set speakers large or small?

Found a few references. In a nutshell, Interaural Level Difference (ILD) and Interaural Time Difference (ITD) are the means for interpreting the location of a sound source. Level difference is mostly used for high frequencies due to wavelength relative to head size, yet the time difference is identical for low frequencies as for high. Time difference allows for low frequency localization. (See the abstract at this link, http://adsabs.harvard.edu/abs/2005ASAJ..117.2391B) The precedence effect also comes into play in a typical domestic room where one is listening relatively close to the source, and level differences are aided by proximity reinforcement. Try listening to sine waves coming from only the left or right channel. You can easily identify which speaker is producing the sound unless the frequency hits a room mode, then ringing may obscure the location cues. More reverberant spaces also make it more difficult to localize, even for high frequencies. Location cues tend to be easily obscured, and room acoustics can affect perception, for better or worse, making it difficult to determine a specific cut-off frequency. It's a convenience to say frequencies under X Hz are nondirectional.

This link suggests learning is involved: http://www.ncbi.nlm.nih.gov/pubmed/7602671

This link goes into greater detail: http://www.aip.org/pt/nov99/locsound.html

One more: http://www.ncbi.nlm.nih.gov/pubmed/1564201

Ojgalli,

Only the abstract of the JASA article seems relevant, and it suggests that the model they tested and human ability to localize LF sounds is poor, especially in other than anechoic conditions. The link didn't get me to the JASA article, so I was unable to see a breakdown of their results, but I'd guess the octave band centered at 125 Hz was by far the most localized.

I did enjoy the reference to the binaural processing model of Lloyd Jeffress. My primary interest when I was a post-doc was binaural processing, and I thought the Jeffress model of neural-spread made much more sense than the equalization-cancellation (E-C) model derived from the theory of signal detection (TSD). Jeffress was a nice guy and helpful in encouraging and critiquing my work. I did a study of the detection of sinusoids in computer generated noise waveforms in which I adjusted the phase of the sinusoid in the noise waveforms. I imagined it would be a crucial test of the theories, and the detectability did vary widely, but not as predicted by either model. By that time, I was losing interest, and never published the work even though I was encouraged by Jeffress to do so.

This research was done with earphones, so we could control the signal that was delivered to the ear, and the lowest frequency was usually 250 Hz. We did have an anechoic chamber that went pretty low, but transducers were inadequate for LF work.

db

wtf are you guys talking about,please come down to my level