This observation is confirmed by curious experiments. Write some everyday
phrase or other on a blackboard; let there be a few intentional mistakes
here and there, a letter or two altered, or left out. Place the words in a
dark room in front of a person who, of course, does not know what has been
written. Then turn on the light without allowing the observer sufficient
time to spell the writing.

In spite of this, he will in most cases read the entire phrase, without
hesitation or difficulty.

He has restored what was missing, or corrected what was at fault.

Now, ask him what letters he is certain he saw, and you will find he will
tell you an omitted or altered letter as well as a letter actually written.

The observer then thinks he sees in broad light a letter which is not
there, if that letter, in virtue of the general sense, ought to appear in
the phrase. But you can go further, and vary the experiment.

Suppose we write the word "tumult" correctly. After doing so, to direct
the memory of the observer into a certain trend of recollection, call out
in his ear, during the short time the light is turned on, another word of
different meaning, for example, the word "railway."

The observer will read "tunnel"; that is to say, a word, the graphical
outline of which is like that of the written word, but connected in sense
with the order of recollection called up.

In this mistake in reading, as in the spontaneous correction of the
previous experiment, we see very clearly that perception is always the