When researchers collect and process spontaneous brand awareness data, it can be captured and coded into different variables. Each of the variables represents the 1st, 2nd, 3rd, 4th mention and so on. Because the variables each have (or should have) the same code-frame (ie: value labels), then they can be combined into a single categorical array.
Users of the research, however, are not so interested in what is the 2nd, 3rd and 4th mention. They are typically interested in "All Mentions". That is, what is the net mentions of a brand across all the variable, regardless of the position. The result of this netting is a multiple response variable, which represents the count for each brand (one brand per subvariable).
It is not recommended to do this in R - this is much better handled by Crunch Automation - see: Transforming multiple selections (fixed column). You can run the Crunch Automation from within R, see: Running Crunch Automation within R.