Many thanks to y'all for your suggestions.
Heres a mini-summary of what I'm thinking so far.
I'll test both water and wine - people can do the water first just in case
I dont see a way to really keep secret what is being tested, and unless I can do so perfectly its not a useful control.
I will test the claims on Roger's website as far as possible. I considered seriously the surface tension experiment, but its somewhat of a pain with respect to equipment and a quickish protocol.
It is hard to know about the question of carrying in glasses from a separate room, versus hidden/similar coasters etc under flasks in plain view. Especially since there seem to be no studies on the half-life of magnetised water or wine! The one set of instructions suggests you keep it in the fridge, and presumably you dont have to drink it while enclosed in the fridge yourself. The advantage of not having the jugs of liquid under hidden coasters in the same room as the testees, is that they then will not know whether their liquid came from the same jug or not. So I am leaning towards a separate room, with rapid dissemination to the testing area.
For the wine I think I will pour a couple of bottles of a medium quality Aussie wine into the one flask, then split them just before magnetisation.
I have no pets, nor do I have ready access to someone who believes wholeheartedly there will be a difference. The one undergraduate mentoree I have who believes in homeopathy etc says he wouldnt be surprised if it worked, but didnt feel strongly about it. He unfortunately has his third year comprehensive exams week after next, and will not be amenable to being part of this test.
Roger sent me the coasters so I could try them myself before rashly taking on such a bet, I've decided to go a little bit further than testing them on my morning cup of coffee (which I have been doing with varying degrees of success...) but I realise that the test will not necessarliy satisy hardcore naysayers.
So heres an overview of the specifics:
I imagine producing a string of randomly ordered triples like this:
1 M
2 N2
3 N1
4 N1
5 N2
6 M
7 N2
8 M
9 N1
etc
where M=magnetised, N1,N2 mean "not magnetised 1 and 2"
Then I imagine labelling glasses (or plastic cups in the case of water) with a number. One student will be conscripted to pour the liquid according to the above schema. A second student will retrieve them from the room, the testees each take their three glasses and try to find the odd man out, which I really think is the simplest test as to whether there is any effect whatsoever. "Better" is too subjective. If there is no difference, as I suspect, then it cannot be better!
I will try and have some folks there who understand wine tasting, but since I dont know many people here yet, plebian physics grad students are my primary resource!
To motivate folks (well, students in particular) I am thinking I will offer a case of beer or a good bottle of wine to any person who CAN find a difference in 4 of an "odd man out of three" tastings. i.e. anyone who gets it correct for both water and wine can have another go at each and if theyre correct again both times, they win.
After all this we head for the pub (no, the testing will not be done in the pub!)
Any further thoughts?