Every sound file was recorded in a male version and a female version although the female was used almost all the time. I asked about this and was told the following:
* Women respond equally to male and female voices
* Men instinctively rebel against petty instruction given in a male voice... unless...
* In a threatening situation (fire, football riot etc) orders in a male voice are trusted more
I don't know how true any of this really is when applied in the real world, but maybe the similar studies have been applied to conversational interfaces.
This is a minor point in the program, though. It's really one of my favorite Nova episodes and I recommend it to anyone.
Usually, telling if a singer is male or female isn't any more difficult than it is for speakers. While in singing there's considerable overlap in ranges between male and female voice types, the overtones produced by a male singer differ from those produced by a female singer. A C5 sung by a tenor will sound very different from a C5 sung by a soprano.
That said, especially in pop music there's the tendency for male singers to use their falsetto register, which tends to sound 'feminine' or androgynous. Moreover, in pop music you also have an abundant use of voice filters and synthesizers such as the infamous vocoders that make distinguishing voice types more difficult.
Lots of [citation needed], but interesting historical background on voice warning systems.