Culture-independent high-throughput sequencing-based methods are widely used to study bacterial communities. Although these approaches are superior to traditional culture-based methods, they introduce bias at the experimental and bioinformatics levels. We assessed the diversity of the human salivary microbiome by pyrosequencing of the 16S rDNA V1-3 amplicons using metagenomic DNA extracted by two different protocols: a simple proteinase K digestion without a subsequent DNA clean-up step, and a bead-beating mechanical lysis protocol followed by column DNA purification. A high degree of congruence was found between the two extraction methods, most notably in regard to the microbial community composition. The results showed that for a given bioinformatics pipeline, all the taxa with an average proportion >0.12% in samples processed using one extraction method were also detected in samples extracted using the other method. The same taxa tended to be abundant and frequent for both extraction methods. The relative abundance of sequence reads assigned to the phyla Actinobacteria, Spirochaetes, TM7, Synergistetes, and Tenericutes was significantly higher in the mechanically-treated samples than in the enzymatically-treated samples, whereas the phylum Firmicutes showed the opposite pattern. No significant differences in diversity indices were found between the extraction methods, although the mechanical lysis method revealed higher operational taxonomic unit richness. Differences between the extraction procedures outweighed the variations due to the bioinformatics analysis pipelines used.