Dead-end stirred cell devices are commonly used in laboratories to characterise ultrafiltration membranes and their separation behaviour. Additionally, protein separation data from such systems are used for process scale-up. Such devices are operated under conditions that are inherently different from those used during the continuous or semi-continuous processing of industrial feed streams. The work presented in this paper compares the rejection behaviour of single protein solutions in a dead-end stirred cell (SC) device with that for a cross-flow system (CF). The effect of ionic strength (20 and 100 mM) and solution pH (4.9, 6.0, 7.1, 8.4 and 11.0) on protein filtration (bovine serum albumin (BSA) and lysozyme (LYZ) from buffered aqueous solutions) behaviour has been investigated using polyethersulfone (PES) membranes with a manufacturer specified molecular weight cut-off (MWCO) of 50 kDa. PES membranes were characterised in terms of dextran MWCO using both the SC and the CF systems. The mode of operation resulted in significant observed differences in the resulting dextran solute rejection curves for the two systems. The observed rejection (R-obs) values for a series of dextran standards were consistently found to be lower for the CF system compared with the SC unit suggesting higher wall concentrations (C-w) due to concentration polarisation effects in the CF unit. Protein ultrafiltrafion studies with the 50 kDa PES membranes highlighted important differences in observed protein rejection behaviour despite operation of the two systems at the same transmembrane pressures (25 kPa). Solution pH was found to have little effect on the rejection of both BSA and LYZ. The solute rejection was found to be more sensitive to ionic strength effects for the SC device both during BSA and LYZ filtration. Convective mass transfer coefficients and hence the true rejection coefficients (R-tr) were calculated for both systems using the stagnant film model to understand the influence of hydrodynamic effects on the ultrafiltration behaviour of the two systems. The magnitude of the membrane Peclet number (Pe) provides a means of comparing hydrodynamic conditions for the two systems and thereby allows differences in observed solute rejection to be better understood. (C) 2008 Elsevier B.V. All rights reserved.