A binary number will have

**odd***if the total number of one-bits in it is***parity****odd**.__Example__: 1101 has 3 1-bits, an*odd*number. A binary number will have**even***if the total number of one-bits in it is***parity****even**.__Example__: 011 has 2 1-bits, an*even*number. As*an error-detecting measure*, many communication channels are configured to have odd (or even) parity. One of the bits in the message is a "parity bit" whose sole function is to take on a state that will assure the overall parity is odd (or even). If a received message has incorrect parity, then the receiving party knows some odd number of errors exist in the received message. If the channel is not very noisy, that odd number is usually 1. Additional information is required if the error is to be*corrected*. Usually the choice of even or odd parity is intended to force at least one change of state in any given message, so that a legitimate message will never consist of all 1-bits or all 0-bits. Most communication channels use word lengths that are even, so odd parity tends to be the norm. The concept of an error-detecting digit can be extended to non-binary numbering. Many account numbers, for example, have a "check digit" that can be used to detect if the digits of the number are mistyped or transposed.