A new program, Fungal Identification System (FIS) was introduced in POISINDEX (PI) in 1991. We studied the accuracy of this program compared to the traditional method of consulting an expert mycologist and also the utility of FIS for the certified specialist in poison information (CSPI) and the emergency department personnel. Fifteen samples were obtained from a Chicago area forest preserve and immediately identified by a CSPI using FIS from PI of summer-fall 1991. The samples were then refrigerated overnight and identified the following morning by a PhD mycologist, using standard mycological identification techniques (SMIT). Another group of 18 mushrooms were collected in August 1992 from rural southwestern Michigan and identified by FIS and SMIT after refrigeration overnight. In 4 cases the mycologist used both SMIT and FIS to identify the mushrooms. Identification by FIS did not match SMIT in 27 of 31 samples (1 was discarded and 1 was identified by FIS but not identified by SMIT). In 2 of 4 cases where the mycologist used FIS, a different answer was obtained from that obtained by the CSPI; however, these were still incorrect identifications. Four types of problems using FIS have been identified which can lead to erroneous results; these are the use of multipart questions, emphasis on color over less variable characteristics, questions which do not account for variability in morphology, and problems which arise from the sequence of questions. These points need to be addressed before FIS can be used as a safe basis for clinical recommendations.