A reproduction design is used to show that temporal intervals containing brief tones appear longer than empty intervals of the same duration, the effect being independent of duration. These and previous data are discussed within a theoretical framework which allows for the interrelation of data from different time perception tasks; and a reversible encoding model is stated which accounts for much of the data obtained with empty intervals. A "chunking" model, in which tones occurring in an interval serve to segment the interval during encoding, can account for the filled-duration illusion if certain conditions are met. Finally, mechanisms that are consistent with these conditions are stated.