Before testing it is important to understand how multianswer question were graded before the patch and how they will be graded after this modification. Imagine the four parts question (each part default mark is 1 so question total default mark is 4 and penalty set to 33.33333%): Match the following cities with the correct state: * San Francisco: (California/Arizona) * Tucson: (California/Arizona) * Los Angeles: (California/Arizona) * Phoenix: (California/Arizona) And that on first try responses were California - Arizona - California - California and on the second try : California - Arizona- California - Arizona Before this change the resulting grade was 4 -0.3333333*4 = 2.67 because penalty was applied to the whole question even if only 1 subquestion was wrong on the first try. After this change the resulting grade will be 1 + 1 + 1 + (1 - 0.333333) = 3.67 because penalty will only be applied to the fourth part that was wrong on the first try I think that if students could vote on this tracker, this issue would have received many votes Create at least one multianswer (Cloze) question. You can use some examples for Moodle docs: http://docs.moodle.org/24/en/Cloze . It would be interesting that at least one of the created questions include subquestions with diffrents default marks to verify grading calculation is right in all cases Don't forget to add several hints to your question(s), for instance "This is the first hint", "This is second hint", ... To fully test this code it is necessary that the number of hint is not too low. The number of tries allowed in interactive with multiple tries is equal to the number of hints + 1. So if you create 3 hints you will have 4 tries allowed. Save and preview the question(s) setting "How questions behave" to "Interactive with multiple tries" As a first very easy test try to answer correctly to all subquestions but one on the first try and to all questions correctly on the second try. verify penalty is only applied to the subquestion that was wrong on the first try. Try to get different subquestions right at different tries (for instance 1 subquestion right at second try and another right at third try) and verify mark calculation is always OK Try the following combination on 2 subquestions a & b Try 1 : response a is right and response b is wrong Try 2 : response a is wrong and response b is right Try 3 : response a is right and response b is right And verify mark's calculation is done deducting 2 penalties for a and 1 for b Test all other combinations you can think of and verify calculation is done in all cases on a subquestion by subquestion basis and considering the last change of response on the subquestion