For a triplet competitive-consecutive halogenation sequence forming mono; di- and trihalogenated products of the form, A + B --> R + B --> S + B --> T, under semibatch operation adding B to A, if perfect mixing could be assumed at all scales, the product distribution would be unchanged on scaling up. However, if the reaction rates are reasonably faster than the mixing rate, the semibatch addition of B to A will be imperfectly backmixed exhibiting macroscale concentration gradients. This partial segregation of the primary reagents is capable of modifying the selectivity and corresponding appearance of R, S and T in the course of the batch. Imperfect mixing is quantified using the networks-of-zones model. The effect of scaling up at equal tip speed is examined for a lab-scale 0.3-dm(3) reactor, a semitech 30-dm(3) reactor and a production-scale 3,000-dm(3) vessel. The intensity of partial segregation is weak at the lab scale, but very severe at the production scale. The lab-scale reactor is therefore close to perfectly backmixed and the primary, secondary and tertiary products appear in sequence. At the semitech scale the increased partial segregation causes the final product to initially precede the secondary product paradoxically but lag the initial product. At the large scale the move severe segregation between A and B gives an even greater paradox, whereby the final product appears ahead of both the primary and secondary ones. The segregated concentration fields of A and B are visualized as sectional image reconstructions for networks comprising on the order of 1,000 zones. Localized intensive plumes of B emanating from the addition point cause the paradoxical reversals of product sequences. The calculations are directly relevant to real industrial miscible liquid halogenations for which product distribution paradoxes have been observed (Haywood, 1990).