GSM divides output of the RPE-LTP speech coder into 3 categories :
1-50 bits called "Very Important bits"
2-132 bits called "Important bits"
3-78 bits called " Not so Important bits"
the 50 bits are very important for having intelligible speech ,so they're applied to both block coder and 1/2 rate convolutional encoder .The 132 bits are important but not as important as the 50 bits so they're applied only to the convolutional encoder .Speech can be understood even if an error occured at the last 78 bits so no channel coding is applied to them .Afterwards, interleaving is apllied to mitigate burst errors .