Optokinetic eye movement (OKR) is evoked in response to full-field visual motion to stabilize retinal images. If periodic visual stimuli are delivered repeatedly for hours, the OKR gain, defined as eye velocity / visual motion velocity, increases towards 1.0 (OKR gain adaptation). In addition, a predictive behavior, named period tuning occurs in the goldfish OKR. After continuous exposure to a periodic visual motion, eye velocity decreases as if timing of the directional changes in visual motion was predicted. In the presence of a constant stimulus duration period tuning occurs even when stimulus onset was randomized, suggesting that the timing of the OKR stimulus end was learned. Herein, we quantitatively evaluated period tuning in the OKR to identify what behaviors were learned other than a prediction of the stimulus end. Goldfish were gently fixated in a cylindrical water tank with previously implanted head posts. In 7 goldfish unidirectional, rectangular velocity random dot visual stimuli were projected on the wall of the water tank to induce horizontal OKR. The stimulus rotated in the clockwise (CW) direction at 20 deg/s for 8 sec, stopped for 8 sec and the paradigm repeated for 3 hours. Angular position of both eyes were measured by using a scleral search coil technique and stored at the sampling rate of 1kHz for offline analyses. We observed a decrease in eye velocity starting a few seconds prior to the end of the CW visual stimulus, and a gradual increase in eye velocity (0.8 to 1.5 deg/s) starting ~ 1 second prior to the onset of the next CW visual stimulus. This result demonstrates that a learned OKR prediction occurs both at the beginning and end of the visual stimulus. When tested with a visual stimulus consisting of a slower CW visual motion (10 deg/s), eye velocity exceeded the visual stimulus (OKR gain > 1), suggesting that the speed of the visual motion was also predicted. Acute cerebellectomy did not block OKR performance to any of the periodic visual stimuli, but it did both prevent and remove acquisition of the predictive OKR behaviors. We conclude that these findings link an intrinsic circuitry of the cerebellum to a predictive eye movement behavior.