Summary A major concern in molecular ecological studies is the lysis efficiency of different bacteria in a complex ecosystem. We used a PCR-based 16S rDNA approach to determine the effect of two DNA isolation protocols (i.e. the bead beating and Triton-X100 method) on the detection limit of seven feces-associated bacterial species of different genera. Glycogen was used in these protocols to improve the precipitation of small concentrations of DNA in ethanol without affecting the sequential procedures. The PCR detection limit of 16S rDNA amplicons on agarose gel from the seven strains tested varied between 8.0 (±1.3) × 10 4 and 4.3 (±1.6) × 10 6 cells for the bead beating method, and between 8.0 (±1.3) × 10 4 and 5.4 (±0.7) × 10 8 cells for the Triton X-100 method. These large differences are most like due to the difference in cell lysis efficiency, since a competitive PCR experiment did not indicate any preference for Gram negative, low G + C Gram positive or high G + C Gram positive bacteria. Denaturing gradient gel electrophoresis (DGGE) analysis was performed to investigate the effect of both DNA isolation protocols on the lysis efficiency of bacteria in fecal samples. A higher diversity in fecal samples was observed with the bead beating method than with the Triton-X100 method. Bands in the bead beating method-derived DGGE profiles corresponding to bands of cloned sequences of the Clostridium coccoides-Eubacterium rectale group and uncultured Fusobacterium prausnitzii were absent or had low intensity in the Triton X-100 method-derived profiles. The applicability of the bead beating method was further investigated by analyzing biopsy samples from the human colon which contain approximately 10 6 cells.