On errors in deducing a cause from an effect