Serial order is fundamental to perception, cognition and behavioral action. Three experiments investigated infants' perception, learning and discrimination of serial order. Four- and 8-month-old infants were habituated to three sequentially moving objects making visible and audible impacts and then were tested on separate test trials for their ability to detect auditory, visual or auditory-visual changes in their ordering. The 4-month-old infants did not respond to any order changes and instead appeared to attend to the "local" audio-visual synchrony part of the event. When this local part of the event was blocked from view, the 4-month-olds did perceive the serial order feature of the event but only when it was specified multimodally. In contrast, the 8-month-old infants perceived all three kinds of order changes regardless of whether the synchrony part of the event was visible or not. The findings show that perception of spatiotemporal serial order emerges early in infancy and that its perception is initially facilitated by multimodal specification.