Reference request: embedded Morse theory

I found the treatment in Goresky-MacPherson's book "Stratified Morse theory" (available from Goresky's webpage here) very enlightening. The focus is not on low-dimensional topology, but the treatment is very geometric.

The main set of ideas that you want to learn is the following description of an m-dimensional manifold Y sitting in $R^n$, in such a way that the standard height function is a Morse function when restricted to Y. Thus this function, say f, gives a handle decomposition of Y; as you pass a critical point of f|Y, you add an index k handle to Y. Simultaneously, you add a (k + n -m -1)-handle to the complement of Y. This is described in reasonable detail in Section 6.2 of the book of Gompf and Stipsicz, "An Introduction to 4-manifolds and Kirby Calculus".

I don't know the original source for this description; I learned it more or less as folklore. The informal explanation that Kirby used to give of this involved sitting in a bathtub (I think the person in the bathtub was supposed to be Y) and watching the topology of the water change as it passed various critical points. (An alternate version, helpful for thinking about knots, was a wire in a bucket of water being filled up.) A nice example to think about is how to build a handle decomposition (or Heegaard splitting) of a knot complement. For instance, for a 2-bridge knot, you should see a handle decomposition with a 0-handle, two 1-handles, and a single 2-handle.

As a side remark, the term relative Morse theory, as I understood it, has to do with the study of the Morse function on a manifold Y induced by a Morse function on a larger manifold. In this form, it was extensively studied (in the PL case) in the 60's, in order to give results on embeddings. For instance, various theorems of the form "concordance implies isotopy" in high codimension are proved in this way.