Original, reference example. This speech excerpt was recorded using 16 bits at 16 kHz. For comparison purposes, we will call this data rate 1.0.
Downsample to 11 kHz. Some of the compression schemes I wanted to use did not offer a 16 kHz rate, so I downsampled to 11 kHz. This is one way to reduce data rates. The normalized rate is: 0.69.
CCITT G.711 (8-bit uLaw), in an AVI file. Since this uses 8-bit samples, we expect half the data rate. The normalized rate is: 0.36. I assume this is more than half of 0.69 because of the AVI file overhead.
ADPCM (IMA ADPCM in an AVI file). This appears to be ADPCM using 4-bits per sample at 11 kHz, giving it about half the size of the uLaw version. The normalized rate is: 0.19.
TrueSpeech 8.5. TrueSpeech is a proprietary system. A reasonable guess would be that this is a vocoder or lpc-based compression technique, given the low bit rate and the use of "speech" in the brand name. The normalized rate is: 0.048.