Publication bias and statistical significance

Chris of Mixing Memory has blogged a great discussion of statistical significance, scientific publishing, and publication bias - the lower likelihood of getting accepted for publication if your study has null results, i.e. fails to show a statistically significant difference.

For more details on publication bias, consider the Cochrane Collaboration description and the results from a quick PubMed search on the topic ("publication bias" has been a Medical Subject Heading (MeSH) term since 1994) .

David Wallace of Fayetteville State University has developed a good description of Type I and Type II statistical errors (PDF file).