PKP(DF) travel times reported by the International Seismological Centre (ISC) are analyzed to study the depth dependence of inner core anisotropy. Plots of PKP(DF) travel time residuals versus range for different subsets of the data clearly show large anomalies at ranges beyond about 149¿ for those paths with inner core ray angles within 30¿ of the rotation axis. On average, the times are about 3 to 5 s earlier for these paths than for rays at other angles, consistent with recent waveform measurements of anomalous PKP(DF) times by Creager (1992) and Song and Helmberger (1993), and indicative of inner core anisotropy in which the P velocity is about 3% faster for N-S paths than E-W paths. The size of these apparent anomalies in the ISC data are much larger than those inferred from several previous analyses of ISC PKP(DF) data which indicated 0.5 to 2 s anomalies at these ranges. The earlier studies applied binning and averaging schemes to the data using windows that were overly tight, thus excluding some of the most anomalous data points and biasing the results toward smaller anomalies. To avoid this problem, I apply a method for computing summary ray residuals that finds the maxima in smoothed histograms of the residuals and does not rely on a fixed interval to window the data points. Standard errors on each summary ray residual are estimated using a bootstrap technique that randomly resamples the data. Plots of the summary ray residuals for several range intervals reveal a small number of independent paths nearly parallel to the rotation axis for which the arrivals are 3 to 5 s early.

Although the data are sparse for paths within 30¿ of the rotation axis, anisotropy does not appear to diminish with depth within the inner core and may cause a much larger anomaly near 180¿ than the 2-s value obtained in several prior studies of ISC data. The uniform anisotropy models of Creager (1992) and Song and Helmberger (1993) fit the bulk of the data reasonably well. However, the observed residuals at ranges between 132¿ and 140¿ are smaller than those predicted by most anisotropy models, suggesting that the anisotropy may diminish within the outermost 50 km of the inner core. ¿ American Geophysical Union 1994