In the usage in telecommunication, the number of '1' bits in the binary value is counted. Parity is even if there are an even number of '1' bits, and odd otherwise. Examples: the parity of the value 10111101 is even (there are 6 '1' bits); the parity of the value 01110011 is odd (there are 5 '1' bits).
Parity is sometimes used for error checking due to the fact that it may be calculated easily, however much more robust error checking methods exist.
There are several types of parity: none, marking, even, and odd. 'None' means there is no parity calculated and a zero-bit is usually inserted (that is, the bit is present but unused or ignored). 'Marking' means that the parity bit is always a '1'. 'Even' and 'odd' parity insert '1' or '0' parity bits so that the total number of '1' is even or odd, including the parity bit. The parity bit is 'stripped off' before the data is used, thus a seven-bit character (or data value) requires eight bits to transmit or store - the seven data bits and the parity bit.
Modern error-checking algorithms use CRC or error correcting codes, for example. These codes are more powerful and can often correct errors, while parity can only detect some errors.
See also