I think it is now accepting wrong numerical answers as correct, and allowing second and third attempts at multiple choice questions to be treated as first attempts and the mark being counted. (realclimate.org)
As a measure of the compulsive behaviours (represented in the Y axis) we use index calculated as the sum of numerical answers to questions «Does your dog freeze / chase lights etc.?» (journals.plos.org)
The first four questions require exact numerical answers based on number of hours slept, time sleep was initiated, and time respondent awoke. (dreampadsleep.com)