A predictive model for pressure drop in a pulse-jet cleaned fabric filter and the effect of on-line vs. off-line pulse-jet cleaning on pressure drop were examined. Data were collected in experimentswith untreated polyester fabric and PTFE-laminated fabric. Twenty-four experimental conditions were studied in which each fabric repeatedly filtered limestone dust or flyash for 30, 120, or 480 s and was then cleaned either on-line or off-line. Each condition was replicated, resulting in a total of 48 experiments. Dust inlet concentration and superficial filtration velocity were maintained at 1.69 g m[-3] and 0.075 m s[-1] respectively. Based on conditions studied: 1) dust removal results from a complex interaction of the pulse pressure, pressure drop prior to pulsing, and dust areal density, w; 2) dust removal efficiency of a cleaning pulse, ε, increases as w increases, 3) the rate of increase in ε with w diminishes as w becomes larger; 4) the predictive model examined should be applied only if the dust areal density added during one filtration cycle, W[o], remains unchanged or if the rate of change in ε is small; 5) under some conditions off-line cleaning does not reduce pressure drop below that achieved by on-line cleaning; and 6) pressure drop depends primarily on fabric type and w[o].