This paper presents a comparison of the measured and modeled ionospheric response to magnetic storms at Millstone Hill and Arecibo during March 16-23, 1990. Magnetic activity was low until midday UT on day 18 when Kp reached 6, and days 19 and 20 were quite, but a large storm occurred around midnight UT on day 20 (Kp=7) and it was moderately disturbed (Kp=4) for the remainder of the study period. At Millstone Hill, the daytime peak electron density (NmF2) showed only a modest 30% decrease in response to the first storm and recovered to prestorm values before the onset of the second storm. The model reproduces the daytime peak electron density well for this period. However, the severe storm on March 20 caused a factor of 4 depletion in electron density, while the model densities were not greatly affected. The inclusion of vibrationally excited nitrogen (N>2) in the model was unable to account for the observed large electron density depletions afterward March 20. The storm did not appear to affect the overall magnitude of the electron density at Arecibo very much, but did cause unusual wavelike structure in the peak density and peak height following the storm. The model reproduces the daytime NmF2 very well for Arecibo, but after sunset the model densities decay too rapidly. This study indicates that successful modeling of severe ionospheric storms will require better definition of the storm time inputs, especially of the neutral atmosphere. ¿ American Geophysical Union 1994