The trick was now practically done.  As the slates had been examined so many times and nothing found on them, even after the automatic writing, the majority of persons would testify that there was positively nothing on the slates when the medium left the table.  The majority of persons would never remember that he at one time wrote on the large slate and erased it.  The message being on a small slate, and these being spread around, few would have known that this message really appeared on the particular small slate that was originally next the top of the stack.

Most people would have certified that they cleaned all of the slates themselves, that the medium never touched any of the small ones, and that he only laid his hands on top of the stack a few times.  Some would even forget that the medium handled their writing at all before burning it.

I am sure that the nickeled tube that carried the dripping water into the space over the glass bowl, had a second tube within it; through which his assistant from the adjoining room either blew, or sent by some mechanism, the chemicals (probably potassium) that would take fire and burn on striking the water.

. . . . .

When I perform the slate trick described above, after writing the “automatic” message, apparently erasing it, and replacing the slates, I do not scatter the slates around on the table as this medium did.  Instead, I proceed as I will now describe.

We place our palms on the stack, and after a time examine the large slate for a message, but find none.  I may incidentally remark that this last examination unconsciously verifies in the sitter’s mind the fact that I actually erased what I wrote “automatically.”

I now look on some of the smaller slates for a message, but find none.  When I do this I do not turn these slates over and look on their under sides, but merely take off the top slate to see if there be a message on the upper surface of the one under it.  I merely remark, “Well, there is nothing on that slate,” indicating the second one from the top; and at the same time I drop the top slate (now in my hand) on the table beside the stack.  I immediately take off the second slate and repeat this same performance, dropping it on top of the first one.  I keep on with this performance until I have removed four or five of the slates, and have them stacked in a second stack beside the first one.  Then seeming to grow discouraged, I remark, “I guess there is no message”; and I replace the second stack on the first stack.  This places the message slate four or five slates down in the stack; as the bottom slate of the second stack, being the top slate of the original stack, is now the message slate.