This paper investigates the performance of some state of the art commercial speech recognisers that support continuous speech, speaker independent recognition. Three speech recognition engines were tested using small size vocabulary, continuous speech, and under a variety of conditions. These tests were originally performed in order to select the most suitable speech recogniser for a specific application that runs on a personal computer. Nevertheless, the requirements and operating conditions of this application are common to many other speech recognition applications. Five parameters were considered critical for the performance of speech recognition in the intended application, namely: dialect, speaker vocal characteristics, microphone type, noise level, and loudness level. The effect of each of these parameters on the recognition accuracy was evaluated by a set of tests in which one parameter was varied and the recognition accuracy of the various speech recognition systems measured.