CORRELATIONS ON DEBRIS FROM SILICATE BURSTS 75 fallout fairly well. The Small Boy field was the cigar-shaped downwind area typically associated with such shots. The Johnie Boyfield was very, perhaps atypically, narrow with a very hot line down the center, which was visible on the ground as a darkened streak. The NRDL collection array at Sedan was not sufficiently widespread to define the limits of the close-in fallout field. With each of these shots, one can associate a sort of average, or typical, value (or narrow range of values) for the ratio rg, 4, observed in gross samples. It is to be understood that this range of Igy 9, is much extended when cloud samples, sieve-fraction samples, and samples from the peripheral stations are considered. Nonetheless, the average value is useful for describing the overall degree of fractionation of the shot. For Johnie Boy a weighted average of this ratio for the hot-line stations is around 0.03, indicating very severe fractionation. For Small Boy the values for most stations are in the range 0.1 to 0.2, indicating more moderate fractionation. All Sedan samples were Sieved, but the Teg 95 values for gross samples were reconstructed by properly weighting the values for the sieve fractions and were found to range from about 0.5 to 3.8. Valuesin this range would be anomalous for local fallout from a true surface burst, but they seem to be characteristic of cratering (buried) shots, venting underground explosions, and venting underwater bursts. All of the Johnie Boy, Small Boy, and Sedan r; 9; values have been log—log plotted against rgg 95. Figure 1 indicates how someof the data accommodate themselves to this treatment: These are the data on '4°Ba for the Johnie Boy shot, and they illustrate a particularly satisfactory fit. The slope, the y intercept at log x = 0 (x = 1), and the coefficient of correlation* have been determined by linear regression to be 0.61, 2.4, and 0.98, respectively. Not all of the data, by any means, fit a log— log plot so neatly. Figure 2 shows a much less satisfactory and more typical example. These data are part of the results for *°Sr in the Small Boy shot. The coefficient of correlation here is only 0.684. The data were originally fitted by lumping together data points from all three contracting laboratories. In the process of investigating a poor fit of ‘*'Ce data for Small Boy, we noticed that points from Laboratory A could be fitted fairly well, whereas those from Laboratory B were scattered so badly as to suggest that Laboratory B was experiencing difficulty with this analysis. Figure 3 shows these data. Investigation of the '*Ru and !“Ru data showed a similar situation, but it was reversed with respect to the two laboratories— Laboratory B points correlated notably better than did Laboratory A points. *The square of the coefficient of correlation reflects that part of the variation in one set of measurements which can be explained by their dependence on the other.