r remarkable; they were probably unique. It is exceedingly
likely that there is not in the entire world another similar pair of
spectacles. Whence the finding of these fragments of glass in the
bedroom establishes a considerable probability that Mr. Weiss was, at
some time, in the chambers at New Inn.

"And now let us gather up the threads of this part of the argument. We
are inquiring into the identity of the man Weiss. Who was he?

"In the first place, we find him committing a secret crime from which
John Blackmore alone will benefit. This suggests the <i>prima-facie</i>
probability that he was John Blackmore.

"Then we find that he was a man of normal eyesight who was wearing
spectacles for the purpose of disguise. But the tenant of New Inn, whom
we have seen to be, almost certainly, John Blackmore--and whom we will,
for the present, assume to have been John Blackmore--was a man with
normal eyesight who wore spectacles for disguise.

"John Blackmore did not reside at New Inn, but at some place within
easy reach of it. But Weiss resided at a place within easy reach of New
Inn.

"John Blackmore must have had possession and control of the person of
Jeffrey. But Weiss had possession and control of the person of Jeffrey.

"Weiss wore spectacles of a certain peculiar and probably unique
character. But portions of such spectacles were found in the chambers at
New Inn.

"The overwhelming probability, therefore, is that Weiss and the tenant
of New Inn were one and the same person; and that that person was John
Blackmore."

"That," said Mr. Winwood, "is a very plausible argument. But, you
observe, sir, that it contains an undistributed middle term."

Thorndyke smiled genially. I think he forgave Winwood everything for
that remark.

"You are quite right, sir," he said. "It does. And, for that reason, the
demonstration is not absolute. But we must not forget, what logicians
seem occasionally to overlook: that the 'undistributed middle,' while it
interferes with absolute proof, may be quite consistent with a degree of
probability that approaches very near to certainty. Both the Bertillon
system and the English fingerprint system involve a process of reasoning
in which the middle term is undistributed. But the great probabilities
are accepted in practice as equivalent to certainties."

Mr. Winwood grunted a grudging assent, and Thorndyke resumed:

"We have now furnished fairly conclusive evidence on three heads: we