The most frequently used variable regions for bacterial diversity are V2-V3 and V6. Depending on the kind of study you want to perform (environmental, intestinal, which organism, etc) you might want to select a region already used by other people in your field : that way you will be able to compare and discuss results more easily. For example in human intestinal microbiota studies the V2-V3 region is the most frequent.
Also depends on the technology you are using : 454 is ok for V2-V3 because of the longer fragments. Illumina gives shorter fragments, so it might be interesting to target a smaller region.
The Eldermet study published in Nature using faecal samples from elderly people used primers for the V4 region. But as Sebastien said it depends on the samples you are study and also the platform for analysis.
Depending on your topic and your sequencing technology, you have many choices. Avoid extreme regions (first and last variable regions) as these ones have less representative available sequences in the databases (such as SILVA). Many studies on this topic can be found in publications.