Scour data collected during 1989-94 were evaluated to determine whether pier scour and contraction scour occurred at 22 bridge sites in Ohio. Pier-scour depths computed from selected pier-scour prediction equations were compared with measured pier-scour depths, and the accuracy of the prediction equations were evaluated. Observed pier-scour relations were compared to relations developed through laboratory research. Mean streambed elevations were evaluated to determine the depth of contraction scour. Channel stability was assessed by use of mean streambed elevations at the approach section. Ground-penetrating radar was used at all sites to investigate the presence of historical scour. Pier scour was observed in 45 of 47 scour measurements made during floods; 84 cases of pier scour were documented, 83 at solid-wall piers and 1 at a capped-pile type pier. Estimated recurrence intervals for 27 of the 35 measured streamflows, all on unregulated streams, were less than 2 years. Seventeen pier-scour prediction equations were evaluated. The Froehlich Design equation was found to most closely meet the 'best design equation' criteria for all 84 cases of the observed data. The Larras equation was found to be the best design equation for the observed data where approach-flow attack angles were 10 degrees or less. Observed pier-scour depths and flow depths ranged from 0.5 to 6.1 feet and 3.0 to 19.8 feet, respectively. All pier-scour depths were less than 2.4 times the corresponding pier width. Selected factors were normalized by dividing by effective pier width. LOWESS curves were developed using the 84 cases of observed pier scour. Normalized scour depth increased with normalized flow depth; however, the rate of increase appeared to lessen as normalized flow depth exceeded 2.5. Normalized scour depths increased rapidly as flow intensity approached the threshold value of 1 and then decreased as flow intensities exceeded this threshold. Normalized scour depth was found to increase with Froude number, and a steeper slope was evident for Froude numbers exceeding 0.2. Normalized scour depth was found to increase with median grain size up to about 10 millimeters for bed material near the pier, then decrease for median grain sizes greater than 10 millimeters. Normalized scour depth was also found to decrease as sediment gradation of bed material near the pier increased. The observed pier-scour relations determined from the field measurements tend to support conclusions by previous researchers of streambed scour, except for the previous finding that normalized scour depth decreases consistently with increasing median grain size. Possible factors that may have influenced the observed trends in the relation between normalized scour depth and median grain size in this study are cohesion and scour measurements made at nonequilibrium conditions. LOWESS curves were developed for 45 of 84 cases of observed pier scour where approach-flow attack angles were less than or equal to 10 degrees. These curves were visually compared to LOWESS curves developed from all observations of pier scour. For three relations, differences in the trends of the LOWESS curves were of sufficient magnitude to warrant discussion. Contraction scour was observed in 4 of the 47 scour measurements and ranged from 0.8 to2.3 feet in depth. Analysis of annual mean streambed approach-section elevations indicated that approach sections were generally stable at 18 of the 22 sites. Ground-penetrating radar, a geophysical method that enables subsurface exploration of the streambed when conditions are favorable, was used at all sites to determine whether historical scour had occurred. Results of the ground-penetrating radar surveys at 20 sites in 1990 indicated the presence of historical scour surfaces at 5 sites. At four of the five sites showing evidence of possible historical scour, differences between the estimated depth of historical scour and the maximum observed scour were w