Most studies on the neuronal encoding of sensory stimuli have been performed based on spike responses. The classically considered neuronal response features like spike count, spike timing or ISIs rely on the all-or-nothing characteristic of spike responses. However, little is known about stimulus encoding on the level of integrated postsynaptic responses. In particular, no standard methods exist to analyze stimulus encoding with graded potentials. Moreover, the huge amount of stimulus-independent fluctuations present in intracellularly recorded membrane potentials complicates the analysis of graded responses. In this study, we use interneurons of the medical leech to analyze stimulus encoding based on graded membrane potential fluctuations. The local bend network controls the reflex behavior of the leech bending away when the skin is touched lightly [1]. It consists of four sensory P (‘pressure’) cells, approximately 20 interneurons and four groups of motorneurons, all of which can be identified individually. EPSPs induced by P cell spikes can be recorded intracellularly from the soma of interneurons, where spike amplitudes are small (350ms) and reached lower reconstruction performance (~50% correct). Using the EPSP slope yielded similar reconstruction performances and optimal time scales, but turned out to be more vulnerable to noisy data. Estimation based on maximum EPSP amplitude led to considerably lower performances.