Generally, dividing by n-1 is used to give an unbiased estimation of the population variance. If you're using a sample from some population N, your sample size can be, let's say n. This sample is generally random, but nevertheless is contained in N. Say the true mean (I'll call it U) is at some point. Depending on what sample you take, your sample mean (U') could sit either really close to U, or really far from U - i.e. U could be outside of the n. If we consider this fact, by dividing by n we don't truly consider the implications of the former, i.e. we underestimate the variance of the population. When dividing by n-1 we get a larger variance, and hence a better estimate from n to infer onto N.
I think that makes sense? I would imagine in VCE, yes you would divide my n-1. I do however doubt you'll be expected to provide rigorous meaning behind your choice of choosing n-1 over n.