We present box model calculations of ozone loss rates corresponding to the results of the Match experiment 1991/92. The Match technique infers chemical ozone depletion from an analysis of pairs of balloon soundings that measure ozone in the same airparcel at different points of a calculated trajectory. It allows a quantitative comparison with model results because the exposure of the observed airmasses to sunlight is well known. The model significantly underestimates the loss rates inferred for January to mid-February. Extensive sensitivity studies show that the discrepancy between model and Match results cannot be explained by the known uncertainties in the model parameters. ¿ 1998 American Geophysical Union