Marginal Distribution vs Conditional Distribution

It collapses the table down to a single variable, throwing away the information about the other one, which is why it is read off the margins. Every entry uses the grand total as its denominator, unlike a conditional distribution, which divides by one row or column total instead. Of 200 surveyed students, 80 are seniors and 120 are not, so the marginal distribution of class is $80/200 = 0.40$ senior and $120/200 = 0.60$ not senior. The proportions in a marginal distribution add to 1 across all categories of that one variable.

You restrict attention to one group, then re-divide by that group's total rather than the grand total, so the percentages inside the group add to 1. That re-dividing is the whole difference from a marginal distribution, which uses the grand total and describes everybody. Of 200 surveyed students, 80 are seniors and 45 of those seniors drive to school, so the conditional distribution of driving given senior is $45/80 = 0.5625$ driving and $35/80 = 0.4375$ not driving. Comparing the conditional distributions across rows is how you judge whether the two variables are associated.