In this paper we critically reexamine the interplanetary cosmic ray radial gradients of >70 MeV cosmic rays as observed by the IMP, Voyager, and Pioneer spacecraft between 1978 and 1996. We separate out three distinct effects on the radial gradient which are due to (1) intensity changes, (2) changes in the radial distance, and (3) the influence of the solar magnetic polarity. The new characteristics that we find, which are at odds in several important ways with many earlier studies, can be summarized as follows: First, the value of Gr depends strongly on the cosmic ray intensity in periods of negative solar magnetic polarity (e.g., 1983.0--1989.5). This intensity dependence is only weakly evident in periods of positive polarity. During the negative polarity periods this is the dominant effect changing the gradients. It masks the radial effects which show only a weak dependence at this time. Second, the radial dependence of the gradient has two regimes. Beyond 10--20 AU the gradient is only weakly dependent on r and is quite small. Between 1 and 10--20 AU the gradient exhibits a very rapid radial falloff during the minimum modulation periods of 1978, 1987, and 1996. This decrease could be a factor of 10 or more in 1978 and 1996. Third, the values of Gr show a strong solar magnetic field polarity dependence. Inside 10--20 AU this is manifested in a much more rapid radial falloff in the 1978 and 1996 positive polarity periods as compared with the 1987 negative polarity period. Beyond 10--20 AU the average radial gradient in the positive cycle in 1996 is only ~0.5%/AU or about 1/3 of that (1.5%/AU) in the negative polarity cycle in 1987. In the positive polarity period the gradient actually increases at large r. This increase, which is predicted in some models as one nears the boundary, may provide a new way of estimating the distance from the modulation boundary. Some features of the behavior of the radial gradient that we observe are predicted by current modulation models, but other aspects of this behavior require further modeling for a better understanding. ¿ 1999 American Geophysical Union