Participants lifted pairs of successively presented weights and compared them for heaviness, using the constant method with 2, 3, or 6 judgment categories. The standard weight (St) was 100, 200, or 300 g, either roving or fixed within a block. For each St, there were 5 comparison (Co) weights. The lifting orders were St-Co and, with 6 categories, Co-St. Time-order errors were negatively related to St magnitude, particularly with roving St. In terms of Hellström's sensation-weighting theory, this result was accounted for by a smaller weighting coefficient for the first-presented stimulus than for the second. Time-order errors were negative on average, which was explained as the result of this weighting in conjunction with a low position of the reference level because of light background heaviness. With roving St, the dispersion of the subjective intrapair difference increased with St magnitude, providing evidence for Ekman's law (G. Ekman, 1956, 1959).