A common assumption in manifold learning is that a vector data set lies on or near a manifold. This talk will illustrate experimental use of several multi-scale techniques on real world data sets for the purpose of identifying their multi-manifold structure and distinguishing real world multi-manifolds and time series graphs. In addition, we will present a theorem that guarantees that a large class of measures on the unit circle can be visualized as Jordan curves. This work was done in collaboration with Devasis Bassu, Rauf Izmailov, Peter W. Jones, Allen McIntosh, David Shallcross, Patricia Medina, Melanie Weber, and Kara Yacoubou-Djima. The work was partially supported by grants from ONR and AFOSR.