The 12-month moving averages of solar indices and CR (cosmic ray intensity) were compared. The long-term variation had a negative correlation (already known). From the hysteresis plots for five cycles (odd cycles 19, 21, 23; even cycles 20, 22), it was seen that whereas odd and even cycles showed different patterns, the detailed evolution in similar cycles (even or odd) was not alike. The reason for complication is as follows. On medium-term time scales, in some sunspot cycle maxima (notably cycles 22 and 23), there were two peaks in solar indices separated by about 2 years. These were reflected in CR as two distinct minima, some with a lag of a few months but others almost coincident with maxima of solar indices. The implication would be that CR modulation as observed on Earth contains two parts: (1) long-term variation due to magnetic field structure in total heliosphere, determined by convection-diffusion and drift mechanisms (this part has time lags of several months and is characterized by hysteresis phenomenon) and (2) short-term and medium-term variations caused mainly by accumulated shock waves (Forbush effects and assotiated phenomena) with time lags of a few days, which in the scale of months or a year or two, would show a time lag zero). The second component would be especially important in maximum of solar activity.