The consensus of soil fertility specialists working in the northeast USA was that soil testing and recommendation systems for P needed to be reexamined because of recent changes in soil testing methodology in the laboratory and corn (Zea mays L.) production technology in the field. Soil tests (M-COL, MM-COL, B-ICP, M1-ICP, and M3-ICP) were performed by either colorimetry or inductively coupled plasma (ICP) emission spectroscopy on samples from soil test calibration studies conducted during 1998 to 1999 at 51 experimental sites chosen to represent a range of soils, including Ultisols, Spodosols, and Alfisols, in northeastern states (Connecticut, Delaware, Massachusetts, Maryland, Maine, New Hampshire, New Jersey, New York, Pennsylvania, Rhode Island, Vermont, and West Virginia). The mean P measured by M-COL, MM-COL, B-ICP, M1-ICP, and M3-ICP was 8.3, 6.6, 148, 66, and 121 mg P kg−1, respectively. Production practices followed local state extension recommendations at each site and included P fertilizer treatments: none, 15 kg P ha−1 banded, or 60 kg P ha−1 broadcast. Combined analysis of variance over sites showed that plant height at 35 d after planting, silk emergence, grain yield, and grain dry down were enhanced by the broadcast P treatment. There were yield increases (P < 0.10) to the band treatment at only four sites and to the broadcast treatment at nine sites. Cate-Nelson statistical analysis of relative yield in relation to soil test P failed to identify soil test P critical levels for any of the soil test methods. The percentage of experimental sites that had soil test P levels below the currently used critical levels in the region ranged from 14 to 65% of the sites. Results showed that 17 to 47% of those sites testing below the critical level exhibited a yield increase (P < 0.10) to broadcast P. Some of the yield responsive sites had soil test P above currently used critical levels. The calibration data obtained from the present study and the relationships examined between soil test P and relative yield do not necessarily validate the currently used soil test P critical levels nor does the data enable much refinement. This study shows that the current critical levels frequently permit both types of errors in soil test prediction; indicating a need for P fertilization when it may not be needed and not indicating a need for P fertilization when it may be needed. The second type of error is usually avoided by recommendations for crop removal rates of P.