<"...Horowitz won from Reuben Fine in 27 moves. The latter went astray in a book variation of the Ruy Lopez, which proved to be inferior. Horowitz won his queen, and thereafter played carefully until victory rewarded his efforts."> -- New York Times, August 15, 1934.

Well, it's not like Horowitz won the queen, but let that pass.

<"Time pressure again affected several of the players last night. The sensation of the second round was the defeate of Reuben Fine by I. A. Horowitz, who recently lost a set match to the youthful Western champion. Fine, on the black side of a Ruy Lopez, drifted into a losing variation which, strangely enough, had been recommended by several book authorities.

"Fine, depending upon published analysis, found himself in hot water when Horowitz steered off from the beaten path. After losing his queen for two minor pieces, Fine made the best of a bad bargain, but resigned just as his clock was about to register his first two hours. It was Horowitz's second success in two days."> -- New York Sun, August 15, 1934

And it's not like Fine lost the queen, either. But proceeding to the opening.

Horowitz's <8.Ng5> was a newer line in this variation, and Fine's <8...d5> (now the main line) made its debut. The critical point appears to arise after <9.f3>:

Opening Explorer

Judging by the results in the Opening Explorer, we can now assess Fine's <9...e3> as clearly inferior to 9...h6. The latter move was first played iin M Shishov vs Nezhmetdinov, 1947 and disappeared until the 1970s, but once it came back the 9...e3 variation went out of favor.

Perhaps Horowitz's <10.f4> was the "nonbook" move that Fine overlooked. He probably spent time analyzing the straight pawn sacrifice after 10.Bxe3, and didn't anticipate such a paradoxical move.

What can we conclude? Fine was probably knocked off balance when his preparation went askew, so an hallucination (as <jerseybob> suggests) is possible. Or he may have just panicked, though his situation doesn't seem to be that bad as yet.