We report the recovery of a spectroscopic event in eta Carinae in 1997/98 after a prediction by Damineli (1996). A true periodicity with P = 2020+-5 days (0.2% uncertainty) is obtained. The line intensities and the radial-velocity curve display a phase-locked behavior implying that the energy and dynamics of the event repeat from cycle to cycle. This rules out S Doradus oscillation or multiple shell ejection by an unstable star as the explanation of the spectroscopic events. A colliding-wind binary scenario is supported by our spectroscopic data and by X-ray observations. Although deviations from a simple case exist around periastron, intensive monitoring during the next event (mid 2003) will be crucial to the understanding of the system.

13 pages, accepted by ApJ Letters (January 2000)