On the GMAT you just have to be able to accept that when the question reads "students were surveyed and each song was selected 4 times" that however many students there were, the survey resulted in every song being selected 4 times total. If you surveyed 100,000,000 students, sure you could get more than 4 songs, but you have to accept as fact the statements made in the question stems.
We could attack this survey on all types of angles.
Were the students allowed to abstain if they hated all 12 songs?
If they must vote, were they instructed to pick the 2 songs they disliked the least?
Were the students responses a result of leading questions?
Was there any type on control group with which to compare these results in order to determine if any bias exists among those surveyed?
We can think of all types of things that would be relevant to a survey, but in the end, we have to live in the GMAT world and only respond to what information is given and in the method it is provided.
Here, after the survey is complete, each song was selected 4 times. Thus, we know that 12 songs * 4 selections = 48 total selections / number of selections allowed per student = 24 students participated in the survey.