Kruskal observed that this same phenomenon seems

to occur with any sufficiently large piece of text—count-

ing forward in this way from any choice of beginning

word lands you at the same place at the end of the page.

This provides an amusing activity for several people to

perform simultaneously, all working with the same text,

but starting with different choices of initial word.

The phenomenon can be explained as follows:

imagine that on a first run through the text, all the

words encountered are circled. Landing on any one of

these circled words in a subsequent run of the experi-

ment will take the player to the same final word. What

then is the likelihood that a second run through the

experiment will not “hit” any of the previously circled

words? If the passage is sufficiently large, say involving

20 steps, the chances of “missing” a circled word every

time is likely to be very small, especially since the typi-

cal English passage contains reasonably short words.

Estimates computing this

PROBABILITY

show that it

likely has a value smaller than 0.1 percent. Thus, with

about 99.9 percent certainty, two paths through a pas-

sage of text will coincide at some point.

298 Kruskal’s count