(1) Given a fundamental group representation of a hyperbolic surface, i.e. $<a_j,b_j|\prod[a_j,b_j]=1>$, and given an element in this group, can we determine whether this element can be represented by a simple closed curve?

(2) More specifically, if we consider only a commutative element, whose abelizaions is trivial in the homology group, is there some method to determine whether this element can be represent by a simple closed separating curve on the surface?

For example, $a_1a_2a_1^{-1}a_2^{-1}$ cannot be represented by a simple closed separating curve. I guess the geometric interpretation of a simple closed separating curve is that it is the boundary of the neibourhood of a group of chained curves. Here a group of chained curves means a set of curves $\{c_1, \dots ,c_{2h}\}$ satisfying the following conditions: the geometric intersection numbers are $i(c_j,c_{j+1})=1$ and $i(c_j,c_{k})=0$ for $k-j>1$. But even this criterion is true, the choice of $c_j$'s in the fundamental group are diverse. From an element in the fundamental group, we may not easily see that whether it is the boundary of a group of chained curves. Are there some easy way or algorithm to determine it?

The previous discussion is for question (2). What if we consider the elements in question (1)? I learn from Farb and Margalit's book that there is a neccesary condition on the homology group that the abelization element of a simple closed curve in the homology group must be primitive. What about the representation in the fundamental group?