Two methods are described to account for varying temperature during creep. Both employ the modified superposition principle. One uses a reduced time involving a shift factor which is a function of both stress and temperature history. The other considers the strain to be a function of the current values of stress and temperature. Experiments on polyurethane include constant stress creep and recovery at several temperatures in the nonlinear range and an experiment in which the stress was held constant while the temperature increased at a constant rate, then the stress was removed and the temperature decreased at a constant rate. The strain in this experiment was predicted by means of the theories from the results of the constant temperature creep tests. The strain in the constant temperature creep and recovery tests were described by means of the multiple integral representation and the modified superposition principle. Most of the nonlinearity and temperature effect were found in the coefficient of the time‐dependent term.