternatives in §4 summarize approaches discussed in the SEP formal-belief entry; they are not offered as a consensus or exhaustive taxonomy. The distinction between joint inconsistency and believing an explicit contradiction follows the SEP epistemic-paradoxes discussion. The preface-paradox attribution is based on Sorensen’s account of Makinson (1965) and Wilder’s preface. Questions about whether statistical evidence can support knowledge are included as a related issue, not as a result of the lottery argument. Questions and prompts are editorial framing.
Research check: 2026-11-01. This educational explanation is not a substitute for the primary text or for the broader scholarly debate.