When a mixture of sound from many sources arrives at the ear, the auditory system attempts to segregate it into different perceptual streams. For sequences of sounds, the effects of basic acoustic properties (e.g., frequency separation, rate) on streaming are well understood, but much less is known about the effects of more complex acoustic attributes. Dynamic variations in frequency spectrum are known to have an important effect on the timbre of sounds. We investigated whether dynamic variations also affect stream segregation. Periodic tones were used, comprising harmonics 1–6, and presented in long sequences (ABA–ABA–…). Tones A and B always differed in fundamental frequency, a factor known to influence streaming, and had either the same or different patterns of spectral variation. The total amount of variation (spectral flux) was held constant. Listeners judged moment-by-moment the grouping of these sequences, and the measure used was the proportion of time a sequence was heard as segregated. We found that sequences of complex tones with different patterns of spectral variation are more likely to segregate when this results in different patterns of change in the frequency centroid over time (Experiments 1 and 2), but not when it does not (Experiment 3).