The Voyager 2 photopolarimeter (PPS) counted the number of photons from the δ Scorpii at a sampling interval of 10 msec as it passed behind the rings of Saturn, yielding a trace of ring opacity at 100-m resolution. We have found that the statistical properties of this data set constrain the size distribution of the largest ring particles. Each PPS count k samples the filling fraction of a "box" on the rings ∼ 150 m long and ∼ 25 m (one Fresnel Zone) wide. If the ring were a uniform gray screen on this scale, then k would obey Poisson statistics, in which case the variance σ2(k) = k, the expected number of counts. However, the ring is actually composed of particles of nonzero size, leading to a stochastic variation in the fractional area of a box blocked by material. Because of this variation, σ2(k) can be significantly larger than that expected from Poisson statistics. Such non-Poisson behavior is indeed seen in the PPS data, especially in the A Ring. In this paper, we derive an expression for the excess noise in the PPS scan due to large particles, and use the measured noise to constrain the upper end of the size distribution throughout Saturn's rings. This method is found to complement the few other methods available for determining ring particle sizes, but is applicable at a much higher spatial resolution (∼ 20 km). The C Ring and Cassini Division hold the smallest proportion of large ({greater-than or approximate}2 m in radius) particlesm whereas the A Ring holds the largest. However, the A Ring shows some evidence that its particles are arranged in a clumpy manner, which could invalidate our conclusions. The central B Ring is too opaque for reliable size determinations. On a finer scale, significant gradients in the size distribution are observed in the transition regions adjacent to the inner edges of Rings B and A. However, the many abrupt changes in optical depth which characterize the C Ring and Cassini Division are not reflected in the relative number of large particles. © 1990.