A mathematical model was developed to predict signal attentuation and signal broadening in flow injection-hydride generation atomic fluorescence spectrometry owing to gas-liquid separation in order to optimise the performance with regard to sensitivity and throughput. Experimental data for arsenic measurements matched well with the model so that theoretical optimisation of various mutually dependent instrumental parameters (sample loop volume, carrier solution flow rate, GLS headspace and purge gas flow rate) was possible. A minimal GLS headspace and a maximal carrier solution flow rate resulted in the highest sensitivity and throughput for the system set-up under study.