We studied the effectiveness of sonication and filter-size-fractionation to efficiently remove and separate bacterial leaf-surface colonizers from other members of the microbial community. Leaf discs from red oak and paper birch leaves were colonizedin situfor 8 and 22 days in Little Rock Lake, WI, and sonicated at 90 W power output for 60 seconds, followed by filter-fractionation (3.0 μm) to remove and isolate bacteria. This technique consistently yielded the highest bacterial activity (relative Electron Transport System activity) and resulted in bacterial recovery efficiencies from leaf surfaces ranging from 96.5% to 98.8%. In addition, comparisons among various methods currently used to estimate bacterial densities (e.g., malachite green-INT, acridine orange direct counts, scanning electron microscopy approaches) were completed to assess the utility of the MINT method for counting bacteria removed by sonication. Bacterial counts derived from MINT and AODC methods were never significantly different, but always were significantly higher and less variable than counts from SEM. Finally, an attempt was made to address sources of variability in counting precision using the MINT method for bacteria removed from leaf surfaces. For two leaf species, on two sampling dates, most of the variability in MINT cell counts was due to variability among microscopic fields counted, while little was due to either subsample or filter used in cell counts. Thus, the combination of sonication and filter-size-fractionation to remove bacterial leaf surface colonizers appears to be an effective approach, allowing reproducible and precise measurements of bacterial densities and activities.