How per position posterior probability is calculated in HMMER?

Question

[m.taheri]

The HMMER software has a hmmalign command that performs a multiple sequence alignment of a set sequences by aligning them individually to a profile HMM. The alignment result is outputted in STOCKHOLM format. The result contains a PP line that stands for posterior probability for each residue. In its documentation it is said that this line represents the probability that each residue is assigned to the HMM state corresponding to its alignment column, as opposed to some other state. But I could not understand exactly how this value is calculated for each residue?

Thanks a lot.