In recent decades, early warning systems using tilt sensors to detect the pre-failure stage before the slope failure have been developed and used in slope monitoring because of the simple installation and low cost of those systems. However, few studies were carried out to investigate the pre-failure tilting behaviors of slopes, and the method to predict the occurrence of the slope failure based on the slope tilting measurement is also unclear. In this paper, the pre-failure tilting behaviors of slopes were studied by conducting different types of model tests as well as a field test in which the slope failure was induced by applying artificial rainfall. A linear relationship between the reciprocal tilting rate and time was observed in the accelerating stage of tilting before the slope failure, and the equation for this linear relationship was also proposed in this study. The slope failure time can be forecasted and corresponds with the time when the reciprocal tilting rate becomes 0 min/o. This paper provides a method to potentially predict the occurrence of landslides and reduce the risks of damage induced by landslides with the slope tilting measurement.