Thirteen listeners received eight dichotic-listening trials. Each trial consisted of 30 pairs of consonant-vowel nonsense syllables spoken by a male talker. The average percentage ear advantage varied by no more than 0.6% over four two-trial blocks, but the test-retest correlation coefficient between the first and second two-trial blocks was only +0.66. The same data, plotted in cumulative percentage form, were distributed approximately as a normal curve with a mean right ear advantage of 6.1%, but with a standard deviation of 17.0%. As a result, 33% of the 104 trials yielded a left ear advantage. The variability in percentage of ear advantage was relatively similar for all listeners, independent of the size or direction of the average ear advantage.