Matching subject and verb voices