The error term in the equation defining the permanent income hypothesis is sometimes viewed in the literature as transitory consumption (generated by the agent), and is sometimes viewed as measurement error (generated by the observer). In both cases, it is usually assumed to be white noise. Sometimes it is ignored altogether, assumed to be identically zero. Using empirical evidence that combines both monthly and quarterly data, this paper shows that the error term in the permanent income equation cannot be identically zero, and that the presence of both transitory consumption and measurement errors can be consistent with the data. The evidence also confirms that the best model for this term is indeed a white noise process.