Artificial head (or dummy head) recording and playback systems are increasingly used. The main advantage of these systems is their ability to reproduce three-dimensional spatial aspects of sound. However, the exact reproduction of direction has been questioned from time to time. The study presents an objective and subjective evaluation of eight artificial heads. In the objective test, the head-related transfer functions of the heads were compared to those of humans. Significant deviations were seen for all heads, and none of them can be characterized as a mean, median, or typical head. In the subjective test a panel of listeners indicated direction and distance to sound sources recorded by the head and reproduced by means of carefully equalized headphones. The reproduction of direction was in general poor, especially in the median plane.

The localization performance was studied when subjects listened (1) to a real sound field and (2) to binaural recordings of the same sound field, made (a) in their own ears, and (b) in the ears of other subjects. The sounds to be localized were loudspeaker reproductions of female speech at natural level, from 19 different positions in a standard listening room. The binaural recordings were reproduced by carefully equalized headphones. With individual recordings the performance was preserved compared to real life, whereas nonindividual recordings resulted in significantly more errors for sound sources in the median plane. Errors were seen in terms of confusion not only between nearby directions, but also between directions further away, such as between sound sources in front and behind the subject. However, nonindividual recordings made in the ears of a carefully selected ‘‘typical’’ subject resulted in a performance much closer to the real-life performance, although still inferior.