According to Coren, Ward, and Enns (Sensation and Perception, fifth
edition) citing Gulick, 1971, we have an auditory localization error of
from 10 to about 18 degrees, depending on frequency. To relate this to
your question, there would be an abolute upper limit of 36 sources (360
degrees divided by 10) we could discriminate. In practice, of course,
the actual number would probably be much less, because sources would
mask each other, and localization accuracy is probably not equally good
at all azimuths.
