A theory linking barrel fill length and air bubble entrapment during extrusion of starch has been proposed. Experimental data suggested that when barrel fill length (defined as length of barrel filled with material pressurized to at least 689 kPa) was greater than the melting region-to-die barrel distance, then air bubble entrapment (as measured by porosity) in the unexpanded extrudate was favored; if the reverse was true, then air bubble entrapment was hampered. Bubble number density in the unexpanded extrudate was not only a function of the total entrapped air (porosity) but also of the type of starch used (pregelatinized versus native amylopectin) where pregelatinized amylopectin gave much higher bubble number density values than the native counterpart. Screw speed increased bubble number densities at constant entrapped air, which seemed to suggest bubble break-up by increased shear of the rotating screws. © 2001 Elsevier Science Ltd. All rights reserved.