Recent accounts of Vickers hardness tests in which the variation of hardness values with indentation size was suppressed by lubrication are re-examined to elucidate the nature of the size effect. It is shown that a simple constant error in either the measured indentation size or the force actually exerted would not explain the effect. A good description of the effect is obtained by reference to an indentation-size-dependent load error which is sensitive to both friction conditions and the plastic properties of the test material.