With current CVS, we get 1.0b0*(-1)^(1/4). But in 5.17, bfloat(sqrt(%i)) -> 7.071067811865475b-1 %i + 7.071067811865475b-1 Shouldn't bfloat(sqrt(%i)) give a bfloat result? We have similar problems with float(sqrt(%i)). Ray