I enjoyed hearing the variety of accents that are recorded. A good range of diversity on the site. What interests me is the number of self-repairs in the speech of the samples. Might these be due to a parsing effect instigated on the subjects by the exigency of reading the passage?