The number of input data bits directly affecting each output code bit in a convolutional encoder. The higher the constraint length, the more powerful the error-correcting code, but higher constraint lengths require more processing power to decode.