Abstract. This article examines retrospective public-supplied precursor reports statistically, and confirms published hypotheses that some alleged precursors within 100km and within a day prior to the large 1995 Kobe and 1999 Izmit earthquakes, may be valid. The confirmations are mostly at the p<0.001 level of significance. Most significant were alleged meteorological and geophysical precursors, and less often, animal reports. The chi-squared test used, for the first time eliminates the distorting effects of psychological factors on the reports. However it also shows that correct reports are diluted with about the same number which are merely wishful thinking, and obtaining more reliable data would be logistically difficult. Some support is found for another published hypothesis in which other precursors occurred within the ten days prior to the earthquake.