This paper employs bootstrapping to correct for bias and to construct confidence intervals for Malmquist TFP indices derived with DEA. It uses these results to investigate the productivity change in Polish agriculture during a crucial period of the country's transition to a market economy, 1996-2000, when Poland was preparing for accession to the European Union. The bias corrected estimates show regress in productivity at an annual rate of 4 percent. The confidence intervals suggest that between two-thirds and four-fifths of the sample farms (250) in different years might have experienced no change in productivity. The cluster analysis based on confidence bounds reveals three paths of productivity change. Farms which recorded an increase in productivity at least in the last year of the analysed period, are larger, more capital intensive, run by younger farmers, and more integrated in factor and product markets. However, they account for only 19 percent of the sample farms. The most important for Poland now is to unlock the forces that can drive ahead structural reform and thus productivity growth.