Weird attenuator circuit

http://postimg.org/image/459i6l03p/

Why / how does this circuit work?In my modest level of understanding of BJTs, this shouldn't work :-D
The two BJTs emitters connected together, but not to anything else? How can any current flow there?

complete circuit:
http://postimg.org/image/6arbcsgqz/

as LTspice file here:
http://s000.tinyupload.com/index.php?file_id=45691190684901314743

Doing AC analysis, looking at the "AudioOut" shows that the feedback path does what it's supposed to do.
The frequency response gets a peak around the cutoff, the higher the voltage at the control input, the higher the peak.

(Only the 2x BJT attenuation part is from the manual as said in the spice file, I just cobbled together a fixed 4 pole LPF before that to see the familiar response)

BJTs have a zone for very small Uce ( < 0.2V) where they act like triodes, that means voltage controlled resistors. It's basically the same as for FETs, where that zone is up to 10 times higher (Uds < 2V).

You can see it in the data sheet, better said in the Ic/Uce vs Ib curves. We are talking about the small zone on the left side where the curves aren't yet almost horizontal.