This paper presents practical experiences and results we obtained while working with simulators for artificial neural network, i.e. a comparison of the simulators' functionality and performance is described. The selected simulators are free of charge for research and education. The simulators in test were: (a) PlaNet, Version 5.6 from the University of Colorado at Boulder, USA, (b) Pygmalion, Version 2.0, from the Computer Science Department of the University College London, Great Britain, (c) the Rochester Connectionist Simulator (RCS), Version 4.2 from the University of Rochester, NY, USA and (d) the SNNS (Stuttgart Neural Net Simulator), Versions 1.3 and 2.0 from the University of Stuttgart, Germany. The functionality test focusses on special features concerning the establishment and training of connectionist networks as well as facilities of their application. By exemplarily evaluating the simulators' performance, we attempted to establish one and the same type of back-propagation network for optical character recognition (OCR). A respective quality statement is made by comparing the number of cycles needed for training and the recognition rate of the individual simulators.