INTRODUCTION: This study examines the accuracy of using membrane models to predict activation thresholds for chick heart cells during field stimulation. METHODS AND RESULTS: Activation thresholds were measured experimentally in ten embryonic chick heart cells at 37 degrees C for stimulus durations 0.2 to 40 msec. Activation was assessed by observing the mechanical twitch of the cell. The heart cells ranged in diameter from 15.0 to 26.7 microm. Since the electric field required for activation depends on diameter, the thresholds were expressed as the maximum field-induced transmembrane potential, Vth = 1.5 a Eth, where a is the cell radius and Eth is the strength of the electric field at threshold. A cell model was created using a singular perturbation method and membrane models describing the ionic currents of a heart cell. The study used membrane models of Ebihara and Johnson (1980), Luo and Rudy (1991), Shrier and Clay (1994), and their combinations. The results show that for stimuli longer than 1 msec, theoretical activation thresholds were within one standard deviation of experimental thresholds. For shorter stimuli, the models failed to predict thresholds because of a premature deactivation of the sodium current. The modification of the m gates dynamics, so that they closed with a time constant of 1.4 msec, allowed to predict thresholds for all durations. The root mean square error between experimental and theoretical thresholds was 6.14%. CONCLUSIONS: The existing membrane models can predict thresholds for field stimulation only for stimuli longer than 1 msec. For shorter stimuli, the models need a more accurate representation of the sodium tail current.