Enumeration of Clostridium perfringens spores was compared using 4 culture media. Duplicate 1 g portions of 35 stools (25 from C. perfringens food poisoning outbreaks and 10 from normal stools) were heat treated 20 min at 75 degrees C and tested on tryptose-sulfite-cycloserine (TSC) agar, trypticase-soy-blood (TSB) agar, lactose-sulfite (LS) medium, and iron milk (IM) medium. Dilutions were plated directly onto TSB and TSC, and a 3-tube most probable number determination was made with each specimen in LS and IM incubated at 45 degrees C. TSB was easiest to use and nonhemolytic food poisoning strains were readily differentiated from the normal hemolytic biotype on this medium. Confirmed counts on TSC and TSB were similar for all specimens, but counts of 8 of 25 outbreak specimens were 2-4 log units lower in LS and IM than on plating media; spores in specimens associated with 2 of 5 outbreaks were intolerant of the elevated temperatures. Results showed that elevated temperature MPN methods in LS and IM are inappropriate for the examination of outbreak stools.