Dear Joelle,

What does happen exactly? Does model estimation freeze at this stage?

The message as such is nothing special, although it shouldn't stop at chunk 20 of 76. Temporal nonsphericity refers to the fact that errors terms are likely not temporally independent (in other words, the error term for voxel x in volume n might correlate with the error term for voxel x in volume n+1), which runs counter the assumptions. Therefore we try to account for these issues e.g. with an autoregressive model, which is enabled per default during model specification. For statistical details see e.g. Purdon & Weisskoff (1998, Hum Brain Mapp, https://dx.doi.org/10.1002/(SICI)1097-0193(1998)6:4<239::AID-HBM4>3.0.CO;2-4 ) and Friston et el. (2000, Neuroimage, https://dx.doi.org/10.1006/nimg.2000.0609 ).

Best

Helmut