lexipixel - 7:13 am on Aug 31, 2011 (gmt 0)
I've seen more and more examples that are indecipherable
We may be down to slim pickin's on the official "reCAPTCHA" images as many of the easier to read ones have probably been solved.
reCAPTCHA uses two-words, and serves a different purpose than the simple "type this word" CAPTCHA's that were used and duplicated everywhere (and cracked) when the idea first came out.
"reCAPTCHA" is used to help improve OCR results for book scanning.
You don't actually have to get both words correct.
The "word" on the left is some indecypherable OCR scan that Google Books of other entity needs help decyphering, the word on the right is a known word (and is usually easier to read).
The idea is that if they show the same reCAPTCHA to enough people and a large percentage say the word on the left is "EXAMPLE" and those same people correctly retype the word on the right (which varies from test to test), then they can safely assume the word on the left is "EXAMPLE" (even though the script producing the two-word reCAPTCHA never knew what the word on the left was).
For this reason you could type anything for the word on the left, and only need to match the word on the right -- and still pass the CAPTCHA test.
Try it next time you get one of the unreadable (left) reCAPTCHA words.
Google reCAPTCHA info: [google.com...]