According to Wikipedia, RAID 2 uses a [7,4,3]2 Hamming code to implement error correction. So, for a 7-disk RAID 2 setup, every 4 bit sequence of incoming data should be encoded in a 7 bit sequence and each bit placed on one of the 7 disks. (Is that correct? I couldn't actually find the RAID 2 specification.)
The [7,4,3]2 Hamming code is a 1-error-correcting code, so it has the neat property that if one of the disks is suddenly filled with garbage or disappears, the decoded data will still be correct. This seems like a great property to have! However, that Wikipedia article also claims that RAID 2 is rarely used. Why not?
Aside: (hidden so that it doesn't derail readers from my actual question above)
The Wikipedia article on RAID levels claims that "All hard disks eventually implemented Hamming code error correction. This made RAID 2 error correction redundant and unnecessarily complex.". While this may be true, error correction within a single disk won't handle a failure of the entire disk (for example, a circuit failure or motor failure). A Hamming code across multiple disks handles this situation, so the quoted argument seems to be invalid.