The invention makes it possible to combine a scaleable audio coder with TNS technology. According to the inventive method for encoding time signals (x1) sampled in a first sampling rate, second time signals (x2) with a sampling rate smaller than the first sampling rate are generated (12). The second time signals (x2) are then encoded (14) according to a first coding algorithm, and written into a bit stream (xAUS) (16). The encoded second time signals (x2c) are then decoded (14) again and are transformed (23, 24) into the frequency range, as are the first time signals. TNS prediction coefficients are then calculated (25) from a spectral representation of the first time signals (X1). The transformed output signal (X2cd) of the coder/decoder (14) with the first coding algorithm and the spectral representation (X1) of the first time signal are subjected to a prediction of the frequency (27) in order to obtain spectral residual values for both signals using the prediction coefficients calculated on the basis of the first time signals alone. These two signals are evaluated against each other (26, 28). The evaluated spectral residual values (Xb) are then encoded by means of a second coding algorithm in order to obtain coded evaluated spectral residual values (Xcb). These evaluated spectral residual values are written into the bit stream (xAUS) in addition to side information with the prediction coefficients.