Figure 9:
Linear regression between n of small letters and n of only height errors.
