Drug interactions with cardiac ion channels can disturb normal electrical activity in the heart which can result in the occurrence of fatal arrhythmias. hERG channel block has been linked with increased pro-arrhythmic risk. Consequently, screening of hERG, to quantify the amount of drug-induced block at different drug concentrations, forms a routine part of cardiovascular safety assessment. Recent initiatives led by the US Food and Drug Administration (Sager et al., 2014) are encouraging the use of computational approaches within standard cardiovascular safety assessment. It is important that hERG channel currents are accurately represented within mathematical models used for this purpose.A number of existing models have been proposed to describe hERG channel kinetics (Bett et al., 2011). These models have typically been constructed through fitting to patch clamp data acquired following application of a series of voltage-step protocols. Each of these models are able to replicate behaviour observed in experiment when subjected to voltage-step protocols (similar to those from which they were parameterised). However, different models can exhibit quite different behaviour under other protocols; including physiologically relevant action potential clamps or clamps mimicking arrhythmogenic behaviour.Here we present novel voltage protocols, designed to determine a robust model of hERG channel kinetics. The protocols were designed to maximise the difference between simulated currents from different possible hERG channel models. This identifies those models which are able to replicate behaviour observed in experiment, and the model features required to achieve this. The protocols have been performed in manual patch clamp experiments, using hERG-1a transfected Human Embryonic Kidney (HEK) 293 cells. The transition rates between distinct states represented within the model have been characterised using Bayesian fitting techniques.This improved representation of hERG channel kinetics can be incorporated within mathematical cardiac electrophysiological action potential models. This will potentially enhance their ability to describe normal electrical activity in cardiac cells. In future work, this hERG channel model will be used as a basis for describing hERG channel kinetics in the presence of drug compounds. An improved representation of drug-hERG channel interactions is likely to be important in predicting drug-induced pro-arrhythmic risk.