Jasso (1978) proposed a universal law of justice evaluations describing a logarithmic relationship between the perceived injustice of a reward and the ratio between this reward and the just reward. In applications this model is treated as if it were exact, whereas analogous models in psychophysics have empirically established degrees of uncertainty. In this article I make the first assessment of the magnitude of error in the logarithmic model of justice evaluations, using published data and a novel experiment. For the standard application of the model, where just rewards are inferred from justice evaluations, I find that the inherent inaccuracy leads to errors of about 15% on average. I also compared the logarithmic model to 2 nonlogarithmic models. Almost 20% of my respondents made justice evaluations that were more consistent with one of the latter models, suggesting that no single model is really universal.