Survey suggests that a high frequency of omissions and mistakes could cause more damage to scientific integrity than occasional cases of fraud