The many-to-one situation that each vibration moment corresponded to the scoring moment for a time lag was transformed to a one-to-one situation that each vibration process had only one representative time lag. The effect of the time lag on subjective scorings was tested and regressed. A fractional order calculus was then introduced to capture the effect of vibration history and the order that gave the best prediction precision for subjective scorings was chosen as the best order. Test results revealed that a short time lag does not influence the subjective scorings, while a longer time lag can have a significant effect. A structural behavior function was then adopted to arrive at a better description of the relationship between the time lags and the subjective scorings. If 0.7 and 0.3 order integration are used for the frequency weighted principle acceleration and angular velocity respectively, an 11.9% improvement can be achieved in the precision error. It reflects that humans are sensitive to cumulative vibration.