Peak ground acceleration (PGA) as a measure of earthquake motion intensity is an important factor in the earthquake-resistant design and reliability analysis of structures. The purpose of this article is to examine the spatial variability of PGAs (recorded at the same epicentral distance) as a function of separation distance. To do this, we define PGA ratios as spatial intraevent variations of PGAs and examine their statistical characteristics. We analyze the probability distribution of the ratios and formulate equations for their probability density functions, mean values, standard deviations, and percentiles. These statistics are then estimated using accelerometer arrays of the Chiba, SMART-1, and SIGNAL databases. Then, the relationship between these statistics and the station separation distances is analyzed. We found that the means and standard deviations have an almost linear relationship with the logarithm of the station separation distances ranging from several meters to 100 km. Finally, based on the fiftieth and ninety-fifth percentiles, the differences between PGAs at two different sites due to future earthquakes are discussed.