: Or thin cap strips with a layer of glass...

This may be tempting, but does one not need some thickness to get the oval final shape?

: I think I'll do four cores - one with all 1/4" strips and poly glue. One
: with 3/8" x 3/4" core strips, 1/4" cap strips, and poly
: glue. One with 7/8" x 3/8" core strips, 1/8" caps w/4oz
: glass inside the caps only - and one with 7/8 x 1/4" core strips, and
: 1/8" caps w/4oz...

When you make test cores, always make them the same length. Also, get a little black book and write down all the dimensions and wood data, and results. Then you can add new tests years from now without needing to repeat anything you've already done before.

: I think that it's reasonable to assume that the blades themselves add roughly
: the same strength to the finished paddle, no matter what the core
: construction. Therfore, simlply putting the first 1/4" blade strips
: on should be a sufficient imitation of the 'core structure' including the
: strength that the blades add to it.

I'd assume that the loom is the weakest area, and just disregard any blade strips, since they will only add strength, some more, others less.

: If I laid up these four cores, and set them up on a lever at the midpoint -
: representing an extended paddle brace, which commom sense tells me should
: put the most stress on a paddle - after all, I did break one that way once
: - then adding weight to the end until it fails oughta give me a reasonable
: indication of the relative strength of each technique, wouldnt you say?

I'd support the cores at each end, always the same distance apart, and hang a bucket at the middle, and slowly fill it until something happens. I'd make the distance greater than a typical loom length, then it would break sooner. That would also give a sort of guaranteed to survive value for the shorter loom on the actual paddle. This is in fact the same test you described, but easier to implement. Only the force pulls down instead of pushing up, just like in a C-to-C roll.

: Especially if I made all four cores out of the same board - eliminating
: material differences.

You can also test for different materials, later on, by freezing the dimensions and structure instead. Both can be easily done. But not everything the first time you do this, naturally.

: Whats a good way to add weight? A bucket of water?

I'd add sand or gravel or pebbles - the last would be the cleanest, no dust. Weigh the bucket and pebbles afterwards, or maybe also at intervals.