Help with perfcurve parameters!!

I want to use 'perfcurve' to generate Roc Curves for some classification data, but am unsure of the exact parameter requirements.

I am using LibSVM to classify three sets of data using labels, -1, 1 and 2. As stated by the documentation, 'perfcurve' calls the following: labels, scores, & posclass.

I know that 'labels' is the correct set of labels used to test the data, and mine is a vector sized '60x1'.

I assumed that 'scores' is the LibSVM 'svmpredict' output 'prob_estimates' but mine is a matrix sized '60x3' and the documentation states that 'scores' and 'labels' must be the same size? So what is 'scores'?

I understand that 'posclass' would be either -1, 1 or 2 but I don't understand which one I should pick?

Thanks in advance.